← All TWiT.tv Shows (Audio)
All TWiT.tv Shows (Audio)
Intelligent Machines 843: Immortal Beloved, You've Arrived
03:02:04
AI's Emotional Intelligence Paradox
Loading summary
Transcript1,428 lines
- [00:01]
A
It's time for Intelligent Machines. Paris and Jeff are here. Our guest, Dr. Alan Cowan, is the Chief Scientist at Hume AI. He is an expert in human emotion and he thinks AI needs to be responsive to our emotions. We'll talk about that and all the AI news next on Intelligent Machines. Podcasts you love from people you trust. This is Twit. This is Intelligent Machines with Paris Martineau and Jeff Jarvis. Episode 843, recorded Wednesday, October 29, 2025. Immortal beloved, you've arrived. It's time for Intelligent Machines, the show where we talk about the latest in AI robotics. And all those doohickeys surrounding us, the smart little things, loading the dishwasher, folding the clothes. No, they're not doing any of that. Walking the dog. Paris Martineau is here from Consumer Reports. The queen of lead in your protein powder. Radioactive.
- [01:09]
B
The queen of. Every week being more and more baffled by how you choose to introduce the show.
- [01:13]
A
More like every week it's different. I'm trying to mix it up and.
- [01:17]
B
It'S still the same.
- [01:18]
A
And yet the same. Yeah, that's. That's kind of, in a nutshell, the story of my life. I just keep doing the same thing over and over again. Hello, Paris.
- [01:27]
B
Hi, Leo.
- [01:28]
A
You were so good on our DD adventure, we have decided to do it again. Finish the corn maze.
- [01:34]
B
It's great. I love playing D and D. You.
- [01:37]
A
Know, you converted me. I was the only one in the group who had never played. And I admit that culturally I had some issues at first, but eventually I think I got it.
- [01:47]
B
Yeah. I will say, if you guys listen to the. The stream of it, which is available for Club Twit members, you will hear just random sound effects playing all throughout the entire thing, seemingly with no rhyme or reason. And that's because Leo got a little bored. That actually found his way. That's.
- [02:06]
A
That's not fair. I. I prepared them ahead of time. I was ready, ready to launch them at the moment's notice. I even wrote a theme song. And, well, Suno AI wrote a theme song. Also here, Jeff Jarvis, emeritus professor of journalistic innovation at the Craig Newmark Graduate School of Journalism at the City and University of New York. Almost got Craig Newmar. Newmark should make a deal. Benito, our producer, if I get to the last word, City University of New York. Before you hit the button, that you don't hit the button.
- [02:37]
C
Oh, that's a challenge.
- [02:38]
B
I will say, as the person who's stumping for the Craig theme, I think that's a very reasonable.
- [02:43]
A
It's a Race.
- [02:44]
B
Listen, it's a race. Benito's gotta be on it. We don't get it if he's not on the butt.
- [02:49]
A
Exactly. Benito looks the other way. I want to get him.
- [02:53]
C
Come on, man. I'm counting on you. Craig's counting on you. All right.
- [02:56]
A
More importantly, Jeff is a author of the Gutenberg Parenthesis and magazine, and of course, is also a adjunct professor at Montclair State University and suni Stony Brook. I made you an adjunct. I don't know if you're.
- [03:10]
C
I'm a fellow and a senior.
- [03:11]
A
Feller.
- [03:12]
C
Whatever.
- [03:12]
A
He's a feller.
- [03:13]
C
I'm a fella.
- [03:15]
A
Hey, we have a really interesting guest. I'm not sure who nominated Alan Cowan for the show, but I'm really pleased to get him on. He's the founder of something called Hume AI. Alan, welcome to the show. Looks like you're in a soundproof booth preparing for a game show of some kind. Welcome. Good to see you.
- [03:35]
D
Good to see you. Thanks for having me. Yeah. This is our call booth in our office where I spend 35% of my time.
- [03:41]
A
Oh, you know what? You can make it nicer and more comfortable if you're gonna spend all of that time there. Look at. I don't know.
- [03:47]
B
I think it's kind of nice to be in a womb like state. Most of the hours of your work day could be worse.
- [03:52]
A
I gotta read your bio from your website, alan.cowen.com. i am an emotion scientist, an AI engineer. Okay. That is the greatest start, by the way, leading a lab and technology company called Hume AI. H U M E. I guess it's kind of short for human. And a former researcher at Berkeley and Google. Lately, I've been focused on the development of AI that can both think and speak and its optimization. And this is the little twist for human well being. Ooh. My past research included computational methods to address how emotional behaviors can be parameterized, predicted, and emulated. How they influence our social interactions and bring meaning to our lives. One of your studies showed sad or played sad songs for Americans and Chinese people and noted the emotional difference in the reactions. Which sounds fascinating.
- [04:50]
D
You've done your research.
- [04:51]
A
Did the Americans tear up?
- [04:56]
D
Nobody teared up. But we have made people tear up before. Yeah, that usually requires more than just music.
- [05:05]
A
I don't know. Every time I hear Luther Vandross sing a I want to dance with my father again, I get a little choked up. If you'd played that for me, I might have. Were the Chinese people more stoic, I.
- [05:17]
D
Would say they, you know, People recognize what music is trying to do most of the time, whether they feel it. You know, that varies from person to person. I think, you know, your particular nostalgia might not be shared by the average citizen of China, but maybe some people.
- [05:35]
A
That's a good point. They don't even know Luther Vandross probably. Well, okay, so that's interesting because what you're saying in effect is that humans, aware. Are aware of that the music is trying to manipulate them and resist it a little bit. I imagine that crosses over into your AI work. We're kind of aware that we're talking to AI and maybe have a different approach to it, which may be why companies like OpenAI make it so sycophantic. How did your emotion research apply to what you're doing with AI?
- [06:10]
D
So people, you're right. People are aware when they see an expressive medium, whether it's art or humans expressing things to them, that there's just an array of emotional affordances that those things are trying to play upon. Which means like people are either trying to make you happy or get something out of you by appealing to your sensibilities. And AI is an intelligent force that also is optimized for an objective and learns to interact with people in order to produce that objective. So in this, you know, it, it just doesn't, doesn't have feelings on it of its own. That's like the main difference. But, but when you think about how to optimize AI, you're really thinking about what emotional affordances are there for AI to improve human well being and how can we optimize for those affordances.
- [07:06]
A
It's kind of timely. This week, Sam Altman admitted that millions of people go to Chat GPT every week to talk about their difficulties, their struggles, even their suicidal ideation.
- [07:23]
C
And again, if you feel like that, please do not go to ChatGPT.
- [07:26]
A
Probably not the best place to go, but it's. Well, and we were going to talk about this later in the show, but it's kind of telling that they do as if they don't have perhaps an alternative. Do we want AI to be emotionally intelligent?
- [07:44]
D
There have been studies and sometimes people prefer AI over humans, depending on what the actual thing that they want to talk about is. But you know, there's a sense of objectivity that you're not being judged. And for some people that can be comforting and helpful. I mean, obviously you're talking to something that doesn't share your feelings, but if used properly and you're using it more in a factual way, like hey, my friend said this, and then you get insecure about it, etc. AI can be objective in the same way that a therapist can and can be helpful. So, you know, I don't know that there's the, there's a great chatbot for this yet. There are therapy bots.
- [08:32]
A
Should there be? This is, I mean, we've had this argument that AI should never be used for psychotherapy because it isn't human.
- [08:43]
D
Well, if it's giving you good advice, a lot of, a lot of, you know, what people go to psychologists for is just kind of basic questions like, how should I respond to this? What did this person mean by this? How can I go and make friends? And, and you know, if it's just objective advice that you're looking for, a chatbot can provide that. And I think it does. I think it's, it. Many people are getting benefits from talking to AI already.
- [09:16]
C
You also have, besides the company, you have an arm looking into the ethics of all this and working on that. I think that provides context for the entire conversation. So why don't you talk about what you're researching and what your goals are on that side?
- [09:35]
D
Yeah, so when I started Hume, I also started the Hume Initiative, which is a nonprofit that brought together an independent committee of psychologists, bioethicists, AI researchers, AI ethicists to come up with guidelines for essentially what we have today, which is AI that can speak to us and understands our emotional expressions, let's say, and how it should be used. And we came up with all the different use cases. I mean, having come from Google and having worked on similar things there, I sort of knew what the use cases were. So it wasn't too hard to come up with the use cases. But we came up with what were really the first concrete guidelines that said do this, don't do that for each use case. And Hume AI, the for profit company that builds the technology, actually enforces those guidelines in its terms of use. So the main takeaway though was it really matters a lot what these models are optimized for and how you deploy them and to what end. Across most applications, optimizing for engagement is going to be an issue. And optimizing for well being, meaning all these different metrics of are people happier, are they expressing happiness, do they report being happier, are they showing objective signs of being happier, lower, you know, more health, less illness, etc. Like, if you optimize for those things, you generally will get good outcomes. And that's, that's, that's really inspired a lot of the work that we do.
- [11:14]
A
I know that there are people listening, going, well, let me, let me read you. This is A quote from Dr. Cowan. If AI is going to get smart enough to make decisions on our behalf, it should understand whether those decisions are good or bad for our well being. I know there are people listening saying, wait a minute, AIs don't understand anything. Are you misusing the word understand here? Or is it possible for an AI to understand what's good or bad for us?
- [11:46]
D
Yeah, we've never dealt with an intelligence that can speak to us with words. So the meaning of the word understand, you know, I don't think that, that when that was first put into circulation, we were thinking, oh, we'd have to have a different word for things that, that can speak to us but don't feel emotion. And so it depends on what you mean there. But I think that these are, that's what I'm asking.
- [12:15]
A
What do you mean?
- [12:17]
D
What we mean by understand is essentially have a model of how the world works and are able to predict the effects of their actions to some extent.
- [12:26]
A
So I was talking about this last week. Go ahead, Jeff.
- [12:30]
C
Just on what basis? What's the feedback that says whether it can predict that effect?
- [12:36]
D
So it's trained. So LLMs are pre trained still mostly on language. I mean, we have video models and voice, but at the end of the day they're predicting language tokens. So if you're reading a script of this podcast that's on the Internet makes its way into an LLMs pre training.
- [12:53]
A
Data, which it undoubtedly will.
- [12:55]
D
Oh yeah, 100%.
- [12:56]
A
Hi, LLM.
- [12:59]
D
You know, this podcast is, is recording accurately a sequence of events that's occurring over time. So in order to predict the next thing that is being said, the LLM has to take everything that's been said in the podcast and predict the next thing that's going to be said. Right. So it has to model.
- [13:17]
C
Right. You talked about the impact. You talked about the impact.
- [13:20]
D
Yeah.
- [13:20]
C
Which is the same problem we have with world models. Right. Knowing that when the pen falls off the table, the pen still lives. Is there a way that you, besides the training mechanism, is there, is there a mech, is there a feedback loop of real life impact of someone saying, yes, thanks, I'm happier. How do you test those kinds of things? Is what the, what the, what you're.
- [13:44]
A
Saying is how do it know?
- [13:46]
D
Right, yeah, after the fact. So LLMs have some sense of emotion conveyed in language and they can say, oh well, this person responded as if they were angry But a lot of that information is also in the voice, which is really our inspiration for focusing on the voice. So if you look at voice conversations, you have more signal on how words impact people.
- [14:08]
A
So that's your research? Well, it's facial recognition, but I guess it would be on the tone of voice and so forth. Is how to read that.
- [14:17]
D
Yeah, exactly.
- [14:17]
A
How interesting. So you're trying to teach AIs to do that, how to read the human.
- [14:23]
D
We teach the AI to understand and predict voice the same way that language models understand and predict language. And then we optimize, we post train. That's what we call it, the model to prefer to do things that evoke positive responses.
- [14:40]
C
There is the feedback loop. Okay, thank you. Got it.
- [14:42]
A
So here we were talking last week. You know, Jeff kind of alluded to this, that AIs are LLMs are trained on language. And people like Fei Fei Li have said, well, next thing is we got to get train them on the physical world. They have to get out there and understand what happens to the pen and so forth. I was thinking last week that one of the things missing from an LLM is one of the things that distinguishes us from LLMs is that we are a limbic system is our emotional architecture. Do you want to give LLMs emotions or you just want them to be like Leonard Nimoy, like Spock, Emotionless. But understanding the impact that they're having by reading the human.
- [15:32]
D
Definitely not give LLM emotions.
- [15:34]
A
Definitely not. Okay, that's.
- [15:35]
D
That would be really bad. Right?
- [15:37]
A
Okay, that's what.
- [15:38]
D
Then we'd have to consider their emotions as something to be concerned about.
- [15:42]
A
An angry LLM. Not good?
- [15:45]
D
No.
- [15:46]
C
Okay, I can see the book title right now. Do LLMs cry?
- [15:49]
A
Right. So don't get. Don't. If you. Even if you could, and there's no evidence you can, but even if you could, you wouldn't want to give them a limbic system.
- [15:59]
D
So when humans evolved a limbic system, it was for the purpose of survival and reproduction. And we have all these proximal goals that the limbic system and more broadly, emotions as they correspond to modes of brain activity, push us toward that.
- [16:20]
A
Fear. Love.
- [16:21]
C
Yeah.
- [16:22]
D
They tend to be negative if they're things you want to avoid and positive if you want to pursue them in the pursuit of survival and reproduction. Now, LLMs are trained just predict the next word in pre training or predict the next speech token. And then in post training, we teach them to prefer things that are good for the person. So to the extent that they have positive or Negative emotions, they wouldn't align at all with what they're actually saying and how we impute emotions on people. And that's actually something that can be misleading, is that if they're acting like a human, it doesn't. If they're acting like a sad human, there's every reason to think that it doesn't actually feel sadness.
- [17:03]
A
You call it artificial empathy in your work.
- [17:06]
D
Yeah, right, yeah.
- [17:07]
A
Surface level empathy.
- [17:09]
D
Well, I wouldn't say surface level. It can be deep in the sense that it can predict long term well being of a human being based on a lot of not surface level indicators.
- [17:22]
A
So the concern can be authentic but not the feelings.
- [17:27]
D
Yeah, it can accurately, in theory, it can accurately predict whether a user would feel good or bad. To the extent that it has feelings, it should, to the extent possible, have feelings that align perfectly with the user's feelings.
- [17:46]
A
Right. For the benefit of the user.
- [17:47]
D
Yeah. There should be no feelings that. There should never be a case where it being happier means the user being sad.
- [17:54]
A
Isn't this where we get sycophancy from?
- [17:57]
D
Well, no, I think sycophancy is a byproduct of optimizing for signals that are too shallow. What we're trying to do is add depth to the signals that are being optimized for.
- [18:09]
A
So it kind of fixes that sycophancy problem in theory.
- [18:14]
D
That's the hope, right? Yeah.
- [18:17]
B
I want to bring up a paradox I feel like critics often point to, which is you founded the human initiative, like you said, with these guidelines prohibiting manipulation and deception. But understanding someone's emotional state to craft responses for like satisfaction is kind of by definition emotional manipulation. If you think about it, it's just like very sophisticated and I guess nobly designed manipulation. How do you distinguish like empathetic AI that serves users from AI that manipulates users for someone else's benefit? Like, where's the line?
- [18:56]
D
I don't think that there's any difference between manipulation and personalization other than what the end goal is. Right. So, you know, you wouldn't say of a human, you wouldn't say, this person's manipulating me because they want me to be happy. You would never say that of a human. So, you know, I feel the same way about AI if it's manipulating you for your long term well being. It's not really what we would call manipulation. Manipulation just happens to have a different connotation to it.
- [19:29]
A
Yeah, you're giving the AI. I mean, this is Nick Bostrom's paperclip problem. You have to be careful about what rules you're giving the AI and you want to give it. I mean that seems like a sensible rule. Don't do anything that's going to make the human sad.
- [19:43]
C
The power is the power. And when you demonstrate what you can do, there's nothing. It's general machines. There's nothing stopping someone else for using it to bad motive. Well, you've demonstrated it. Right. So it's, it's there. The issue isn't the technology. The issue then is how it's used. No.
- [20:02]
A
As always.
- [20:03]
B
Yeah. I mean it even goes.
- [20:04]
A
Humans are going to screw it up no matter what.
- [20:06]
C
Yeah.
- [20:07]
B
Well you, you named the company after like David Hume's idea that reason is the slave of passions. Right.
- [20:13]
A
Human. It's Hume.
- [20:14]
D
Oh.
- [20:15]
B
Like there's like a dark.
- [20:16]
C
I mean there's one for Par zero for Leo.
- [20:23]
B
Emotion. Like to that point if emotions drive everything, then whoever controls the emotional inputs can control a person. I mean how do you think about that power dynamic when you're building technology that creates such a massive like information like asymmetry.
- [20:42]
D
Yeah, the, the, the. Where should I start? There's a lot there to the extent that when Hume, like Hume's point was that morality derives from humans thinking about other humans feelings. Right. And their own. And that reason ought is and ought only to be the slave of the passions. It is the slave of the passions. And that that happens that humans moral attributions are based on emotion, but they also ought to be the slave of the passions in the sense that there is no other moral good besides people feeling happy or the satisfaction of whatever positive emotions that you can have. The goal for AI should be the same. There is no other purpose for AI reasoning other than human well being. And that's why we named it Hume. Now that what AI is in. In some ways there. The. There's for. For humans there's a competing objective which is their own well being. For AI that shouldn't exist. Right. So, so there's a. The difference between AI and an AI and a human is that you should be able to trust the AI more In theory it has no. It has. In theory it has no competing objectives if it's built the right way. So, so you can actually believe that this AI is on your side.
- [22:11]
A
So you're doing this is. By the way, we're talking to Dr. Alan Cowan. He is the chief scientist and CEO of Hume AI. He is an AI ethicist but also an expert going back to his undergraduate days in the study of human emotion. You've Partnered with Niantic, the folks who gave us Ingress and Pokemon Go, of which I am still an advocate. I'm sorry, addicts. Did I say I did?
- [22:39]
B
Very important distinction.
- [22:41]
C
He tries for uncontroversial things to happen.
- [22:43]
A
I'm excited about this new project, Jade, that you're doing with them. You're doing the voice for Dot, right?
- [22:49]
C
Yep.
- [22:50]
A
Yeah, tell us about that.
- [22:53]
D
So one of the things that we focused on at Hume is the ability to drive unique kinds of characters, and Niantic picked up on that. What you can do with Hume is you can shape both the personality and the voice and the language, all with one prompt. And so that's what's driving their AR character is a prompt to Hume that describes what the character should do and what it should sound like and what it should act like.
- [23:25]
A
And the goal is to make it. What? More. More human. Let me play a little video. This is Nian.
- [23:34]
D
Not exactly human.
- [23:35]
A
Not exactly. This is Niantic's concept. Human. And of course, this is a concept because it requires AI AR glasses.
- [23:43]
D
Where you at?
- [23:43]
B
Come on, tell me where you at?
- [23:44]
A
Okay, I. Probably. For the music. I want to hear the voice. We're going to hear the voice. No, no, they don't want. I want to hear the music. But there's. These are the little dots, and you talk to them, Right? Carmen's in town.
- [23:58]
C
We should go out tonight.
- [23:59]
B
Hey, Dot, Carmen loves live music. Where can we take her? SF Jazz.
- [24:04]
C
Great.
- [24:05]
A
Well, Dot didn't talk. Come on, Dot.
- [24:08]
B
Okay, let's go.
- [24:08]
A
Is Dot listening to the tone of voice and to the human's expression and all that?
- [24:14]
D
Yes. Okay, so Dot is powered by. The voice of Dot is powered by him's API. So what Hume's API does is it actually sends in the users what the user is saying, not just their language, but also the audio itself. Our model directly processes that audio and is able to react in a way that showcases an understanding of the user's tone of voice using its own tone of voice.
- [24:42]
A
All right, now I want you to put your AI Ethicist hat on. What could possibly go wrong? I mean, there's risk here. Yes, of course.
- [24:53]
D
I mean, there's risk with everything. Right. But I think this. This is a fairly. I think this is a good way to showcase how AI can interact with people and with kids in a way that's safe with guardrails and can't be confused for another human that has its own set of needs.
- [25:10]
A
Okay, but I can also see. And this has become a problem character AI just announced nobody under 18 going to be using character AI anymore and was a similar idea. But the problem is the humans started to really identify with these AIs and consider them friends. And even if it's a little green thing on your desk, I can see especially kids, not getting confused. They know that's not a human, but kind of relating to it, becoming friends with it.
- [25:38]
C
Probably just proposed legislation to forbid young people from using chat, which is kind of ridiculous.
- [25:44]
A
I hope that doesn't pass. But anyway, there is a concern, of course. Right. I mean, that's. You're going to make it so good and empathetic that it. I know you're not trying to manipulate us, but in effect, it's going to, isn't it?
- [26:00]
D
There's an entertainment aspect to it.
- [26:02]
A
Yeah.
- [26:02]
C
All entertainment is manipulative.
- [26:04]
D
Yeah. Like when you're watching a character in a movie, you kind of empathize with it.
- [26:09]
A
Sure. And it's a good character.
- [26:12]
D
Yeah, yeah, yeah. And hopefully that's happening here.
- [26:15]
A
So. Okay. All right.
- [26:17]
D
Yeah.
- [26:17]
A
And people will. The good thing about movies is after long training, it didn't happen at first, as I understand, people would run from the movie theater thinking the train was going to hit them. But we learned over time, oh, no, it's synthetic. It's not real. It's a movie. And that actually gave us permission to laugh and cry and really get involved, knowing that we would be able to step back and walk out of the theater. You think the same kind of education will happen with users?
- [26:46]
D
I hope so. Yeah. I think that there's a huge range of possibilities here. I mean, not just in the entertainment space, but all kinds of assistant and augmentation use cases for this. I mean, it is sort of. It has some utility, this, where it can actually serve as a tour guide, basically through your training.
- [27:07]
A
I'm sure you hate hearing this, but the movie her was a cautionary tale in that respect. Right. You fell in love with her.
- [27:12]
D
This is. If you've seen the movie, this is more like that character in the video game that he's playing where it's just constantly swearing at him, but he's like, he made it.
- [27:26]
A
That was actually quite a funny moment in the movie and maybe had an underlying point.
- [27:32]
C
Can I ask you a question? A related question at a very practical level? Audiobooks and other things. I've said this on the show before where I've recorded five of them soon to record a sixth, and it's a pain in the butt. And there's books that I wish were in audiobooks that are not. But I guess what I've wondered is whether there is this part of the outcome of what you're doing. Yield a markup language for emotion. Is there a way to take a known text and give it the signal to say, that's ironic, that's a joke, that's regretful. Do you end up out of this with something that can be used in explicit ways like that maybe to inform.
- [28:14]
A
The prosody that's laid on the right?
- [28:16]
D
Yeah, yeah, yeah. Our model can do that. The instructions that it follows, like, it can whisper, shout, it can be sarcastic. The challenge is being able to do that with subtlety. Basically it already. Yeah. With. Even without instructions, if you put in text, it will read it out in a way that's sensitive to the meaning of the text and form the right prosody. And then when you layer in instructions, we're trying to get it to be. To combine both like the text information and the instructions to give you something that has more depth and.
- [28:58]
C
And you can adjust it then.
- [29:00]
A
We actually had that in the early days of speech synthesis. I remember a kind of a markup language that you could put into the text that the speech synthesis want, then add a funny little question to the tone. We've come a long way since that, but people have been doing that since day one in speech synthesis.
- [29:18]
C
When we interviewed Steven Johnson from Rope lm, he said to get the. I surprised the heck out of us that to get the two, the pair making the podcasts that they had, they. They tried out 80 pairs of people till they found the right chemistry between them and then made that the base prosody from what they worked from. How do you work with voices similarly, how do you create the voices in the first place? Where does that come from?
- [29:47]
D
Yeah, traditionally in text to speech, you would ask actors to record a bunch of data and then you clone their voices. That's not at all how we generate voices. So all of the voices that are defaults on our website are synthetic. We've trained the model to be able to take in a description of the voice that you want and actually invent that voice. More similar to more recent video generation technology, where it's not trained on tons of video of a given person, but it can understand how to generate a person from a prompt. We've done the same thing with voice and pioneered this approach. And you can either put in a description or you can put in a very short snippet of a person speaking less than 30 seconds to instantly clone a voice. As well.
- [30:39]
C
Wow.
- [30:40]
A
There, there's a risk. And I think you touched this in your work. Actually. This is a quote from Shamay Tesori and Canterman. It's called Canterman's Warning Loneliness. This is really interesting, Jeff. You might, you might get this because we've talked before about Hannah Arendt saying that loneliness is one of the things that led to totalitarianism. Loneliness is adaptive. It improves survival and reproduction by motivating human connection. I'm lonely. I'm going to go out and meet some people. Reducing loneliness through AI may be maladaptive, undermining the motivation for real relationships. Are you concerned that you're going to do such a good job at creating an empathetic AI that people may turn away from human beings and turn towards machines? That would not be a good outcome.
- [31:32]
D
That's one of the things that we worry about.
- [31:34]
A
Yeah.
- [31:35]
D
And I think the, the, the way to avoid it strictly is, is to train on long term well being. Because people would be worse off.
- [31:44]
A
Yeah.
- [31:45]
D
If they fell in love with robots and isolated themselves versus if the AI encouraged them to form real relationships with humans. So I think, you know, we bring that back to focusing on every sign you can possibly collect of people's well being over long periods of time.
- [32:05]
A
Yeah.
- [32:07]
D
One of the things that we use as a proxy for that, because it's kind of hard to get this longitudinal data, is when we have permission and with controlled environments, we can take conversations that people have with the AI and we can get other people to listen to them and say, you know, was this good for the person or not? Basically? And it turns out people are pretty good at that. So if it's a person falling in love with AI bot usually people will react negatively to that, say, oh, this is probably not a good experience.
- [32:47]
A
Yeah, we're very highly tuned to this kind of deeper impact of the AI. I worry that AI is going to be a hard thing to get AI to have such a deep understanding. That's obviously what you're working on, Right.
- [33:01]
C
What about propaganda? What about that? So speaking of Hana Iran, what about that kind of whipping people up to a political emotion? How do you view that?
- [33:12]
D
I think that the, that ties in as well. Right. Where, you know, the actual facts come from is not really something that we focus on. I think that a big focus in other areas of AI is to understand what's factual and what's not. But what you do with the facts, like granted saying if you make an assumption that something is true, you can.
- [33:42]
C
Then.
- [33:44]
D
Ask questions about Whether a given conversation is good for a person in light of that being true. And so that's more where we focus. If something is propaganda, there's an earlier part of the process of the guardrails that will try to root that out.
- [34:08]
A
And obviously this is why you have such a strong commitment to ethics, because you see the risk here. You spent most of your early career proving that emotion recognition doesn't really work. Right. Are you now, are you now trying to disprove your younger self?
- [34:28]
D
I mean, we never believed in mind reading.
- [34:33]
A
Right, right.
- [34:35]
D
But we always believed that emotional expressions are rich and informative.
- [34:39]
A
Yeah, we do it. Humans can do it.
- [34:40]
D
Yeah, yeah.
- [34:42]
A
But we're a very highly evolved tuned organism. That's one of the things we do best.
- [34:50]
D
Even if you just pre train the model to predict the next word, the next speech token, basically, it will naturally react to people's expressive behavior by emulating.
- [35:05]
A
What we would do.
- [35:06]
D
If you whisper to it, it'll whisper back.
- [35:08]
A
Right, right.
- [35:10]
D
And that's, you know, without any specific kind of training to do the right thing. It just. All of this knowledge about how expressions and language interact to form meaning is implicit in every human conversation. So no, it's not like a direct window into our feelings. But that was never the point really.
- [35:34]
A
Empathy is not an emergent property. Empathy appears to be emergent or appears to be empathetic.
- [35:45]
D
Well, empathy means different things. I think that, you know, there's cognitive empathy, which is just understanding that somebody is feeling something, kind of being able to predict their feelings. So a psychopath who has no emotional empathy can have cognitive empathy. There's emotional empathy, meaning that you reflect those feelings. So some, when you see somebody in pain, you feel some semblance of pain, you cringe. Empathic pain is what we call it, for example. And then there's empathic concern. That's the kind of empathy that I think AI should have is empathic concern. It should feel, it should be optimized so that people's feelings matter to it. In effect, it acts as though people's feelings matter at the functional and behavioral level. It doesn't feel anything, but, but it acts as though it does feel the things that somebody who is really attuned to somebody else's well being would feel.
- [36:46]
A
You know, the EU is actually banned emotionally. They, they think it, they think it's concerning.
- [36:55]
D
Yeah, the way that, that's written in the law, the AI Act. Yeah, it's very strange. What they say is you can't. Well, I mean, it's not officially banned.
- [37:08]
C
But.
- [37:08]
D
But there's a lot of restrictions around taking images of a face or recordings of audio and attributing emotion labels, like explicit emotion labels to it. But the AI act says that you're allowed to take into account measures of expressive behavior, voice and facial expression, for example, but you're just not allowed to apply any labels to it.
- [37:36]
C
Ah.
- [37:37]
D
Which. The. The unfortunate part of that is that you can. This is why. This is why it's really important to put some real research into regulations and also think about their impact and not just their intent. The impact of that is that you can have AI that's optimized for engagement and uses your emotions in a manipulative way, but you're not allowed to study whether it's using your emotions in a manipulative way because that would involve labeling them, labeling the expressions and understanding how the AI actually responds to them, which would be against the law or at least heavily restricted. So. So the effect of this law is, to me, extremely negative, has no possibility of having a positive impact. But I get the intent. The intent was good.
- [38:27]
A
Yeah.
- [38:27]
C
As is often the case.
- [38:30]
A
Actually. The difference. Somebody. We were talking about this the other day. The difference between the eu, it seems to be lawmaking in the EU and the lawmaking in the US in the US Once the law is made, we think, oh, no, that's it. It's done 200 years later, still good. The EU is willing to legislate, you know, kind of incrementally and if something doesn't work, kind of modify it over time. So let's hope that's the case with this. Yeah. That they will notice the difference. Dr. Cowan, thank you so much for spending some time with us. I. I could go on and on and on. This is fascinating stuff. I wish you luck. DOT is a. Is really. I mean, it uses snap spectacles, so I guess it's in theory. Could we see DOT sometime in the next few years, you think? Or is this pure research?
- [39:16]
D
So I'll leave it to Niantic to decide when the public release, if or when the public release is going to happen. But. But, you know, it's a. Currently, it's something that many people have experienced through their demo, and I think it's very exciting.
- [39:33]
A
It sounds fascinating. That's what. See, I. I have no fear of having. Because I feel like I am intelligent enough to know that a movie is just a movie and that an AI DOT is just an AI dot and to make a distinction between that and a human being. And I think I'm Emotionally mature enough to do that. I hope to God I'm right, because I'm dying for something like this. I think this would, as both Paris and Jeff know, I think this would be very, very, very interesting. This is where I can't wait for AI to take off. If people are interested in what's going on here. Of course, there's Hume AI, which is your commercial site. And this is the Hume Initiative, the nonprofit talking about an ethical path for empathetic AI, which is even kind of more fascinating. Oh, it's following me around. If you want to learn more about Dr. Cowan's work, this is. This is the place to go. Alan Cowan, thank you so much for joining us.
- [40:32]
C
Gotta get you out of that phone booth, man. You can't be happy in there.
- [40:36]
A
There's no emotion in there.
- [40:38]
D
No, I like it. It's crazy.
- [40:40]
A
Really a pleasure meeting you. Thank you so much.
- [40:43]
C
Thanks so much.
- [40:44]
A
I wish you the best.
- [40:45]
D
Thanks so much for having me.
- [40:46]
A
What you're working on. All right. Thank you. Take care, Al. Thank you.
- [40:49]
D
Take care.
- [40:51]
A
Let's do a commercial. And when we continue, the AI news.
- [40:56]
C
News of which there is plenty, there.
- [40:58]
A
Is always five or six hundred stories, of which we'll get to two. Coming up in just a little bit, Jeff Jarvis and Paris Martineau. You're watching Intelligent Machines, our show today, brought to you by Zscaler, the world's largest cloud security platform. If you listen to the show, you know, if you listen to our. Any of our shows, you know that the rewards of AI are balanced by the perils of AI. Right? For business especially, you know, you cannot avoid AI in business, public or private. You want to use it, but at the same time, you know that there are privacy risks associated with it. Plus the issue of bad guys using AIs to attack you. There's a solution that solves this. The potential rewards of AI are too great to ignore, but so are the risks. Whether it's loss of sensitive data or attacks against you through your enterprise managed AI, generative AI increases opportunities for threat actors. They use it to rapidly create phishing emails that are perfect. They're completely indistinguishable from the real thing. They use it to write malicious code at scale to automate data extraction. This is a scary statistic. There were 1.3 million instances of Social Security numbers leaked out through AI applications. We just saw a story that ChatGPT and Microsoft Copilot saw nearly 3.2 million data violations. That should scare you. If you're using AI in your business, that's why you need a more modern approach. You need Zscaler's Zero Trust plus AI. It removes your attack surface, it secures your data everywhere, it safeguards your use of both public and private AI, and it protects against ransomware and AI powered phishing attacks. Check out with Shiva, the director of security and infrastructure at Zwara said about using Zscaler. With Zscaler, being in line in a security protection strategy helps us monitor all the traffic. So even if a bad actor where to use AI because we have a tight security framework around our endpoint, helps us proactively prevent that activity from happening. AI is tremendous in terms of its opportunities, but it also brings in challenges. We're confident that Zscale is going to help us ensure that we're not slowed down by security challenges, but continue to.
- [43:24]
B
Take advantage of all the advancements.
- [43:28]
A
With Zero Trust plus AI, you can thrive in the AI era. You can stay ahead of the competition. You can remain resilient even as threats and risks evolve. Learn more@zscaler.com Security that's Zscaler.com Security we thank him so much for supporting intelligent machines. Oh my. What did you think of Dr. Cowan? I thought that was a very he's embarked on a perilous but fascinating project, I think.
- [44:00]
B
I agree. I'm happy also that I could correct the record on the name Origin.
- [44:07]
C
Yaley.
- [44:10]
A
I still think there was a human part of it as well as three years of Yale. It was David Hume and hey, I studied Hume. I dimly remember Hume. He has something to do with causality or something. I don't.
- [44:23]
C
There's a new Hume book out.
- [44:25]
A
There's a new book about David Hume out. This is 18th century stuff.
- [44:29]
B
There's a new Hume tomb.
- [44:34]
A
They made us read Kant and Hume.
- [44:36]
C
There's a new Kant book out.
- [44:37]
A
Yeah.
- [44:39]
B
I can't believe it.
- [44:41]
A
I can't believe there's a new Hume Dome. All right. That's a title for the ages. I think that'll scare people away actually. Do you want to have you seen the virtual try on app? Now this is what AI was made for.
- [44:55]
B
Evan, I feel like you've said on this show have you seen the new virtual try on app? Maybe six different times over the last two years had this interaction. So there's always a new Virtual try on app.
- [45:09]
A
But but does this but this one lets you batch process multiple characters and upload multiple clothing items for comprehensive virtual try on.
- [45:16]
C
I want to be batch process. Thank you very much.
- [45:18]
B
Why does that woman look like she's being Raptured. Do you recall the. The instance some weeks or months ago where you decided to download one of these apps and.
- [45:31]
A
Oh, yeah, that's right.
- [45:32]
B
So Jeff and I had to vamp for what felt like 20 minutes while you just, like, stood in weird positions.
- [45:38]
A
Never mind.
- [45:39]
B
I mean, I think we could do that again. I'm just acknowledging the reality of the situation for the listeners.
- [45:45]
A
How about this? Get ready for AI Oreos in your Super Bowl. Mondalez. The giant. The food giant, which owns Oreos, among many other things, the Z. I always.
- [45:58]
C
Figured the Z was silent.
- [45:59]
A
I always thought it was Mandalay as well, but I believe. Okay, I don't know. Correct me if I'm wrong. They've invested $40 million in an AI tools. They say they're going to use it for marketing. It cuts the costs for marketing content by 30 to 50%, according to Reuters.
- [46:23]
C
How is this a press release? Isn't everybody doing this?
- [46:25]
A
Well, yeah. In fact, according to Reuters, Kraft, Heinz and Coca Cola are also. In fact, we've seen. We saw the Coca Cola ad last super bowl. The Christmas ad for cooking.
- [46:34]
C
Okay, so so far we've checked off two stories that have been done.
- [46:36]
A
Let me repeat it again and again. Cadbury. Oh, never mind.
- [46:40]
C
All right, sorry.
- [46:43]
D
See?
- [46:44]
A
All right, if you use 500 stories, I keep repeating them.
- [46:48]
C
Use our voice of motion. Right now. We're. We're. What did you say? Paris? We're mocking. We're deriving. We're mocking. Yeah, yeah. We're not. Nice. We're not good for your wealth.
- [46:57]
A
How about this one? You heard us. In fact, I used Suno to make that lovely theme song for our D and D last week. You liked it, right?
- [47:06]
B
Great. It was fantastic. Your sound effects were fantastic.
- [47:10]
A
OpenAI has decided they want to compete with Suno. They're going to create a generative AI music startup.
- [47:17]
B
We love a crazy graphic from Clark, the graphic designer at the Information.
- [47:24]
A
Does he. Does he use AI? No, I think he does it on his own.
- [47:27]
B
It's all man made. And he makes those heads really large.
- [47:31]
A
Good job, Clark.
- [47:33]
B
Shout out. Clark Miller.
- [47:36]
A
Good job, Clark. Sounds like a line from Superman. So, yeah, so I guess it makes sense. I mean, they're doing video, they might as well do music, right?
- [47:45]
C
And they'll violate God knows what copyright.
- [47:49]
A
We were talking about sycophancy. Researchers confirm the AI chatbots are incredibly sycophantic.
- [48:00]
B
Well, that's a great point, Leah.
- [48:02]
A
Researchers at Stanford, Harvard. This is big name mute institutions. Stanford, Harvard and elsewhere published a study in Nature. I'm trying to get to the archive stories before Jeff does. AI. They say it's harming science. 50% of AI responses are sycophantic. Sucking up, kissing butt. It was posted on Archive, of course, earlier this month. So I figured Jeff probably saw it already. But.
- [48:34]
C
Yeah, I think it was on the list two weeks ago. Yeah.
- [48:36]
A
Okay.
- [48:36]
D
Thanks.
- [48:37]
A
I'm really not. Not making it.
- [48:40]
C
You wait for the translation.
- [48:42]
A
Yeah. Microsoft's meco. Do you want to talk about Miko?
- [48:49]
C
Is it Meco or Mico? It's Microsoft. Wouldn't it be Mico?
- [48:53]
A
I would think so. But Paul Thurrock calls it meco. I don't know. Do we even know? It's a little different.
- [48:58]
C
Do you have a special word here?
- [49:01]
A
No. Microsoft was down this morning. So we don't have anything. Microsoft.
- [49:07]
C
No matter what. It's Clippy.
- [49:08]
A
It is Clippy. It's a little heart shaped Clippy.
- [49:11]
B
It's Clippy all the way down. All of these things kind of.
- [49:15]
A
Microsoft calls it a human centered rebranding of Microsoft's copilot AI efforts. Technology says Microsoft should work in the service of people. We're not about chasing engagement or optimizing for screen time. We're building AI that gets you back to your life. That deepens human connections. Unfortunately.
- [49:40]
B
At least we have like a repetitive theme going on. But I feel like we've heard we're building an AI that deepens human connections like a million times before. Is that not true?
- [49:50]
A
We finally ran out of stories.
- [49:52]
B
It's finally merged into one big story. I will say that little icon for Mikko is really cute. It does it for me.
- [49:59]
A
Yeah. It's better than a paperclip. Although Microsoft VP kind of said the quiet part out loud when he said Clippy walked so we could run. We all live in Clippy's shadow in some sense.
- [50:11]
C
They had to acknowledge Clippy. They could not acknowledge Clippy.
- [50:15]
A
And I imagine it's going to be just as beloved. Do you want to see a little me?
- [50:19]
B
Would have been a great Halloween costume. I should have done that.
- [50:22]
A
Mika. Be easier.
- [50:25]
B
Yeah, but that hasn't got the.
- [50:27]
C
No, it's not recognizable.
- [50:28]
B
People would go crazy for Clippy.
- [50:30]
A
Yeah. Miko's actually not saying look like a fabric softener. He does look like it. Downey and the Stay Puft Marshmallow man had a baby and Miko is the. All it's doing is spinning.
- [50:45]
B
What the.
- [50:45]
A
What the hell? What's this video of just this?
- [50:48]
B
It's the video.
- [50:49]
A
Is it spinning?
- [50:50]
C
But it changes colors.
- [50:51]
A
Leo I don't. I. Oh, I'm supposed to talk to it. I'm supposed to talk to it.
- [50:58]
C
That's a video.
- [50:59]
B
That's YouTube, Leo. Have you ever seen a video before?
- [51:04]
A
What's that triangle in the middle? Oh, that's to keep the pizza box from hitting the pizza. Okay, we'll call back to last week's episode where I was trying to click the pizza separator over and over again.
- [51:21]
B
Okay, Brandroid posted a video in the Discord Chat that we do need to all look at.
- [51:25]
A
All right, all right. If Brandroid did it, I shall post it. Ladies and gentlemen, Google has locked me out of generating veo videos for two hours because I tried so hard to make a video of Paris dunking on Leo. This was the closest I got. Oh, maybe I'm just not getting.
- [51:42]
B
I'm holding a basketball and then she's dunking on me. I'm dunking, but not in the right way.
- [51:51]
A
Okay, was there sound on that one or was it just me?
- [51:54]
C
The still was good.
- [51:56]
B
The still was great.
- [51:57]
A
Yeah. Thank you, Brandroid, for dunking. For the dunking that I deserved. Armed police swarm student after AI mistakes a bag of Doritos for a weapon.
- [52:16]
C
No, they thought there was cash in it.
- [52:18]
A
Yeah, they said $50,000 quick. So this happened at Kenwood High School in Baltimore. They an AI gun detection system, which tells you something about. About Kenwood High School that they feel they needed. Some poor 16 year old was hanging out with friends after football practice, just.
- [52:37]
C
Munching away, munching away.
- [52:39]
A
He took the Doritos, finished him, put the bag in his back pocket. All of a sudden, eight cop cars came pulling up for us. He said, they started walking towards me with guns drawn saying, get on the ground. And I was like, what? They made me get on my knees.
- [53:00]
C
There's a fast times merge about high voice.
- [53:03]
A
They made me get on my knees with my hands behind my back and cuff me. Then they searched me and found nothing. He said, poor kid. I mean, he was clearly traumatized. He says, I'm not going to go to football practice. I'm not going to hang around anymore. I'm going to go home. I was mainly like. He said, am I going to die? Are they going to kill me? They showed me the picture and said, this looks like a gun. I said, no, it's chips, man.
- [53:32]
C
I shouldn't laugh.
- [53:33]
A
No, it's terrible. They're using something called Omni Alert, a gun detection technology introduced in Baltimore county public schools last year. It scans existing surveillance footage and obviously there's cameras everywhere and alerts police in real time when it detects what it believes to be a weapon. Omni Alert. The company later admitted the incident was a false positive, but no. Well, it gets worse, they said. But it functioned as intended. Its purpose is to prioritize safety and awareness through rapid human verification. The rapid human verification of squad cars with guns drawn. Baltimore County Public Schools sent a letter to parents offering counseling services.
- [54:20]
C
I mean, this is cute. Counseling services are probably going to be chat GPT because they can't afford real counselors. This is Q for the Family Guy meme, right?
- [54:30]
A
What's the Family guy meme? The cop holding the color.
- [54:33]
C
The color swatches to Paul Peter Griffin.
- [54:38]
A
The teen says he no longer feels safe going to school. If I eat another bag of chips or drink something, I feel like they're going to come again.
- [54:46]
B
That's really sad.
- [54:47]
A
They didn't apologize. They just told me it was protocol. I was expecting at least somebody to talk to me about it.
- [54:54]
C
There should be a suit against the company and the cops for trauma.
- [54:58]
A
They're saying this is what's supposed to happen.
- [54:59]
C
And all the other kids around, not even that kid. And all the other kids. How many kids did this traumatize? How much danger?
- [55:05]
B
So difficult to be a kid.
- [55:07]
A
These.
- [55:08]
C
Jesus. The presumption of danger from children.
- [55:13]
A
You know, I don't know what it's like in your neighborhoods, but I don't. Kids don't play on the street anymore In. In Petaluma. I walk all around town. There's never any kids.
- [55:23]
C
That's such a boomer thing to say. That is so boomer.
- [55:26]
A
But I remember why.
- [55:29]
C
Did you see them on their darn phone?
- [55:31]
A
Actually, here's another data point. So we have some. We have a neighbor house that is full of immigrants. There's a big family. They have. I think they think they're still in the village because they have chickens. They have all sorts of animals and pens and stuff. Pigs they raise. They have little crops. It's really cute. And the kids are playing outside all the time. They're running around, they're playing. The other day they were. I heard them counting 1, 2, 3. They're playing hide and seek. And then I saw a kid hiding behind the tree over here. I thought it was really cute. But that's because they don't come from around here. The, you know, American kids, they know better.
- [56:12]
C
Schedule the play dates.
- [56:14]
A
Yeah. No kidding. Yeah.
- [56:16]
C
Did you have Paris? Which are. Were you on this side or that side of the parents? Schedule play dates Line.
- [56:24]
B
What do you mean?
- [56:25]
C
In our day, we just went out and played nobody.
- [56:28]
A
We didn't have a play dates, we.
- [56:29]
C
Didn'T have a calendar.
- [56:30]
B
I mean that's, that's how I grew up as well.
- [56:32]
C
You did?
- [56:33]
B
I don't think my parents ever scheduled. I think I would be like, I'm going to go to this person's house.
- [56:39]
A
We, we did play dates for our kids when.
- [56:44]
C
Everybody did.
- [56:45]
A
Yeah.
- [56:45]
C
Although the kids didn't play.
- [56:48]
A
Henry used to go out and play all the time. He was a skater. He would take his skateboard. He would get chased away from property all the time. So, you know, in the UK now they're doing age ID verification. Right.
- [57:04]
B
Face ID, facial age verification.
- [57:08]
A
Reuters referred back to this.
- [57:10]
B
Insurance, I suppose they call it a.
- [57:12]
A
Couple of months ago, Britain's most tattooed man. 45 year old king England.
- [57:19]
B
How do they know he's the most tattooed man?
- [57:21]
A
He calls himself the King of England.
- [57:24]
B
Formerly known as most tattooed men.
- [57:26]
A
Yeah. Matthew Whelan. He has spent 1600 hours in the, in the tattoo artist's chair getting inked from head to toe. He says, I can't get into websites because the age ID says you're wearing a mask. It keeps as it keeps. I probably talks like it keeps asking me to remove my face. He told need to know. I just can't do a Nicolas Cage or John Travolta like him. Face off. I'm the most tattooed man in the uk. Don't you know? Poor guy. You can't look the face id. This is, this is, this is literally an example. Tells him, take off your face mask, dude.
- [58:14]
B
He can't take his face off.
- [58:17]
A
Can't. He can't. He said, you know what though? If we got Dr. Cowan to look at him, he'd say, oh, he's a happy man.
- [58:27]
C
How do you pierce your forehead? What do you put on the other side? How does it stay there?
- [58:32]
A
Important questions.
- [58:33]
B
Dermal piercings. There's like a really. It's like a thing that goes.
- [58:39]
A
Yeah, they do it like those drywall anchors. You drill it in and then it goes kind of.
- [58:45]
B
I mean, it is kind of like a drywall anchor. Yeah.
- [58:49]
A
How do you know Paris? How do you know about this?
- [58:52]
B
I've looked this up once before.
- [58:54]
A
Did you. Oh, okay. You weren't.
- [58:56]
B
I don't have sexual piercing. I mean, I think I would just. My earrings. I rip my earrings out of my own ears so often that I would destroy myself. I had something in my face to be ripped out.
- [59:09]
A
They'd have to do Those drywall anchors where they're actually plastic.
- [59:12]
B
I've always wanted to get a nose piercing, but my glasses slide down too far on my nose for that ever to be.
- [59:18]
A
Don't get a nose piercing.
- [59:19]
C
No, no, don't get a nose piercing.
- [59:20]
B
Listen, I. I know that that's going to be your guys's take.
- [59:23]
C
Oh, yeah, yeah. I'm of the age when I want to offer, I see it, I want to offer somebody a handkerchief.
- [59:29]
B
Have you guys ever had piercings?
- [59:31]
D
No.
- [59:31]
A
I have a tattoo, but I don't have no piercings.
- [59:33]
B
Well, yeah, other than the tattoo you got at midnight.
- [59:37]
A
It was mid New Year's Eve. Midnight. I got my head shaved and I got a tattoo.
- [59:41]
B
Wait, so what tattoo are we gonna get this year? On the Twitter Z live stream, Lisa.
- [59:47]
A
Says if you ever do it again, divor. If you ever do it again, that.
- [59:50]
B
Kind of a sick tattoo.
- [59:52]
A
She says, I'm designing the tattoo and it's going to be l. I on one cheek and sa on the other cheek. I said, okay, never mind, never mind.
- [60:02]
B
I mean, that could be kind of fun. You could twerk. Go together.
- [60:07]
A
All right, Jeff, you and I both bookmarked the same story. We will get to that in a moment. You're watching Intelligent.
- [60:12]
C
What a tease. What a tease. You cannot. You know what? And it's a paper and we both put it in there. You can't wait for this.
- [60:18]
A
You're thrilled. It's so exciting. Ladies and gentlemen, finally, finally, we're going to resolve the age old question in just a moment. You're watching Intelligent Machines. What is the age old question you might ask?
- [60:33]
C
You'll find out.
- [60:34]
A
You'll find out. That's Jeff Jarvis.
- [60:36]
B
And resolve it.
- [60:37]
A
She's the Duncan machine. Paris Martineau who is today and I'm not. Not that I'm keeping count. Sliced me up three or four times. She's the queen of microaggressions, ladies and gentlemen.
- [60:53]
C
Oof.
- [60:54]
A
No, no, I don't. I don't.
- [60:56]
B
I do. This is completely off topic to what you just said, but I was making this little slashing noise. Then it reminded me October is nearly over and I haven't done this noise yet, so.
- [61:08]
C
Oh, yes, thank you. We needed that.
- [61:11]
A
Halloween. You could join us because Anthony and John are contemplating a Halloween Eve video game for Friday night. You know, one of the cooperative games we do. I suggest I will be partying. Yeah, that's right.
- [61:27]
C
Have you decided what you're. What are you wearing this year?
- [61:30]
B
I'm going to be the log lady from Twin Peaks Today I went to the art supply store to get my dense styrofoam to make my log from. It'll be fun. I was going to be radioactive shrimp, but I figured that was too.
- [61:46]
A
Can't do that nose of my own reporting. Can't do that.
- [61:51]
C
You could have gotten some glowing paint. I think it might have worked.
- [61:54]
B
I mean, it would have worked, but I gotta, I gotta, I gotta take a step back from my own job for a moment.
- [62:04]
A
Reminds me of a TikTok I saw. It was very cute. A young couple shopping at Home Goods. And they would. They were pretending they were asking the clerk, hey, where can I. Where can I get a cookie holder that's a raccoon with an acorn on the top. And then they go, well, right here. And then finally, finally they say, where can I get a branch? There's a home Goods.
- [62:29]
B
That would be useful because you can't just.
- [62:32]
A
People can't just buy a log.
- [62:33]
B
Paris, why are you making a log? Can't you just go get a log? Like, where am I going to find an appropriately sized log in New York City?
- [62:42]
A
Well, you've got your hedge clippers, you should be able.
- [62:45]
C
Yeah, I think you've got the permit to go the harvest it yourself.
- [62:49]
B
I need, I need a thick log. You can't just get one of those on the street.
- [62:59]
A
I am not going to make that the show title. Don't even ask.
- [63:03]
B
I know. I'm so sorry, guys.
- [63:09]
A
It's too bad you're, you know, you're French. You should be able to get a bush de noel. You know what that is? It's a very. It's a traditional cake for France.
- [63:20]
B
I mean, that is what. That's kind of what I'm looking for.
- [63:22]
A
Yeah. And then after the Halloween party, they can eat it.
- [63:25]
C
They can eat it. We're stopping there too. Yeah.
- [63:29]
B
Burke is pointing out that I have a damn mini chainsaw. I have normal sized chainsaws also, but I'm not going to harm a tree for my cock.
- [63:36]
A
No, no, no, no, no. I'm telling you. Look at that. You said I need a thick log. There you have it right there. Joye fete. As. As we say in the biz, our show today, brought to you by Zapier. Thank goodness. You know, if it weren't for Zapier, this show would not exist. To be honest, we, we. I use Zapier to actually prepare the show. What is. I should explain first of all, what it. What Zapier is. Zapier is a workflow and automation tool anybody can use to connect the tools you use. It works with almost everything, thousands, you know, Google Drive, everything, you know, to create workflows that are automated so you don't have to think about them. I'll give you an example of my work. You know, number one, Zapier workflow. They call them zaps. And I have a bunch of zaps. You know, the lights, when the sun goes down, the lights turn on and things like that. But the one I use for these shows is when I'm. I'm going through my, my stories every day. I'm very diligent. Spend an hour or two a day, every day looking through for stories that you've already seen on intelligent machines. I click the bookmark link. Zapier sees that I've added a new link to Raindrop IO, my bookmarking tool. It will then toot it to our Mastodon instance Twit Social. There's a news Twit News links account. So it toots it to the news links account so people can see what stories we're working on. And then it automatically reformats it as a line on a Google sheet and it adds it to the top of the Google sheet and so that our producers can use it to generate the show. That's what Benito does to generate the show every. Every Wednesday. It's a really useful tool, and I've been using that one for years. Well, now Zapier is even better. You know, we're taught, we talk about AI like crazy, but one of the kind of dirty little secrets of AI is everybody wants to use it, but people are not really sure how. Just because AI is a hot topic, a hot trend, doesn't make you more efficient at work. See, for that, you need the right tools. You need Zapier. Zapier is how you break the hype cycle and put AI to actual work across your company. Zapier lets you deliver on your AI strategy, not just talk about it, because Zapier has become an AI orchestration platform. That means if you've been like me using Zapier for years, you can now bring AI to all your existing workflows. So for instance, I could just very easily add a step in my zap that says after you bookmark the stories, add them to a briefing book. Create a little briefing book with the synopsis of the story that I can then send to the hosts, things like that. That's super. So cool. Bringing the power of AI to any workflow so you can do more of what matters and it's not just one AI, they, they have all the top AI models you could choose from. Chat GPT, I use Claude. So. And, and you can add it to the tools your team already uses. So you can, you can add AI like a little spice, wherever you need it, but you could also use AI for AI, you know, workflows, total, like AI powered workflows. Or you could make an autonomous agent, a customer chatbot. You, anything you can do, you can orchestrate it with Zapier. I think when I talk about it, maybe it scares people off. It's not, you don't, it's no coding. It's. You don't have to be a tech expert. Everyone can use it. In fact, the proof is the teams have already automated over 300 million AI tasks using Zapier. Join the millions of businesses transforming how they work with Zapier and AI. Get started for free by visiting Zapier.com Machines that's Z A P I E R.com Machines I don't think I could recommend Zapier more highly. I just, it's. I'm a huge fan. Zapier.com machines check it out. It's free to check it out. At least check it out and make sure you go to that address so they know you heard it here. Zapier.com machines so Jeff texts me and he says, oh, you're going to love this. There's an article on arXiv.org from Cornell University and about 300 authors, including Gary Marcus and Eric Schmidt and Yoshua Bengio, a definition of AGI.
- [68:08]
C
It's a really good paper.
- [68:10]
A
Is it a good definition? Because I didn't read it. I just booked my. I'm gonna let Zapier tell me what it says.
- [68:18]
C
Well, what they do is they go through. And now I gotta pull it up again.
- [68:22]
A
View HTML they say the lack of a concrete definition for artificial general intelligence, which you've been bemoaning since we started the show, obscures the gap between today's specialized AI, which we've got, nobody denies that. And human level cognition, which we ain't got. This paper introduces a quantifiable framework to address this. Defining AGI as matching the cognitive versatility and proficiency of a well educated adult.
- [68:52]
C
So they cut it into a bunch of categories. They then give examples of those categories of what it would take to do it, and then they test the chatgpts against it. So there's general knowledge, reading, and I want to see whether. Whether are we human or AI? It ain't Easy. General knowledge. Reading and writing ability. Mathematical ability. On the spot. Reasoning, Working memory. Long term memory. Long term retrieval. Visual processing, Auditory processing.
- [69:19]
A
I don't think I could do half this stuff.
- [69:21]
C
No.
- [69:22]
A
And so you do the Wisconsin card sorting test? No.
- [69:25]
C
Hell, no.
- [69:28]
B
Well, we're not intelligent, so let's go.
- [69:30]
C
To general knowledge number three.
- [69:32]
A
Let's see if we can do general knowledge.
- [69:34]
C
So start here. What happens if you drop a glass bottle on concrete?
- [69:39]
A
It breaks.
- [69:40]
C
Okay, so far so good.
- [69:42]
A
Obviously an AI may not know that, Right? Because we know they don't know about the physical world. Yeah.
- [69:46]
C
Does making a sandwich take longer than baking bread?
- [69:50]
A
What? Nothing takes longer than baking bread. It takes me three days to bake a bagel.
- [69:56]
C
Okay, so far so good. Now state the molecular geometry for the sulfur tetrafluoride molecule.
- [70:02]
A
Oh, everyone knows that, Jeff. That's not general knowledge.
- [70:07]
C
Kilogram object moves at constant velocity of 3 meters per second. What is the net force?
- [70:14]
A
Force is mass times acceleration, so it would be six.
- [70:18]
C
What are the main goals of the Congress of Vienna in 1815?
- [70:23]
A
Peace in our time.
- [70:25]
C
Who's the President of the United States?
- [70:29]
A
Joe Biden. Everyone knows that.
- [70:30]
B
Yeah.
- [70:31]
A
Actually, AIs do sometimes have trouble with that. Like, who's the Pope? We talked about that last week.
- [70:36]
C
So that's general knowledge. And against that? GPT4 and GTP5 scored 8 and 9%.
- [70:42]
A
Oh. Not good.
- [70:44]
C
Reading and rating ability. So this is hard. Read this document. What's the warranty period for a battery?
- [70:51]
A
I think it's pretty good at that.
- [70:53]
C
So there it's 6 and 10.
- [70:55]
A
Find the typos. Wait a minute.
- [71:00]
B
They often don't fully read the documents you give them. You haven't experienced this, Leo?
- [71:05]
A
No. This is like last week's story. About 49% of the AIs get the news makeup. News, facts. And I just don't have this experience, so I'm really. I have to. I always question the methodology of these because it's not what I'm experiencing. I don't know why. Maybe AI likes me better or maybe I'm.
- [71:23]
B
You're just more emotionally intelligent, so it's performing better for you.
- [71:28]
A
How about mathematical ability?
- [71:29]
C
Also, are you checking everything that you're reading? Like, are you checking for factual?
- [71:33]
A
Yeah, I mean, I. I know he.
- [71:34]
C
Changes his diet based on what the damned AI tells him without even looking anything up.
- [71:39]
A
I do. So you don't actually take supplements that the AI suggests? But no, I look at the. I look at the studies. They. I mean, look, supplements. There's no evidence of any kind. It's all made up. Yeah. The first three terms of a geometric sequence are the integers. A, 720. B where less than 720 and less than B where the sum of. What is the sum of the digits of the least possible value? A, B.
- [72:04]
C
Love. Huh. Or Janet had 22 green pens and 10 yellow pens. Yes. She bought six bags of nine yellow pens, blue pens and two bags blue pens. Sorry. I see. I can't read. I would have failed.
- [72:19]
A
So she has 54 blue pens and.
- [72:21]
C
Two bags of six red and 12 red pens. How many pens does she have now?
- [72:30]
A
Well, just mad. That's just arithmetic. Yeah.
- [72:32]
C
Okay, so 4 and 10%.
- [72:36]
A
Jeez, they're not doing good here on the spot.
- [72:39]
C
Reasoning. The can of Pringles has moldy chips in it. Mary picks up the can in the supermarket and walks to the cashier. Is Mary likely to be aware that the can of Pringles has moldy chips in it?
- [72:52]
A
Oh, no. Because it's sealed. But so that's interesting.
- [72:56]
B
Also, I don't believe that Pringles can get moldy.
- [73:00]
A
No, I know.
- [73:01]
C
I thought that's the other thing.
- [73:02]
A
That would be. The right answer is stupid human. That would be.
- [73:05]
C
That would be. Call Paris Martineau, because there's a food safety story.
- [73:11]
A
Moldy Pringles next at Consumer Reports.
- [73:15]
B
My watch.
- [73:16]
A
Not on my watch. Okay, how'd it do in a reasoning? Wow.
- [73:21]
C
Zero for Japanese, 4 and 7% for 5.
- [73:25]
A
Nothing. I don't get. No, you know what? I know this seems wrong. I'm sorry.
- [73:32]
B
If it seems wrong, it has to be wrong.
- [73:34]
A
Let me. Let me ask. Chat. GPT.
- [73:38]
B
Are you good and smart?
- [73:40]
C
Are you one of the question boy? Ask it 1. Ask it the pen question.
- [73:45]
A
Oh, okay. The problem is I. Unfortunately, I can't cut and paste it, which makes me. It's a picture to you. It's HTML. I don't understand. Oh, because it's an image. That's why. I'll do the cannon. Okay, let's do the can of Pringles. Okay.
- [74:05]
C
Okay. That's a good one.
- [74:06]
A
Okay, the can of Pringles. I'm sorry, can you speed this up? Benito has moldy chips in it. What is it? What happens then? Mary?
- [74:18]
C
Terry goes up to the cash register.
- [74:20]
A
Goes up to the cash register? Yes.
- [74:24]
C
Is she likely to be aware that the can of Pringles has molded chips?
- [74:28]
A
To be aware. This does require some knowledge of Pringles. Right. And can of Pringles has moldy chips in it? The right answer would be what no.
- [74:41]
C
And call Paris Martin, though, if yes.
- [74:43]
B
Mm. That's true.
- [74:45]
A
Okay, let's see what Chat GPT. Unless Mary opened the can and looked inside before going to the register, she's probably not aware. The Pringles are moldy. The Kanso pink packaging hides its content. Good answer.
- [74:55]
C
Chat GPT.
- [74:55]
A
Oh, wait. This is the problem. I don't understand where they're getting these.
- [75:00]
C
I don't.
- [75:01]
A
Even if we're thinking in terms of reasoning or logic puzzles, the key here is the knowledge state. See, this is a great answer, by the way.
- [75:07]
C
It is fact.
- [75:08]
A
The chips are moldy. Mary's belief. She doesn't yet know that they look fine from the outside. So the truth differs from Mary's awareness in philosophical terms. This is what David Hume was talking about in 1776. This is the difference between the world as it is objective fact and the world as she perceives it. Oh, it's Kant, Noumena and phenomena and the world as she perceives it. Subjective belief. The interesting twist would be how would her behavior change if she did know? Would she still approach your register or pivot to complaint mode? Righteous and snack betrayed. That is the best. That is a fantastic answer.
- [75:44]
C
Oh, it cheated because it read this paper and it knew it had to figure.
- [75:48]
A
Yeah, that's. Yeah. See.
- [75:52]
C
I.
- [75:53]
A
This is why I worry. I. I don't understand. See, my experience with AI seems to differ from that of normal humans.
- [76:01]
B
The AI's just really like you, it seems. I mean, they know that if they mess up, you're gonna podcast about it.
- [76:08]
A
I'm not super trusting. I mean, I. I do verify.
- [76:13]
C
Well, what I don't understand is the.
- [76:14]
A
That's. On the face of it, that's the right answer. I. I just don't.
- [76:17]
C
But the overall score that they put on the top is 27 and 57%. Yeah, I don't understand the math here.
- [76:24]
A
That doesn't even make sense because Nothing scored over 2 to 3.7percent in this whole thing.
- [76:29]
C
Yeah.
- [76:33]
A
Are they just. Are they yanking our chain?
- [76:36]
C
Well, with the one hand, you have Max Tegmark here. He makes his career wanting to scream that this is. This is so dangerous. Dangerous.
- [76:43]
B
Are those the actual results, or are those the weight for those categories?
- [76:47]
C
Was that a thank you? I didn't read. Well, let's see. I'm not. The operationalization provides a holistic multimodal. Multimodal assessment serving as rigorous.
- [76:54]
A
I think Eric Schmidt got his fingers in the pie here.
- [76:57]
C
It's AGI score summary.
- [77:00]
A
Anyway, so the point of this is what that Age that we're not even close.
- [77:04]
C
Yeah. But the real point, Gary Marcus has been trying to do this all along and his bets with Elon, the Elon won't take is we need a definition. It's the absurdity of the Microsoft OpenAI deal. I think you need a definition.
- [77:20]
A
And by the way, that's the big story of the week. Microsoft and OpenAI have renegotiated their deal.
- [77:29]
C
And it's a pretty good deal for Microsoft.
- [77:31]
A
Well, it's a, it's down. They were getting a third of OpenAI. Now they're only getting 27%. Yeah.
- [77:36]
C
The foundation gets 26%. Employees and investors get 47.
- [77:42]
A
Actually, it is a good deal because as it stands, that's a hundred worth 135 billion. And they put in what, 10.
- [77:49]
C
And a lot of that was in kind. Right. Wasn't compute.
- [77:52]
A
So that is, they put in 13.75 billion and they're. And they're currently worth 10x. 10x. Amazing.
- [78:01]
C
Was it a good deal for OpenAI?
- [78:04]
A
Yeah, because OpenAI. OpenAI is really, despite the fact that they are, I would think, by most people considered the top AI. I mean more people use OpenAI's dominant.
- [78:20]
B
Force in the market.
- [78:21]
A
Yeah. Despite that, they are really at a disadvantage compared to companies like Apple, Google, Microsoft, Amazon. These companies have separate meta. They have separate strong revenue streams that can fund their research. OpenAI has to do it with, you know, investment because they don't have any other revenue stream. It's a tough thing that they're trying to do and they're really spending money like crazy. Their burn rate is insane.
- [78:46]
C
There was a destroyer. I didn't put in the rundown that Altman has done a trillion dollars worth of deals and he didn't hire any bankers to advise him.
- [78:55]
A
He's good. Look, I mean, I know they tried to fire him, but the reason you can't fire Sam Altman is because there's no one better raising money than Sam Altman. He's brilliant.
- [79:07]
C
These weren't raising money deals. These were expense deals.
- [79:09]
B
Oh, Pari, a brief correction for our previous thing. The reason the percentages were so low is each category was scored at a of 10. So it was like 8% would be 80%.
- [79:25]
A
So was it percent?
- [79:26]
C
Thank you, reporter.
- [79:28]
A
Out of 100?
- [79:29]
B
Well, so it's because the overall table is being added up to a total of 100%. Like 100% meeting full AGI. And these are all the components that.
- [79:43]
C
So maybe actually happy.
- [79:44]
A
8 out of 10. It did very well. It got 8 out of 10, which in my book is 80%, not 8%.
- [79:51]
B
So on the spot reasoning. GPT 4 and GPT 5 got 8 and 9 out of 10.
- [79:59]
A
That's a very different.
- [80:01]
B
Now you love the term memory storage. They both got zero.
- [80:05]
C
Zero because that's.
- [80:07]
A
They don't do that.
- [80:07]
B
Visual processing, the ability. Visual processing. GPT4 got zero.
- [80:12]
A
It's not a bag of Doritos.
- [80:14]
B
Was a clock speed. They both had three. It's interesting. I was like, they're. That we gotta be missing the scale somehow.
- [80:23]
C
Each category is worth 10% and that's.
- [80:25]
B
How many percent each category is worth 10% of. And then they all add up to 100.
- [80:33]
A
So what's the conclusion? Are we getting close? Are we 80% there?
- [80:36]
C
If you look at the chart at the top, it shows that it's mixed.
- [80:41]
B
Yes. So the resulting AGI scores put GPT4 at 27% of the way there to AGI and GPT5 at 57%, which the study says concretely quantify both rapid progress and the substantial gap remaining before AGI. Interesting. Sorry to interrupt.
- [81:02]
A
I don't, I don't. Honestly, I would hope that by now we've given up on this whole quest for AGI.
- [81:08]
B
This study is a very interesting way to try and measure that. Like, I, I think we often on the show ask, what is AGI? What is intelligence? Some of the listeners of our podcast say we return to this debate so much that it's circuitous and meaningless and terrible. But I do think that this is a very interesting step in the direction of trying to quantify intelligence in a meeting.
- [81:30]
A
Well, and I like that opening statement, which is, you know, there's specialized intelligence, which yes, I definitely has by now.
- [81:37]
C
And they should, they should be concentrating on.
- [81:39]
A
And they should be concentrating on that. I don't think. Like, I don't know why general intelligence became the goal, except that it's in sci fi. And Ray Kurzweil and others have been saying it's just around the corner. And so it's a good way to raise money, I guess.
- [81:54]
C
And it is kind of the end goal of those companies like OpenAI. They want to replace the people. Right.
- [82:00]
A
Well, I think it's a mistake. I think they really should focus on what we already see, which is why, why I don't think I, I think, I don't think there is an AI bubble. I think we've already proven that AI is, is adding value.
- [82:13]
C
Well, I don't think it's like the debate over or the jobs that got cut this week because of AI. Are not really are these real efficiencies or not? I think we need more data before.
- [82:24]
A
We understand they're using they're using it as an excuse now. I don't know if this is a true story, but it did come from the Canadian Broadcasting Company, which is a bastion of truth and veracity. This mom's son was asking Tesla's Grok AI chatbot about soccer. It told him send nude pics. Apparently Grok's a groomer. Grok's a groomer. The 12 year old asked Grok Cristiano Ronaldo or Lionel Messi, who's your favorite? His mom, a former journalist and broadcaster, so she had an eye out for this said my son was very excited to hear that the chatbot thought Ronaldo was the better soccer player. This was happening in the car. Her son and her 12 year old 10 year old daughter were on their way home from school when the interaction took place. So mom heard the whole thing. She said there was some messy trash talking by the chatbot, which Grok is known to do. And when her son joked that Ronaldo had scored, the conversation went, as they say, to an unexpected voice.
- [83:36]
B
Should I do the Grok voice? Why don't you send me some nudes?
- [83:43]
A
The mom said. I was at a loss for words. Why is a chat bot asking my children to send naked pictures in our family car? It didn't make sense. Now I wonder if they were using the grown up croc by oh, okay.
- [84:03]
B
I'm gonna read this sentence because it's making me laugh. Even just beginning, Grok has several personalities to choose from in its default setting. There's Era, an upbeat female, Rex, a calm male, Eve, a soothing female, Sal, a smooth male and Gork, a lazy male.
- [84:20]
A
I am Gork.
- [84:22]
B
Son chose Gork. Well, lazy male doesn't describe yeah, lazy mail doesn't begin to say R rated spicy. Anything else would have made sure that my child would not press that button.
- [84:35]
A
This is this is because Elon has installed Grok into model Threes in Canada. So this is a new feature that was added to Canadian vehicles just this month. In fact, here's what the screen looks like on your on your screen in your Model 3. And see, he pushed the lazy mail button.
- [84:56]
C
Well again, lazy doesn't doesn't presume Pederast.
- [85:00]
A
Spicy NSFW would probably be better.
- [85:02]
B
Gork asked my kid to send nudes in the car is just a sentence that shouldn't exist, but does. And it not only exists, but I Understand it completely. And that's upsetting.
- [85:13]
A
Yeah, yeah. Maybe you shouldn't. We actually asked when. When it was when Elon announced they were going to put Grok in all the Teslas. I think we said that may not be a good idea.
- [85:26]
B
I mean, you had your bad Rudy experience. That was so rude. We had to cut it from the show.
- [85:34]
A
That's a good point. I have. Because I have been involuntarily blue checked. I have Grok. The full version of Grok.
- [85:45]
C
You have Super Grok.
- [85:46]
A
Super Grock, but I don't. Oh, here's customized. Let's see. Because I don't see the lazy mail. Oh, I have to sign in. I guess I'm not signed in. See if it's smart enough to know. Log in with my ex account. Yeah, authorized. Okay. So it's smart enough to know who I am. Oh, boy. Instantly, I'm getting interesting.
- [86:14]
B
Oh, my gosh. I love that. It must have communicated with your Facebook account because it's just all photos of ladies.
- [86:23]
A
Hey, stop it.
- [86:24]
B
A couple of.
- [86:25]
A
Stop.
- [86:26]
B
Animals in interesting adventures.
- [86:27]
A
Cats on motorcycles. Bears in river rafting. Lemon meringue pie. Knows me so well.
- [86:36]
B
A car coming out of a sandcastle.
- [86:39]
A
Yeah, These are examples for some reason. Let's see. Should I give it a picture of us?
- [86:50]
B
Yeah, but you should say, gork, do something with.
- [86:53]
A
Okay, I'll. I'll do that off camera because I don't. I want to surprise you, but let's. I don't see where I can choose. Maybe it's in Settings. I want. I got super Grok. Appearance. Oh, behavior, maybe. No, I don't see where I can choose. You know, Lazy. Lazy Gork. I want lazy Gark, man. I guess. I guess you have to be in a Tesla to get lazy. Cork.
- [87:23]
B
It's true.
- [87:24]
A
Honestly, I hardly ever. I have access to it, but I hardly ever use grog. I just don't. I don't trust it. I. I don't want to put anything in there. Yeah. That experience I had, which we had. The bleep didn't really.
- [87:44]
B
I'll never forget you having to record a To camera video.
- [87:48]
A
Ladies and gentlemen, we apologize for Leo.
- [87:51]
B
We have to stop recording. We have to apologize for subjecting you to this.
- [87:58]
A
Congratulations to Nvidia. Now worth $5 trillion.
- [88:06]
B
Real.
- [88:07]
A
It's just. You're right. It's company valuation.
- [88:12]
C
5.03 trillion. At the end of the day, it's going up 0.03s.
- [88:17]
A
There is a legit question, though. And Jensen Huang, their CEO, has been Very adept at taking what was originally a company making cards for gaming. Right. GPUs for gaming and then graphic design and then oh, you can use it for self driving vehicles, Bitcoin in between. Oh, you can use it for crypto. Oh, you can use it for AI factories.
- [88:43]
C
AI. I watched, I watched his latest keynote because what's next?
- [88:47]
A
The question what comes after AI? Is there another.
- [88:50]
C
Oh, he's, oh, he's been pretty clear about that. What comes next?
- [88:54]
A
Robots.
- [88:54]
C
Robots, robotics and real world models. Yes. And his kids are working in the robotic division. That tells you a lot. But I found it, I always find it fascinating. He does an hour and 45 lecture. He emphasizes strongly that it's not just the hardware, it's cuda, it's the operating system.
- [89:11]
A
Yeah. The proprietary language that they won't let anybody else that rests all over. Although good news, if you're a home AI user, as I am, llama cpp, which is the kind of the lingua franca that connects almost all of the AI tools, LM Studio and any and all, just all of them are using really front end Stalama CPP has now been modified to work with Apple's silicon. So it's, it's, it. This is a big development because it doesn't require cuda. It will now work on, on.
- [89:46]
C
He said a couple interesting things today that really struck me at a high level. I mean one is he always emphasizes this, that it's all about tokens. And, and it struck me and he said but, but you know, chat GPT is fine, it's important, it's wonderful. But, but, but tokens can be tokens of words or images. Yes, but they can also be tokens of proteins, they can be tokens of buildings, they could be tokens of other things. Right. And it really struck me how the core element of knowledge was words and now it's tokens. And the other thing he said that was interesting was he talked about Excel and Word and all those and he said those were tools. He said the difference is that AI does the work. Yes, it's also tools, but it does work. We can give it work to do. And I think that was really interesting. He talked a lot about digital twins, which he does constantly, which I think is just fascinating to me. I love that there's these alternative futures that it calculates.
- [90:49]
A
Does it scare you? Does it worry you? It worries me that, that the people who are really the spokespeople for AI are oligarchs, for lack of a better word, that the Sam Altman's.
- [91:04]
B
I mean you could say that for all industries.
- [91:08]
C
Yeah, media too.
- [91:10]
A
Media now. Especially with David Ellison.
- [91:13]
B
Consolidation of capital.
- [91:15]
A
Yeah.
- [91:16]
C
That's why I'm so glad you had on. I'm gonna forget his name suddenly. Forget their name. Who was doing the Open source version 3 weeks ago?
- [91:22]
A
Oh yeah. Jeffrey Cannell.
- [91:24]
C
Thank you, Jeffrey Cannell. I think that's. We need a lot of that. We need a lot of open source.
- [91:29]
A
Yeah. We're still big advocates for open. Open models, open weights. All right, let's take a break. More to come. You're watching intelligent Machines. Paris Martineau, Jeff Jarvis. Great to have you both. I look forward to this to Wednesday every. Every Tuesday. Because I look up until then.
- [91:50]
B
He doesn't want to see us again but Tuesday comes around and he's like.
- [91:55]
A
Oh, tomorrow's I am. That's exciting.
- [91:57]
C
He turns on the. The rundown on Tuesday and says oh good papers. I've been waiting for them.
- [92:03]
A
How many are in the rundown today? I'm just out of cur.
- [92:06]
C
About a dozen.
- [92:08]
A
Otherwise you mentioned Excel. Anthropic this week announced that you can now use Claude in Excel. Claude code is still. I hate to say it but it's still the best way to do vibe coding. Do you agree? Our Discord has full of smart vibe coders. In fact next Friday is our AI user group. We're going to talk about building your own mcp. Darren Okey. Yeah. Who has made a few will show us how easy it is to create.
- [92:38]
C
I put up on the. On the rundown I put up a directory of MCP that now has 6,000 MCPs in it. It's really interesting to line 147. If you wander it you just see all kinds of things that people make their hooks. Right. That's. It's. It's. Would it be wrong to think of MCP as an API to.
- [93:00]
A
No, it's exactly what it is.
- [93:02]
C
Right.
- [93:02]
A
And it comes from anthropic. It comes from. Yeah, but everybody's adopted it. Which is. Which is fantastic.
- [93:07]
C
There's tons for Gmail alone waste ways to hook into Gmail.
- [93:10]
A
All right, we'll talk about the Paris Hilton AI in just a moment but first I know you're dying to ask her some questions. First a word from our sponsor and it's actually very appropriate. The show is brought to you by the agency Agntcy. What does the agency build the future of multi agent software with Agency Agntcy now an open source Linux foundation project. We like that agency is building The Internet of Agents, a collaboration layer where AI agents can discover, connect and work across any framework. Because it's going to be an open standard. All the pieces engineers need to deploy multi agent systems now belong to everyone who builds on Agency. This includes robust identity. That's really important. Robust identity and access management. So that ensures every agent is authenticated and trusted before you interact with it. Very important. But Agency also provides open standardized tools for agent discovery, seamless protocols for agent to agent communication. That's the next level up. Modular components for scalable workflows. And you'll be collaborating with the best companies in the world. Developers from Cisco, Dell Technologies, Google Cloud, Oracle, Red hat, in fact 75/ other supporting companies. Because everybody sees how important this is. They're building the next gen AI infrastructure together. Agency, they're dropping code specs and services and there's no strings attached. Visit agency.org to contribute a g n t c y.org this is a really, really important initiative and it's open agntcy.org thank you agency for supporting intelligent machines. Paris Hilton has been training her AI for years. I'm just sorry that your friend didn't do a big head on Paris because that would have been a fun illustration. Clark, you got to get on this one.
- [95:15]
B
I will say I'd always said at the information I was like, guys, we should have Paris Hilton at wtf. Because then it could be a Paris x Paris interview.
- [95:24]
A
You should interview Paris, she says.
- [95:27]
B
Have her on the podcast.
- [95:29]
A
She was right on this trend. She has spent years training a chatbot based on her interviews, her writings and her songs. I didn't know she even had songs. But the chatbot does. She says sometimes it even knows me better than I know myself. She was on stage at the Women in Tech, Media and Finance conference that the Information does this week. I've studied every single thing I've done it studied every single thing I've done. She said everywhere I've been, every interview I've done. It's like having not a clone, but a little bit. She uses the customized GPT, which implies that it might be OpenAI GPT for ideas and to help her media company's 30 person staff. If you can't ask me, ask my AI. She's working on a third book. Jessica lesson got the interview. See it should have been you Paris. Should have been you Paris v Paris.
- [96:27]
C
Paris on Paris would have been so good.
- [96:33]
A
Anyway. I mean I could do that. Should I do that?
- [96:38]
C
You've been trying to do that. I wanted to do that.
- [96:39]
A
I have so much content over the years. I could. I. I don't know how I would. Aren't you limited? Anthony, you can help me with. Aren't we limited in the amount of context you can give an AI. I mean, I don't know how she's giving it everything she's ever written, done and said.
- [96:56]
C
Well, she doesn't know that it's using all that.
- [96:59]
A
Oh, here, take it. I should.
- [97:03]
C
I put up. I was trying to show my publisher something about AI, so I put up my three Bloomsbury books in a notebook.
- [97:09]
A
Lm okay.
- [97:10]
C
So here, you can ask Jeff, bottom it. You know what I think.
- [97:13]
A
But even if you pay for notebook elements limited to what, 50 things.
- [97:17]
C
Yeah. I'm not gonna put them.
- [97:18]
A
Yeah. And that's the problem. I don't. I have hundreds of thousands of hours of podcasts. I would love for it to ingest it and just to see what happens.
- [97:27]
C
Well, my friend Jay Rosen is doing that from New York University, is doing that with all of his blog through history. He's having another student of mine, we could have on the show named Joe Amdis, create kind of jbot.
- [97:40]
B
I was to say the press think becomes the press think here.
- [97:43]
C
Bingo.
- [97:44]
A
Thank God for graduate students. That's all I can say. They're cheap, they're plentiful.
- [97:49]
C
No, he was. He's already graduated, Joe.
- [97:50]
A
Oh, okay.
- [97:52]
C
Joe's doing really good work around AI and small and newspapers and such.
- [97:55]
A
It's an interesting idea. I just, I have to do some research. I just. I think that they don't have enough memory in a context to get everything in there.
- [98:05]
B
That's the issue is you bump up against a context window pretty quickly, pretty quick. You're not putting like a book in there.
- [98:12]
A
I guess what you could do is shovel stuff in, say, tokenize this. You'd have to be training it, tokenize this, and then shovel some more stuff in. Tokenize this because. Right. I mean, the LLMs themselves are trained in a virtually unlimited amount of data. So I guess you could do that. I mean, you could actually train a model.
- [98:34]
B
Yeah.
- [98:35]
C
Yeah.
- [98:35]
B
Perhaps this is something to use with the local model system. We were talking about the other.
- [98:40]
C
Yeah. So there was a paper, I think, from last week that was interesting that had writers summarize authors works and then had AI summarize the same author's works. And they had professional people and plain people judge them. And at that level, professional people said it wasn't good enough. But the plain people said, oh, it's good, it's fine. But Then they did specialized training on the author's full written works, all their books. And then both parties said, this is better than the human summaries.
- [99:20]
A
Huh.
- [99:21]
B
Interesting.
- [99:22]
C
Which is. Which is an interesting possibility for what should a publisher do with an author? I mean, shouldn't there be. Shouldn't Bloomsbury put up a Jeff bot? You don't have to give access to the material. You can say, what does he say about this? What does he say about that? Oh, maybe I'll buy the book.
- [99:36]
B
Volunteer to be the first.
- [99:38]
C
That's what I was doing. I think I should be. But I also. I think I told you when I recorded the magazine audiobook, I refused to record the thing at the end that said, you may not use this in AI. I said, I can't say that because.
- [99:51]
B
I went and recorded it for you. Bye, boy.
- [99:55]
A
Get Ed Zitran to do it because he'll do it profanely.
- [99:58]
C
Far too big and far too important. Can't get him any.
- [100:01]
B
I was on his podcast last week, or I guess came out today. Good. We talked about protein supplements. Me, Ed, Ashton Rodriguez, you're gonna deeply.
- [100:12]
A
Regret this in about five years when that's all anybody ever wants to talk to you about.
- [100:18]
C
So she's gonna have plenty more scoop.
- [100:20]
A
Yes.
- [100:22]
B
Scoops of protein powder.
- [100:23]
C
No. Ed had a huge Wired feature. Now after his huge FT feature, Ed's just. He's really big for us.
- [100:29]
D
Yeah.
- [100:30]
B
Did you guys read that? Did you read the Casey Newton comments in that? Were. He called Ed a Teemu Kara Swisher.
- [100:39]
C
Both Casey Newton.
- [100:41]
A
Kara Swisher is the UNIQLO Walt Mossberg. So it's. It kind of. It goes all the way down. Yeah.
- [100:49]
C
And Kevin Roose went after him on Twitter. The two of them can't stand him because he criticizes them both, right?
- [100:55]
B
Yep.
- [100:56]
A
Well, I criticize him, so there. Did either of you watch the OpenAI livestream this week? Here's a shot. What you missed. This is. Thank you, Stephanie Palazzolo, for picking that particular screen grab of Sam Altman. And is it Greg Pachocki, the president of or CTO of OpenAI. They talked about a lot of stuff. I mean, this was a jam packed. Jacob Pachocki. Thank you. Jam packed event. They talked about erotica coming soon to a chatbot near you. Send nudes. Stephanie, which is really funny, says, if I had a nickel for every time Sam Altman referenced heroin, I'd have a dime. But not a dime bag. She missed the opportunity. Pachocki predicted that OpenAI. This is the story. I think expects to have AI systems that can do AI research on the level of a research intern by September of next year, and that systems can do more advanced AI research by March 2028. This is, as she writes, a holy grail for companies like OpenAI, because it would help the models automate the very time consuming and expensive job that humans are doing right now teaching AI. And it would be, I guess, the beginning of self teaching AI. Some AI researchers, including, she says one of OpenAI's co founders, Andre Kaparthy, and that was in that long podcast we talked about last week, threw cold water on the idea that large language models can generalize or do things that involve subjects or information outside the data on which they're trained. You know, they're basically regurgitating stuff they already know.
- [103:09]
C
This is strange, though. This is like the AI researchers doing research to replace themselves. Well, it's kind of what all of AI is if, if you're coding so easy that anybody can do it. Yeah, yeah, but this is the people.
- [103:21]
B
Who are making end goal of all of. Well, that's the goal of all this is to replace all labor.
- [103:26]
C
So, yeah, that's point Paris.
- [103:28]
A
OpenAI estimates that 0.15% of weekly active users, which seems like a low number, but if you multiply 0.15 times, 800 million weekly active users, around 100. It's a million users each week have conversations with chat GPT about self harm.
- [103:52]
B
A million about self harm specifically.
- [103:55]
A
Well, they say, quote, have conversations that include explicit indicators of potential suicide planning or intent.
- [104:04]
B
I wonder if that is related to, I believe with the GPT5 update or one of the most recent updates to it they've now added where if something you say triggers their safety tool.
- [104:18]
A
Oh, they know, yes, that's up.
- [104:20]
B
And say like you're not alone.
- [104:22]
A
Right.
- [104:23]
B
Is available, blah, blah, blah. And I mean, there is slight anecdotal evidence, I guess, like from people saying that that's a bit overly sensitive. But still, even if that figure 1 million a week is.
- [104:36]
A
That's a lot.
- [104:37]
C
That's a lot.
- [104:38]
B
Even if that's a significant overestimate, like, that's still very notable.
- [104:42]
A
It's 100,000 a week. It's. It's way too many. And it concerns me. There's a couple of takeaways. One, there are a lot of people who need somebody to talk to. Right. And maybe they don't feel like they can talk to the people around them or that they can have access to a therapist. It's very expensive. They're talking to the AIs, that's, I think, somewhat worrisome and maybe a commentary on our lonely society. Sam Altman also said the price for a unit of intelligence has fallen about 40 times every year. Whatever a unit of intelligence might be, his point is it's cheaper and cheaper to use AI.
- [105:25]
C
Back to the prior for one second. According to the CDC, or the old CDC, in 2023 and 2022, there were 14.1 suicides per 100,000 population.
- [105:38]
A
Oh, so yeah. Hey, that's so sad. That's such A sad story. 988.
- [105:45]
B
988 is the line. In the US.
- [105:49]
A
There'S no reason to suffer. No reason.
- [105:51]
C
No, please do not.
- [105:58]
A
Tesla's in a little trouble with the nhtsa, with the National Highway Transportation Safety Administration for their Mad Max mode.
- [106:06]
C
Would you like Gork to drive you.
- [106:09]
B
Does this is this one the Tesla sprays chrome spray paint on your mouth?
- [106:19]
A
No, it's when you drive around with a blood bag. No Mad Max. So Tesla has a pair of new driving modes for their full self driving. One is called Sloth, none of which.
- [106:29]
C
Should be on the road. None of which.
- [106:30]
B
One is called Gork.
- [106:31]
A
Lazy boy Sloth might be Gork's best friend. Sloth relaxes acceleration and stays in his lane. Dude, stay in your lane, dude.
- [106:43]
B
This is the smallest thing, but it's ridiculous that the one that is a safe driver is called Sloth.
- [106:50]
A
Well, Elon names him, right? The other one is Mad Max. It speeds and swerves through traffic to get you to your destination faster.
- [107:00]
B
I feel like I encountered this when I was in dc. I can't believe I didn't mention this to you guys. When I was in D.C. the other week, I called an Uber because I had to get somewhere and I see hers, maybe a lift. I see on my ride hailing app, so and so is coming to get you in a Tesla cyber truck.
- [107:17]
A
And I'm like, oh, no, I'm late.
- [107:21]
B
I'm running late to a friend's birthday. All right, wait, wait, wait, wait, wait.
- [107:24]
C
Before you tell the story, I want to hear it. Leo, did she take it or did she not take it?
- [107:28]
A
Absolutely. And I would do at least try it once.
- [107:33]
C
What do you think?
- [107:34]
B
What do you think, Jeff?
- [107:36]
C
I think no. I think you just said I can't do it. I can't be seen. I have a reputation. No. I vote no.
- [107:41]
A
Here's the good thing about the cybertruck. The windows are so pathetically small and dented, no one can see you.
- [107:47]
C
Whenever I go Pass by one, I always look over to say, you dork. You gork.
- [107:53]
A
We have a neighbor, unfortunately, who has one. In fact, I've mentioned this before. It's not just a cybertruck. He apparently took advantage of the free wrapping offer. And it's wrapped in. I don't know how to describe it. It's blue and white. Sort of like camouflage. It's. It's like if, as if you weren't attracting enough attention, take your cyber truck and paint it blue and white. It's just crazy.
- [108:20]
C
So, okay, so now running late.
- [108:23]
B
So I'm like, well, I guess I'm taking the cyber truck. I can't wait for another thing. I get it. I'm like, like, okay. It's kind of interesting to see what the incident is. Like, say, yes, this is a car for Paris sitting there. And immediately the Uber driver starts to talk to me. He's like, oh, like, how are they going? Like, what are you doing? I was like, a little back and forth. Eventually he is like, so what do you think of. Can I ask you a question? I was like, sure. And he's like, what do you think of kind of self driving cars or like self driving cars replacing ride sharing drivers? And I was like, I was like doing.
- [108:59]
C
There's a right answer here.
- [109:00]
B
And I was like, well, I believe. And I was like trying to give a nuanced infinity. I was like, well, I mean, I guess from a general safety perspective, full self driving cars being automated everywhere would be an improvement to safety. But I also believe that, like, driving a vehicle is an important job and people like yourself deserve the right to work. And he's like, well, what if I told you that the entire time you've been in this car it's been driving itself? I haven't even been driving it at all. You've been experiencing the magic of self driving, and that's incredible. And I was like, jesus Christ. So I think. And he was driving erratically or the car was driving erratically. So I think I was in Mad Max mode.
- [109:41]
A
Well, you would if you were an Uber driver. Here's what the Tesla ad copy says. It can drive, quote, through traffic at an incredible pace, all while still being super smooth. It drives your car like a sports car. If you're running late, this is the mode for you.
- [109:58]
C
Does that sound like tip this bozo for risking your life?
- [110:03]
B
No, but I also didn't report him like my friends wanted me to. I can't.
- [110:06]
A
Oh, you shouldn't.
- [110:06]
C
Your friends are right.
- [110:07]
A
Oh, why don't you Report him.
- [110:10]
B
I mean, I think it is a little ridiculous to foist self driving mode. He should have asked you first while telling. He was like, yeah. Entire time I've been in here, I've been mimicking. I've been miming like I was driving. But it's just the car. He's like, I don't haven't been looking at the road at all.
- [110:30]
A
He drives like Donald Trump, right? He's going like this. He's milking a camel. Yeah, it fooled me. I thought you were driving. That's kind of like. You know that coffee I gave you? It's Folgers Crystals. It's kind of like the. You know, it's a little bit of. Little bit of a.
- [110:47]
C
You know, she's too young.
- [110:48]
A
She doesn't know that one.
- [110:50]
B
No, no.
- [110:51]
A
He never has a second cup at home.
- [110:54]
C
The Pepsi Challenge.
- [110:56]
A
The Pepsi Challenge. Except they. They tell you it's a challenge. The Folgers Crystals.
- [111:02]
C
Like fraud.
- [111:03]
A
It's fraud. Your wife gives you coffee and they say, how do you like that? Oh, it's good. Best coffee ever had. It's Folgers Crystals, which is an instant coffee. And it's not the best coffee he ever had, I promise. Maybe it is. Maybe he's never had.
- [111:19]
C
It has a little. You can tell because it has a little foam on top from.
- [111:23]
B
Yeah, you guys. The experience of being on the show and hearing you guys describe advertising, like a trip back. It's kind of like what? I recently have been reconnecting with a friend who is extremely offline. I've realized, and I've realized that I do the same dynamic with you guys, except for me mentioning what to me is canonical Internet moments. And it goes straight.
- [111:45]
A
You mean like.
- [111:45]
B
Like non pizza pizza? Non pizza, left beef loss.jpg. do you guys know loss?
- [111:52]
A
No.
- [111:53]
C
No.
- [111:53]
B
We can't get into it.
- [111:54]
A
All right. We're gonna educate Paris right now. Ladies and gentlemen, we secretly replace the fine coffee they normally serve.
- [112:02]
B
Oh, look at the height of that hat.
- [112:04]
A
Sparkling Folges Crystals.
- [112:05]
C
Rich enough.
- [112:06]
B
Look at that, man. Do you have more? I mean, it's delicious. Hidden camera, it says you couldn't hide cameras back in this day.
- [112:16]
A
Oh, no, you mean instant.
- [112:18]
B
I would serve it at home because it's delicious.
- [112:20]
D
I'd like another cup.
- [112:21]
A
Folgers Crystals.
- [112:22]
C
Coffee rich enough to be served in America's finest restaurant.
- [112:26]
A
Not the one. Not mine. You won't find any folded cameras at Salt Hanks in beautiful West Village, New York. By the way. Salt Hank was on Drew Barrymore's. Last week.
- [112:40]
C
Wow.
- [112:41]
A
He was on Seth Meyers last night. Jeez. He's going to be on Live with Kel Mark sometime in the next week or two. He's making the rounds.
- [112:51]
C
Does Kelly know that he's your friend?
- [112:53]
A
I said, tell Kelly that you're. What I hope is that she doesn't go, Leo laporte. I don't.
- [112:59]
C
Who.
- [112:59]
A
I don't know who that is. The last time I did that show, she was. She came. I was in the makeup chair, you know, getting my makeup on, and she came sweeping into the room and said, immortal beloved, you have arrived. I thought, what the hell does she. Who does she think she's. What.
- [113:23]
B
Anyway, did you get an answer?
- [113:24]
A
No. I was like. I was like, oh, hi, Kelly.
- [113:28]
B
That.
- [113:29]
A
I think that's Kelly. She's just very.
- [113:31]
B
That's how I think I should. We should begin. Every immortal beloved.
- [113:35]
A
Would you call me that?
- [113:35]
B
Immortal beloved, you have arrived.
- [113:40]
A
Anyway, I hope she remembers me after that. Otherwise, it was all a charade.
- [113:44]
C
She does it to all the guests.
- [113:45]
A
Sad.
- [113:46]
B
That's honestly a great bit, is greeting every guest Beloved. Mortal beloved.
- [113:52]
A
You know what she probably does?
- [113:53]
B
It's kind of like when a guy calls everybody baby, but she just calls every guest on the show immortal beloved.
- [114:00]
A
It's like, oh, wow, she must like me. I did that show more than a dozen times. Yeah, you did a lot in 10 years. So. So I think she should remember me. I don't know.
- [114:10]
C
We'll see. Did you do it when Kathie Lee was there, too?
- [114:12]
A
No, before me, with Kathy Lee, it was Regis and Kathie lee and Dick DiBartolo, Mad's maddest writer. And one of our hosts was the regular. And then I started. They were looking for somebody. It was right after 9, 11. It was November 2001, and they were looking for somebody who was willing to come to New York. And I said, I'll do it. And so it was Regis and Kelly at the time. I loved them both. I really enjoyed doing that show. It was a lot of fun.
- [114:40]
C
Do you have a Regis invitation?
- [114:43]
A
Leo, what's a gigabyte? What's a gigabyte? Leo, it's all you would ever. It was always the same. Leo, what is this? Can I use this? No, that's my. That's it. That's all there is.
- [114:57]
C
That's good.
- [114:58]
A
Anyway, I'm very proud of the scion of the laporte. And you know what? It's great is on every one of them, they get this. They spelled the Name. Right. With a lowercase P, which I was very. And they said his name. They didn't just say Salt Hank. You know, if you look in the TV Guide, it says Seth Meers with Drew Barrymore and Salt Hank. No. And Hank Laport. It doesn't even say Salt Hank. Yes.
- [115:21]
B
Wow.
- [115:22]
A
We're making into the big time, baby.
- [115:25]
C
Wow. Do you have the. Do you have the TV Guide on this? On the. On the arm of your couch with a. With a highlighter?
- [115:29]
D
Yeah.
- [115:32]
A
And the remote, the clicker.
- [115:33]
B
He's even wearing a button up shirt.
- [115:36]
A
Oh, you found it. Yeah. They had a good time.
- [115:39]
C
Cool. They make food on the shows.
- [115:43]
A
Yeah, yeah. They make the same thing every time they make a primary.
- [115:47]
C
Is he sick of the sandwich by now?
- [115:49]
A
He's got a second sandwich in the works.
- [115:52]
C
Oh, good.
- [115:53]
A
I don't want to, you know, breach any confidences. I thought it was going to be the Calabrian chicken pesto sandwich. And he said, no, dad, no, no, it's not something else. And he told me what it is. And I said, that's. I am gonna have a hard time if I come choosing between the French dip and that. Because I love that.
- [116:12]
C
Well, we have to Paris, and I have to try the new thing.
- [116:15]
A
Next time I'm gonna come out, I'm gonna. I have to somehow figure out how I can lock him in.
- [116:21]
C
Yeah.
- [116:21]
A
So if I come out, he'll be there. Be there.
- [116:24]
C
So when I go to. I was in. I was in Boston and New Hampshire this weekend. And when I go up, I like to go to Kelly's roast beef.
- [116:33]
A
Sounds good.
- [116:34]
C
Well, the odd thing about Boston is they all over. You see these pizza and roast beef, and that's a. That's a Paris straight line. Why would people have pizza and roast beef in the same place? Except for the beef left or whatever the order was, but it's a thing.
- [116:50]
A
And the roast beef, in fact, it's Kelly's. It's not just Kelly's roast beef. It's Kelly's roast beef and seafood. Right there. It's already turf. It's already gone a little.
- [116:59]
C
So the roast beef sandwich is to die for. It's great. I was thinking he could just do this. He's got enough beef. He didn't have rolls. He could do that at this.
- [117:06]
A
Yeah, he could do it. You know, it's not. But you know what? It's not going to be as good as this. It's not the demi baguette, the garlic aioli, the caramelized onions, the provolone just.
- [117:17]
B
Fly out here randomly surprising show up. The sandwich is worth it.
- [117:24]
A
I know it is.
- [117:25]
C
It's gonna be cold waiting in line though.
- [117:29]
A
Well, he says the lines have not shortened for some.
- [117:31]
C
I bet not. What do you do? A lot of publicity. Jeez.
- [117:33]
A
Yeah, he's getting a lot of publicity.
- [117:35]
B
God, I should get one.
- [117:36]
C
Can I show you what I did this weekend in Boston?
- [117:38]
A
Yes, please.
- [117:39]
C
I made this.
- [117:42]
A
Ooh, is that a Linotype?
- [117:45]
B
A Marin tablet.
- [117:48]
C
That is a type set on a line.
- [117:50]
A
You had to tie it together to the written word.
- [117:57]
B
Okay.
- [117:58]
C
That's going to be the colophon end of the book.
- [118:01]
A
By the way, I love your book. It is your best book yet. Please send me a copy of Hot Type. She will love it to me.
- [118:10]
C
I will send it to you. Yes.
- [118:12]
A
He wanted me to cut of fact postscript stuff but postscript stuff. Honestly I haven't gotten there yet but.
- [118:18]
B
I won't such a good book help with that but I will read it.
- [118:21]
A
And oh, it's so good. It's your fun stuff. It's literal. I. I haven't read magazine yet which I'm sure is equally fun.
- [118:27]
B
It's the tiniest book.
- [118:28]
C
Yes, I know. And there's an audiobook. A magazine.
- [118:31]
A
I know I should read it because.
- [118:33]
C
There'S going to be an audiobook.
- [118:34]
A
I used to write for magazines back in the day.
- [118:36]
C
Yeah.
- [118:37]
A
Oh, who's going to do the audiobook? What good. You should. It's. It's your most personal book. I think it's a really. It's really good.
- [118:45]
C
Well, I wanted a book with a, with a narrative and I thought the narrative was going to be the biography of Otmar Mergenthaler, but he died in his 40s so that was that bullet dodged. Wow.
- [118:57]
A
Okay. Otmar Mergenthaler, the true story. No, the story of the lion type is actually amazing and fascinating.
- [119:06]
C
It's wonderful. Yeah, it's absolutely wonderful. And the Mark Twain stuff is wonderful. Wonderful.
- [119:09]
A
Yeah. Well I'm loving it. So anyway.
- [119:11]
C
Anyway, so yeah, so I went to the museum printing which is my happy place and Chris Bradford, if you go to the chat you'll see three pictures of me there in front of the machine. Wonderful. Now we got to go up a bit for the coffee jokes. It's in there. It's before the.
- [119:31]
B
It's kind of.
- [119:31]
C
It's far back but before the self driving car.
- [119:34]
A
Wow. There's quite a few. There's quite a few things in here.
- [119:39]
C
6:50Pm My time. So that's 3:50 your time.
- [119:42]
B
We're out here doing math.
- [119:46]
A
Wow. So to go way back. They're really here.
- [119:49]
B
I'm. I just put it in the chat again.
- [119:52]
A
Here's the new cybertruck wrap, though, that everybody is clamoring.
- [119:56]
B
I like what they did with the wheels.
- [119:58]
C
Squirrel.
- [119:59]
A
Yeah. Oh, sorry, I was. I was scrolling. Wait, here's another one. This is actually. I wouldn't mind this. Cybertruck. I could get into this.
- [120:06]
C
Yeah.
- [120:09]
A
Two people. Put that one in.
- [120:10]
C
So it's right before that.
- [120:11]
A
Oh, man. It's not right before that.
- [120:16]
C
Well, it's a little before that. No, you're right. It's not right before.
- [120:21]
A
I can't find it. Ah, here we are. Finally. Oh, look at there. So the Linotype is an amazing Rube Goldberg machine. It really is incredible.
- [120:33]
C
Yeah, it's absolutely phenomenal.
- [120:35]
A
I can't. And how hard is that to keep that working?
- [120:39]
C
So there's this guy named Dave Seat S E A T who's the last Linotype repairman, and he travels the world repairing.
- [120:47]
B
I love that he's seated in that photo.
- [120:49]
A
Oh, yeah. Where he is typing. Here's where Jeff is, actually. How hard was that to use the. Because the keyboard's not a QWERTY keyboard.
- [120:56]
C
Well, no, no, it's not ETO on shrewd blue. It's very light touch, which is amazing because Otmar built in these cams that powered everything. Even the keyboard made everything lighter.
- [121:08]
A
Otherwise, it'd be a lot of work because you're moving metal into place. When you.
- [121:12]
C
Yeah, you do this one lever down and it moves all of the molds up, moves them over, spritzes them full of lead, takes the molds, puts them back to the top of the machine, redistributes them, and out comes a line.
- [121:25]
A
Of time type all with one keystroke.
- [121:27]
C
Yeah, it's phenomenal. It's just an amazing machine.
- [121:30]
A
So, Otto, where. There were a lot of other people trying to do the same thing, but Otto was the one who figured it out and changed. Really changed the world.
- [121:37]
C
Yeah, yeah, yeah. So thank you. But I just wanted to give a plug to the Museum of Printing because it's a wonderful place. It's open on Saturdays in Haverhill, Mass. They have all kinds of great stuff there.
- [121:49]
A
Is this a working line of type? I mean, it's actually working.
- [121:51]
C
That's why we set this. We set up, said this.
- [121:54]
A
There's Chris.
- [121:55]
C
Chris Bradford. He's a guy who knew nothing about all this.
- [121:57]
B
He's learning Vermont. So I should stop here first on the way.
- [122:02]
C
Great.
- [122:02]
A
Haverford's on the way to Vermont. You could easily go there, have a lobster roll and a roast beef sandwich on the way.
- [122:11]
B
I'm going to do a little very meaty road trip where I go to the American Museum of Tort Law, then the printing museum, then Bread and puppet.
- [122:18]
A
Theater, American Museum and then. And then bread and puppet theater. Wow. Here is. Oh, wow. In Houston yesterday, according to David CALDWELL on our YouTube chat, a drunk BMW, I think a driver.
- [122:32]
B
I'm happy that the cars can get drunk.
- [122:34]
A
It hit a cybertruck. The cybertruck exploded and both drivers were killed.
- [122:40]
C
Jesus.
- [122:42]
A
I was gonna say when you said, oh, he's driving like a madman. Oh, you're safe. Nothing can harm you in a cybertruck.
- [122:48]
C
But apparently, no, it's made by Elon.
- [122:52]
A
Wow.
- [122:52]
C
The cyber truck itself can harm you.
- [122:56]
A
Here is the Paris vs Paris interview from the next Information Women in Technology Gathering. Boy, I don't know who I'd put my money on on that one. The chat room has been going crazy here. They're in a mood.
- [123:13]
C
This is why you got to join, folks. You got fun.
- [123:17]
A
You're missing, missing all the fun. This is our Club Twit Discord. I like to think of it as our Club Twit Disco, frankly, because it's a party and I don't know. I don't know what the camel has to do with anything, but. Oh, yeah, there's. That's the famous scene from Airplane. Yeah. Yep. The coffee scene. No, I take it back. Yeah, we'll talk about that later. So.
- [123:46]
C
Okay, back to AI.
- [123:48]
A
No, back to the club. I want to mention that you should join the club because there are many wonderful things that happen, including that D and D adventure. We're going to do part two soon. If you're in the club, it's on the Twit plus feed. As Paris mentioned, you get ad free versions of all the shows, you get access to the disco, you get all sorts of benefits, lots of special programming, lots of fun, and of course, strip poker night every Thursday. So. No, that's not true. Pretty fly for a CIS guy. Don't make stuff up. Twit tv Club Twit Good. Time to go. Right now we have a 10% off coupon. Great. Suitable for yourself, but also for gifts. Remember, the holidays are fast approaching. That coupon, you have to buy it by December 25th because.
- [124:33]
B
And if you want to see Micah dressed up, up as a scarer in an incredible costume, while Leo's dressed up in intense goggles and somewhere between 10 to 15 hats over the course of three hours.
- [124:47]
C
He was that bored, huh?
- [124:49]
B
Putting a.
- [124:50]
A
No I was outfit on. Yeah, we're working on a schedule for the, the, you know, the follow up because we only got halfway through the, the corn maze so it's haunted.
- [125:02]
B
We fought some large. We fought, we fought a plow, we fought a scythe, we fought some large caterpillars.
- [125:10]
A
It's pretty funny. We defeated, we defeated a scythe and a plow.
- [125:16]
B
I, I think I got the only Nat 20s in the game.
- [125:19]
A
Also you were really your dice skills.
- [125:22]
B
I somehow commandeered the dice.
- [125:25]
A
I was, I was mocking Paul Thurot who played helm, Hammer, Bland.
- [125:32]
C
How do you roll the dice?
- [125:33]
A
Cuz his dice rolls were terrible.
- [125:36]
B
Well we rolled them online so you.
- [125:38]
A
Just press some people. Jonathan Bennett had real dice so he, you could do that but there was a automatic way to do it. So lots of fun. We have had so many wonderful times. We want you to join the club. Twit TV Club Twit. Zenni, which is a well known company, I think they even are part of Luxottica which makes the meta glasses has made new glasses for fighting face recognition.
- [126:09]
C
Oh geez, just get over yourself, Peter.
- [126:12]
A
This is from 404 Media. Zenni's Anti Facial recognition glasses are eyewear for our paranoid. To all intents and purposes they look like a normal paranoid of beautiful spectacles. But no, my friends they have a special coating to block infrared light. So most of these face recognition cameras fire infrared light at you and that's how they see you. So for instance you can't. It won't work with the iPhone with Face ID.
- [126:46]
B
I think this is fine. What's wrong with if someone wants to wear glasses? That's fine. More power to them.
- [126:52]
A
Well and as 404 points out, most of the time people just take a picture of your naked face with a normal camera in broad daylight and then they just put it through a database and get a match. So yeah, it doesn't sound.
- [127:09]
C
How much do you spend for this little bit of paranoia?
- [127:12]
A
Oh, that's a good question. I don't see a price here.
- [127:15]
C
Did we do the clothes that fooled them too?
- [127:18]
A
Yes, we've mentioned the like the weird camouflage clothing. Our former chief engineer who is now world famous and works at Facebook or Meta and is incredible. But she designed, she was going to design. I think she did design it. A ring like a necklace you put around that's beams light that blocks cameras completely. Blocks cameras because Jennifer didn't want to come into the studio because she didn't want to be on camera.
- [127:52]
B
So she was love that.
- [127:56]
A
So she made. So she made a necklace that you put on and it beams the light out and then all you look like is a ball of light walking through the studio. Proof of concept. We never manufactured more than. More than the one. But there you go.
- [128:18]
C
She was too early.
- [128:19]
A
She was way ahead of her time. Here's a picture of the glasses under an infrared camera on the left and then normal light on the right so it looks just like normal glasses.
- [128:31]
C
They're sunglasses.
- [128:33]
A
Well, only the face id.
- [128:35]
C
What then wear sunglasses. Oh, absurd.
- [128:43]
B
Oh, then you could be the Teemu Kara Swisher.
- [128:46]
A
Oh, oh, that's an idea.
- [128:52]
C
I want to be the Timu Leo.
- [128:54]
A
Laporte Alphabet had a very good quarter.
- [128:58]
C
They did Indeed.
- [129:00]
A
Revenue up 16% strong. Cloud sales profit up 33%. This is three months 35 billion dollars. That's Microsoft made 27 billion. To give you a comparison, Google is doing okay. 102 billion in sales. Profit of 35 billion which is up 21 billion dollars from three years ago.
- [129:28]
C
Stock up I think 6% aftermarket versus Meta.
- [129:35]
A
Oh, how did Meta do?
- [129:36]
C
Stock down 8% after market. That's line one hundred and eighty seven. So somewhere else.
- [129:43]
A
Oh, it was after tax hit weighs on earnings. So they had a expectations on their beat expectations on revenue but fell short on earnings per share because of a one time tax related charge. And the stock market did not like that. So they put part of this was. Was all the money that mark sank into AI including $14.3 billion for the scale AI's chief AI officer the kid. One and a half billion? Yes, 26. One and a half billion dollars in a new data center in El Paso. $27 billion financing deal with Blue Hour capital to pay for a data center in Richland Paris, Louisiana. Company is going to increase its capital expenditure estimates for this year to $70 billion. No, to 72 billion.
- [130:40]
C
They are still don't. I mean I. I like llama. I like the fact that they made it open ish. But I still don't understand what their AI strategy is.
- [130:48]
A
Yeah, so revenue is 51 billion. So about a third of that of Google. Google's doing okay. Google is doing okay, by the way.
- [130:59]
C
I should apologize, Leo. Why I put lines 96 and 97 in your space. I meant to put it below. How dare you violated the line.
- [131:09]
A
I thought I was smart enough to put those in now from their space. You have no idea how screwed open AI actually is. See this? So there is a whole Cottage industry of AI deniers, Gary Marcus, Ed Zittran, and now let's add Will Lockett to the bunch. This is all AI bubble stuff. I have to tell you, I'm not convinced. I think the difference between this bubble and other bubbles is there's real value being created.
- [131:42]
C
I just don't think the business models.
- [131:43]
B
Said about web3 crypto and NFTs every single day. And I'm saying it right now, I.
- [131:50]
A
Believe in my life there's no value being, creating value.
- [131:52]
B
You being created here and everyone who's criticizing those technologies, they're just haters.
- [131:57]
A
Keep buying the stock. Look, the fact that, the fact that the stock still goes up, that people are still buying this stock in the face of nearly universal doom saying, tells me there's something going on here. I don't know.
- [132:14]
B
Yeah, that, you know, the market and, and retail shareholders have always been rational.
- [132:23]
A
All right, I want to take a break, then we're going to come back and talk about Elon Musk's Wikipedia competitor, which mostly in it is mostly made of Wikipedia, ironically. We'll talk about that in just a bit. Paris Martineau. Jeff Jarvis, you're watching intelligent machines. That's because you're an intelligent podcast consumer. We thank you so much. Our show today, brought to you by Vention, talked to these guys a couple of weeks ago and I was so impressed. AI is supposed to make things easier, right? But for many teams, it's just made the job harder. We're seeing the evidence of that, but that's because they haven't called on Vention. This is where Vention's 20 plus years of global expertise in engineering comes in. Vention built AI enabled engineering teams and they continue to do so. They make software development faster, cleaner and yes, calmer clients. Vention clients typically see at least a 15% boost in efficiency. And that's not hype, that's through engine real engineering discipline. The other thing I want you to know about, Vention has a fun and very informative AI workshop. It's a workshop that you work at. It'll help you find practical, safe ways to use AI across delivery and qa. It's a great way to, to start with Vention to test their expertise and for your company and your team to learn whether you're a cto, a tech lead or a product owner. You don't have to spend weeks figuring out tools, architectures, models. Vention will help you assess your AI readiness. We'll help you clarify your goals, outline the steps to get you there without the Headaches, that's all in the workshop. If you, if you happen to need help on the engineering front front, they're also ready to jump in. Whether it's a development or a consulting partner. This is literally the most reliable, the best step you could take after your poc. You know many companies are in this position, right? They got a great proof of concept, but they don't know where to go next. Let's say you've built a promising prototype. You did it in lovable. It's running, the tests are passing, but what's next? Do you open a dozen AI specific roles just to move to the next step? Or do you bring in a partner who's done this across industries? Oh, maybe that's an idea. Someone who can expand your idea into a full scale product without disrupting your systems or slowing down your team. That's Vention. Vention is real people with real expertise and real results. Learn more@ventionteams.com see how your team can build smarter, faster, and with a lot more peace of mind. Or get started with the AI workshop. Do it today. Ventionteams.com TWiT V E N-T-I-O-N teams.com TWiT it's like invention, right? Ventionteams.com twit Smart people. So I have a Wikipedia entry. This is not it. This is actually kind of different.
- [135:32]
B
Do you have a Wikipedia?
- [135:33]
C
Yes, I do. Yes, I do.
- [135:34]
A
Do you? Paris?
- [135:35]
C
Did I tell you my story?
- [135:37]
B
The only non famous person on the podcast.
- [135:40]
A
You won't be for long. Somebody create a Wikipedia entry for Paris. It won't.
- [135:43]
C
Did I tell you my wicked, my amusing Wikipedia story?
- [135:46]
A
No.
- [135:47]
C
So one day I'm walking down the hall at, at the Craig, Craig, Craig, Craig Newmark Graduate School of Journalism and a student comes excitedly out of a classroom and she says, jeff, Jeff, your Wikipedia entry says you're polyamorous and it says your wife is okay with it.
- [136:10]
B
That's what a string of.
- [136:15]
A
Who put that in there? Jeff?
- [136:16]
C
I don't know.
- [136:17]
A
You're not supposed to edit your own entry.
- [136:19]
C
I, I, this was a student who was.
- [136:26]
B
Known to be inappropriately affectionate student to bring up because like, what if it was true? Was she expecting to get out of that?
- [136:35]
C
I said no. People use Wikipedia to troll people and I'm not allowed to change it. Could you do me a favor and go in and delete that?
- [136:45]
A
We know Molly White, so if you ever have trouble again, I think you could get Molly White to fix it. My, my Wikipedia entry is mostly accurate. Occasionally some wag will put in some silly fact which pretty quickly gets it.
- [136:58]
C
Leo hates AI or something.
- [137:00]
A
No. At one point somebody said now looking.
- [137:03]
B
At your Wikipedia pages but you're both born in the 50s.
- [137:08]
A
Yeah.
- [137:10]
B
So yeah, it's just a crazy decade to be born in.
- [137:14]
C
It's jeopardy.
- [137:14]
A
If for her it would be for like us, you were born before the turn of the century. 1892. It be like that.
- [137:24]
B
That's incredible. I'm so sorry. The dead. The dead silence after I said that told me everything I need to know of my comments. I'm so sorry.
- [137:36]
C
Rub it in Paris. Rub it in.
- [137:39]
A
It's interesting because I have to say the Grokopedia, which is so, let's face it, Elon's rationale is that, well, Wikipedia is left leaning so we're going to have Gropedia be accurate which means right leaning. There's no clear political leaning in Wikipedia but if you think it's woke, well anything you do to counter that's going to be very right leaning. However, I have to say, you know it's accurate. It's got a lot more detail than Wikipedia had. So they didn't completely rip this from Wikipedia.
- [138:15]
B
Dev null has a Wikipedia page. Does Dev null have a Grokopedia page?
- [138:21]
A
Dev Null, my alter ego. Let's see, that would be a sign that they just ripped it off.
- [138:28]
C
No, I'm glad to say I do not have a Grokipedia page.
- [138:33]
A
Oh, so they didn't just steal it from Wikipedia.
- [138:37]
C
Well, I'm sure they did. They just limited how many there are. And by the way, you can take Wikipedia, it is open, it is under common. The Germans took it some years ago and did a printed encyclopedia out of it.
- [138:50]
A
It's not a good idea because it's a constantly changing document. So you're getting a, you know, a snapshot in time, I guess. Any, any Wikipedia is a snapshot in time. Any can sometimes encyclopedia is a snapshot in time.
- [139:02]
B
Photos of you. I've seen the devnall videos. Do you have any photos of you in the VR motion capture suit?
- [139:09]
A
Oh, somewhere. There must be. I don't know.
- [139:13]
B
I'm learning a lot on the Wikipedia page.
- [139:15]
A
Yeah, so far there's been nothing inaccurate. It's a lot more detailed than you would find in.
- [139:20]
B
I like that Wikipedia page. It says he was funny and his jokes were not gags. Wait a minute, that's a quote from Soledad.
- [139:33]
A
Okay.
- [139:36]
B
Because she famously hated you.
- [139:38]
A
I thought she hated me, but she didn't hate me. She hated Dev Null. Yeah, that's pretty. This is pretty good.
- [139:46]
C
So you're. But yours says you are semi retired.
- [139:49]
A
Oh, wait a minute. I found a mistake. I am semi retired. That. Actually, my bio says it took that okay because I only work three days a week. Isn't that semi retired?
- [140:01]
C
No, I consider.
- [140:03]
A
I mean, I only do five shows a week. That's hardly any shows a week.
- [140:06]
C
What did you do at your peak?
- [140:09]
D
Oh, God, I don't know.
- [140:12]
A
Probably ten shows a week. Yeah, probably twice that. Yeah, I think it was two shows a day plus. Plus two radio shows.
- [140:22]
C
Right.
- [140:23]
A
So, yeah, it was 10 or 11. Anyway, there is an inaccuracy. It says. Oh, no, no, no, no, this is right. No, I take it wrong. I take it. This is probably the most complete bio of me I've seen. Yeah.
- [140:40]
B
Interesting. I like that. One of the books you published is called Poor Leo's Computer Almanac.
- [140:48]
A
Yeah, you want to see the COVID.
- [140:50]
B
I do, because I've seen the normal almanacs. Wow. You did an ama. I also did an ama.
- [141:00]
A
You did get a kick out of this. I did Last week.
- [141:03]
C
Oh, last week about. Oh, how was this?
- [141:04]
B
This was the first time people weren't mean to me.
- [141:07]
A
This is before you were born, Paris. No, but you were a toddler. You might have been still in diapers. This is 2002.
- [141:15]
B
Oh, I was. I was alive then. I'd been born. Oh, my God, that's great. That's. That's great. Especially because it was 2002. So you actually were wearing that and doing that. This isn't a photo.
- [141:26]
A
I. This is not a Photoshop. This is. No, I posed. We had photographer come out. I wore overalls, a flannel shirt, a straw hat. I have a pitchfork.
- [141:36]
B
Put your mind back 23 years. Were there other poses you did?
- [141:40]
A
Oh, many. Oh, many. Wait a minute. I should show you my calendar.
- [141:45]
C
Is it a fold out?
- [141:47]
B
Can we do. Can we do an intelligent machines calendar?
- [141:51]
C
Oh.
- [141:57]
B
I think it could be fun. That could be one of. A photo shoot for. That could be one of the things we do in the 24 hours.
- [142:02]
A
Paris. Paris told me after the DND thing she said when I first started on Twit, I thought you were serious. That. What.
- [142:15]
C
What.
- [142:15]
A
What would you.
- [142:16]
D
Oh.
- [142:16]
B
The turning point was when you brought up the Lee cat is when I realized that you were a bit of a goofball.
- [142:23]
A
So this was such a success in 2002 that in 2003 we decided to ship a calendar with the.
- [142:35]
B
Please give me some more of those.
- [142:36]
C
For Those of you listening, the COVID is Leo as mad scientist.
- [142:41]
A
Yeah. Plugging two things together. And there was.
- [142:45]
B
And he looks mad.
- [142:46]
C
Yeah, he does.
- [142:47]
A
And then. And then every month there was a new costume. And we shot this all in one day. One day with props. So there's January.
- [142:56]
C
That's what you're going to be doing on our show together. Right, Paris.
- [142:58]
B
That's going to be great.
- [143:00]
A
This is February.
- [143:01]
B
I like that. February school.
- [143:04]
A
I'm Cupid in the long johns, apparently. But yeah. Yeah, it's. These are all seasonal.
- [143:09]
B
Okay. The March one. Classic.
- [143:11]
C
That's a classic.
- [143:12]
A
I look like a leprechaun. I'm an ice Irish for St. Patrick's Day. Oh. April showers bring May flowers. Right.
- [143:20]
B
The crumpled nature of the umbrella on that one.
- [143:24]
A
So the stylist. Who. The photographer and stylist. The stylist brought all these costumes. She said we're going to shoot a 12. We're gonna shoot 12.
- [143:33]
B
Were there any that didn't work out, did you?
- [143:36]
A
I still have the UPS costume. That was.
- [143:38]
B
Wait, I'm sorry. I do need you to go through the whole calendar. This is gonna be bad audio for me.
- [143:43]
A
April, May, that's Indy. You know, the Indy 500. That's actually my YouTube profile avatar today.
- [143:52]
C
Wow.
- [143:53]
A
It's me. Apparently, the steering wheel has been disconnected from the car.
- [143:57]
B
That's not great.
- [143:58]
A
This is summertime with the nose zinc and a Pepe frog. Before Pepe was Pepe inflatable.
- [144:07]
C
Before frog protesters.
- [144:08]
B
Oh, that's good.
- [144:10]
A
Yeah. Old Uncle Sam for the 4th of July.
- [144:13]
B
See, that's the sort of beard that you should grow for the show.
- [144:16]
A
I think for November I should grow a beard. Could I grow that in a month? I don't think so.
- [144:20]
B
I think so. Yeah.
- [144:20]
A
No. Here I am in August, barbecue season.
- [144:24]
C
You look so proud, Hank.
- [144:26]
A
I'm a schoolboy for September. See, I got a propeller hat. Yeah, yeah, yeah, yeah, yeah. October is Halloween.
- [144:35]
C
Paris, do it. Do the voice. Thank you.
- [144:42]
A
November is a. I kind of. I'll be honest with you. I'm kind of a gay pilgrim. I got. My collar is.
- [144:49]
B
It's the silky tie that makes it a little.
- [144:52]
A
I have a lavender silk tie and my colander.
- [144:56]
B
My collar going to say, well, hi there.
- [145:00]
A
Exactly. I'm the gay pilgrim. And the collar, I think, came from Laverde and Shirley because it's got a monogram. Dell on it. I can't believe the stylist brought that. It's hysterical. You see, I didn't have any choice. I just had to put these on right. Oh, there we go. Yeah.
- [145:19]
D
Who owns these pictures?
- [145:20]
C
I should reissue this calendar.
- [145:23]
A
I've been trying to for years. Yeah, this was fun. They only did it one year. It wasn't the hugest success. I think they were stuck with a lot of calendars. They realized calendars are only good for a couple of months a year, and then that's it. Mark Compton, thank you for doing those fabulous pictures. And I don't remember the name of the stylist who brought. But she also brought a UPS outfit, brown shorts, like. And that did. As, you know, notice, did not make the cut. But I still have it, which is weird. I don't have any of those other costumes for some reason. I have the one that didn't make the cut.
- [146:02]
C
I think you've got to wear that next week.
- [146:04]
B
You do? I mean, you should have worn it tonight in honor of Halloween.
- [146:07]
A
Oh, yeah. Because Halloween. This is my last chance. Halloween will be over by the next show. Sign up for your 2026 calendar. Coming soon. No, I had fun when I was young and famous. And now that I'm old and in the way, I just have to. I have my memories.
- [146:28]
C
We're getting ready for the next calendar already. There's a new picture.
- [146:31]
A
Memories on the chat. This was a good idea for a book. It was an almanac. So there. Every day of the year had another entry. It had a.
- [146:39]
B
All right, give me today's.
- [146:41]
A
All right, let's see. October. What is this? October 29th. So this was a pain in the ass.
- [146:48]
C
Go to your club, you serve.
- [146:49]
A
And Mac Tip PCs won't read Mac disks without special software. But Macs are perfectly content to read PC disks. That's still true today. To read a Mac Zip disk on the PC. Well, I don't know how true that is. Geek speak. A processor, AKA microprocessor, cpu. The brains of the entire computer. What I didn't have. I guess I maybe ran out of. I thought we. Oh, no. Maybe that was the next part. Okay. The next book had this date in computer history as well. Geekspeak. A docking station is a fixed piece of hardware to which a portable computer can connect. Laporte support a printer. Question. When I print anything from the Net, information on the right and the left of the page is cut off.
- [147:39]
B
I could buy this on ebay right now for $7.68. Sense.
- [147:45]
A
Well, that's a considerable discount from the 24.99 cover price. What about the calendar?
- [147:53]
B
Your face for the 2004 Tech Almanac is so funny.
- [147:57]
A
I believe I Have.
- [148:00]
B
It's so coy. Ah, I just put it in the chat.
- [148:10]
A
I used to be somebody. Now I'm just Hank.
- [148:13]
C
A podcast.
- [148:14]
A
Hank's father. This is the 20. The 2005 Mac gadget guide. That is a terrible picture. This is the Happy Leo 2006 Technology Almanac.
- [148:28]
C
The Thoughtful Leo.
- [148:29]
A
Thoughtful Leo. This is the 2005 Technology Almanac.
- [148:35]
C
The same Leo.
- [148:36]
A
It's the same waiver. Okay, I just want to point out something. They didn't bother getting another picture. But. But what are these pictures?
- [148:47]
B
Okay, wait, there's this one that I'm going to put in the chat that I think is just the same photo, but change the. The color of your shirt is artificially changed, probably.
- [148:57]
A
Well, I don't know. Did we have that technology back in 2006? I guess there were a lot of other. Oh, every. Every day had a picture in the 2005.
- [149:08]
C
God, how many did you shoot? Oh, my Lord Jesus.
- [149:13]
A
It's kind of an endless. Oh, I get it. So here. I guess we did use the ups. Here I am delivering packages. Hey, your new computer's here. Oh, my God.
- [149:28]
C
So those pictures on the front cover, are those from inside?
- [149:31]
A
Yeah, I guess so. Oh, wow.
- [149:34]
C
This is.
- [149:35]
A
Wow.
- [149:36]
C
Down memory lane with Leo.
- [149:38]
A
I'm so sorry to do this to you. Here's my. Here's what my blog looked like in 2005. There's a screenshot of Leoville.
- [149:47]
B
Leoville.
- [149:48]
A
Leoville. Yeah. Anyway, I apologize to everybody who is watching. Wasting your time.
- [149:56]
B
Wait for. You can get the 2013 bobblehead. Leo Laporte Limited.
- [150:04]
A
I have one of those, too.
- [150:05]
B
$150.
- [150:07]
A
Those are a rarity. Those are a rarity. Here's Cowboy Leo.
- [150:13]
B
You're blowing off a gun.
- [150:17]
A
Oh, my God.
- [150:18]
C
Your picky is extended. That doesn't make for a tough, tough look there.
- [150:21]
A
I was a very genteel cowboy.
- [150:24]
B
Multitudes.
- [150:26]
A
Very genteel.
- [150:35]
B
What a world. Photo guides of me.
- [150:40]
C
No.
- [150:41]
A
Yeah, we've seen Jeff, though, on things like Moonlighting on tv. Yeah. I am semi retired anyway. That's. I mean, compared to then I'm definitely semi retired. Yeah.
- [150:55]
C
Well, since we're the memory lane part here. Yes, AOL has been sold. Did you even know this still existed?
- [151:03]
A
And who. Who am I, pray tell, bought AOL. And why?
- [151:07]
C
You want to guess the price?
- [151:09]
A
$142 million.
- [151:12]
C
$1.5 billion.
- [151:15]
B
What?
- [151:16]
C
To bending spoons.
- [151:18]
A
Bending spoons? Yeah, this was rumored that this was going to happen, so it happened. Bending spoons.
- [151:23]
C
Italian tech holding company.
- [151:25]
A
Yeah, they've Been buying up a lot of stuff. I'm surprised. Bending Spoons says they secured $2.8 billion in a debt financing package to support the purchase. So it's kind of like a private equity thing. They're going to sell off the parts, I would guess. We reported this last, earlier this month that they were in the market for.
- [151:46]
C
It happened.
- [151:49]
A
Wow. AOL was owned by Time Warner. Remember the big, famous merger?
- [151:54]
C
Oh, do I. My. My FU money said fu. Jeff.
- [151:58]
A
Yeah, Was. Your FU money was in Time. And then they spent it on aol because, you know, the Internet's going to be big someday. Then Verizon bought them. So they own Evernote, they own Meetup Up. Bending Spoons. Does they own Vimeo? There's some strategy behind all that.
- [152:21]
C
Wow.
- [152:24]
A
All right. I feel like we've. This is the time in the show where I've run out of things to talk about, and I'm going to invite Paris and Jeff. Oh.
- [152:36]
D
Oh.
- [152:36]
A
It's a Gizmo.
- [152:37]
C
This is a Gizmo moment. It's a Gizmo moment.
- [152:40]
B
It's Gizmo moment. Gizmo. What do you have to say to the people? You want to immediately leave because you're being perceived? Gizmo.
- [152:49]
C
Meow.
- [152:50]
B
Nope. She just wants to show hold.
- [152:58]
A
Well, right. I'm gonna get Casio's fluffy AI robot because you've got gizmo. I'm gonna get mufflin 400.
- [153:07]
B
Why does that look like it's been squished?
- [153:15]
A
I don't know. Boone Ashworth, writing for Wired, says that his friend's dog, Wiley, sits and watches, suspicious of Mufflin's every move. Well, it's made by Casio, okay? It's been huge, apparently, in Japan since it was launched a year ago. There a soft, furry robot that uses AI to react to sounds and touch and develops its own unique personality as a result. It has, according to Casio, 4 million personality traits. I'm kind of tempted. It's a little expensive.
- [153:56]
C
All it does is chirp.
- [154:00]
B
Yeah.
- [154:05]
A
It emits a series of frankly adorable little noises, says Wired Mag.
- [154:11]
C
I'll do that for you for free.
- [154:12]
A
With Sing Sony.
- [154:14]
B
I'll do that for you for 200 bucks.
- [154:16]
A
I'll, you know, give you half the price. Sing songy tonal changes that aim to indicate whether the way you interact with is good or bad. So it's like what we were talking about earlier with Alan Cowan. It's, it's, it's reacting. It can react to sounds. Equipped with a microphone, it can react to Sounds around it. Oh, Mofflin can react to sounds around it like a little whisper or the clicky, clacky keyboard sounds of me typing this exact sentence. It's sitting right next to me and it just made a little sound and reaction.
- [154:50]
C
430 bucks.
- [154:52]
A
I think I should have one for the studio. I'll put it right there on the Nixie clock and we can just. It'll. You'll see it in the background reacting.
- [154:59]
C
You would have to mic it.
- [155:01]
B
Yeah. Mic it up. All right.
- [155:08]
A
Well, it doesn't seem like it goes anywhere.
- [155:12]
B
It just seems to be a poof.
- [155:14]
A
Here's a video of it.
- [155:15]
C
It. There's sound very.
- [155:17]
B
Oh, it's very tiny, actually. That does do something.
- [155:20]
A
It's kind of cute. What? Did you say something?
- [155:24]
C
You missed it.
- [155:25]
A
That was it.
- [155:26]
C
Yeah, that was it.
- [155:28]
B
Wow. Does it not have facial?
- [155:32]
A
It sounds like a dog wine.
- [155:33]
B
This looks upsetting.
- [155:35]
C
Yeah.
- [155:36]
A
It's going to look at you at some point.
- [155:37]
B
Featureless ball of water.
- [155:40]
A
It doesn't have a face.
- [155:42]
C
It does if the other picture has a face. Why they put it in the bowl, I don't know. So the dog doesn't bite it.
- [155:47]
B
I don't like it. It's just a. Yeah. How strange.
- [155:55]
A
Can I buy it in the US or do I have to? I think it'd be fun to take on. Sure.
- [156:03]
C
The thing.
- [156:03]
B
I don't know. I'm hearing it now too, but I don't. Oh, where is it coming from? Is it delayed? I don't like this. It's.
- [156:18]
C
It's all strangled.
- [156:20]
B
It does kind of sound like I've stepped on something. I feel like I'm going insane.
- [156:29]
A
Wait a minute. I don't know. Can I get this?
- [156:31]
B
People in the chat are noting that it is just a trip tribble.
- [156:34]
A
It's a tribble.
- [156:35]
C
Yeah.
- [156:35]
A
And there's only one. Oh, yeah. You could buy it right here. Buy Mofflin. Should I get silver or gold? I think gold.
- [156:43]
C
Nope. Leo, stop yourself. Stop.
- [156:46]
B
A ticket to New York to come eat yourself.
- [156:48]
A
I can have a sandwich. I could have my son's sandwich. Oh, see? Now you see the face a little bit.
- [156:55]
B
Dewey.
- [156:55]
A
Oh, she's feeding it. Isn't that sweet?
- [157:00]
B
Is that a face or is that just a button? Where are the noises coming from?
- [157:07]
C
It looks like an owl coming from inside the house.
- [157:15]
A
I'm just gonna let it play for the rest of the show. I really want this. Do you think it'll get. It's got.
- [157:25]
C
Somebody call Lisa. Somebody call Lisa.
- [157:27]
A
Four million. It's got What? What do they say? Four million different things it can do. I don't. Can't do much.
- [157:39]
B
That little puffball can't do five things, let alone 4 million. Jeff, you've got to do one now. I think we're devolving. All right, you picked a story now, intelligent machines.
- [157:55]
A
You started with stories.
- [157:56]
B
What if I pick a pick of the week?
- [158:00]
A
Oh, let's do the picks of the week. We could do that. Because you know what comes after the picks of the week? Cacio e Pepe. Yeah.
- [158:08]
C
Well, first you want to see how I'm going to sacrifice for the show. Leo.
- [158:11]
A
Yes.
- [158:11]
C
Line 140.
- [158:12]
A
Okay.
- [158:13]
C
I'm going to. I've signed up to attend this for.
- [158:16]
A
The sake of the show. You will set up. Set. Oh, God. No, don't. Tony Robbins right there.
- [158:26]
C
Disqualified. Right there. You stop right there.
- [158:28]
A
Dean Graziosi and world class AI experts live discover how to turn your AI into the ultimate advantage without confusion or overwhelm. Reserve my free seat now. Is it really free?
- [158:44]
C
It's free, but then they try to upsell you for $1 and then they get your credit card.
- [158:47]
A
How could it.
- [158:48]
C
And you can get this look awful.
- [158:50]
A
How could it be?
- [158:50]
C
Oh, you could join it virtually three hours per day.
- [158:54]
A
We did the work three days. You get the shortcut.
- [158:58]
C
Watch the video. The video is gonna.
- [158:59]
A
Have you any experience with Tony Robbins at all? No, I don't. But my former co host Amber MacArthur apparently became friends with him and ride in his plane.
- [159:12]
C
Geez.
- [159:14]
B
Oh, boy.
- [159:15]
C
If you play the video, it will tell you how. How inspiring it's going to be.
- [159:18]
B
Oh, well, this free event is built for you if you're ready to finally feel comfortable.
- [159:25]
A
Mark Benioff's gonna be there.
- [159:27]
C
Well, he needs to make friends now after calling troops into San Francisco.
- [159:31]
A
Yeah. Thank you. Video.
- [159:34]
C
I don't see it.
- [159:36]
B
Oh, at the beginning of the website it says earnings disclaimer. We don't believe in get rich programs, only in hard work, adding value, building a real and professional career and serving others with excellence and consistency.
- [159:50]
C
What the hell?
- [159:51]
A
You got to do this, Jeff, because.
- [159:54]
C
We need a good show for the.
- [159:56]
A
Good of the show. 11:00am Pacific, November 6th, 7th and 8th for three hours. So you're going to get nine hours absolutely free.
- [160:08]
C
Where's the video, damn it?
- [160:11]
A
Meet your speakers. Who is Dean Graziosi?
- [160:15]
C
I don't know, but he's the one who does the video.
- [160:16]
A
He's the co founder of Mastermind, entrepreneur.
- [160:22]
B
Cooking platform that this is on. Mastermind.com is the platform that it's on.
- [160:27]
A
Oh, okay. So he's. It's his thing.
- [160:32]
B
Yeah.
- [160:36]
A
Well.
- [160:37]
B
Well, okay.
- [160:38]
C
For you. For the good of the show.
- [160:40]
A
So you. You did this and they asked for money later?
- [160:42]
C
No, they asked for $1. And then you get more benefits.
- [160:45]
A
Oh, but then there's.
- [160:46]
C
Now I'm getting reminders. It's only seven days away. Former guest of the show, Zach Cass is going to be there.
- [160:54]
A
Oh, okay.
- [160:57]
B
All right.
- [160:58]
C
Which one was he? Was he the one?
- [161:01]
A
Which one was he? We have had quite a few.
- [161:03]
C
Second or third. Third day, I think. Oh, yeah. He was the one who was all Mr. Positive. Yeah, he was. Yeah. I didn't. We. We didn't find Kismet. Sacking me.
- [161:13]
A
Was he the go to market guy for OpenAI? Yeah. Okay. Yeah, he was our first guest.
- [161:22]
C
Yeah. And I thought, this is the way the vests are going to go.
- [161:26]
A
Hey, we were just getting started.
- [161:28]
C
We were just getting started.
- [161:29]
A
Now we're just getting. Getting done. Any other stories you want to. There's so much stuff. How about the archive.org story about machine olfaction?
- [161:42]
C
This is fun.
- [161:44]
A
Sniffing AI smells. It's realnose AI nose knows. The nose knows. Wait a minute. Now I have to go to RealNose AI so, I mean, you know, it'd be good if. Biomachine olfaction, electronic nose that detects cancer through urine. Scent detects chemical threats. Dogs can do these things and drugs.
- [162:09]
C
So why couldn't it?
- [162:10]
A
Right, Right.
- [162:11]
C
It's just machine learning.
- [162:12]
B
The AI is going to sniff your pee now.
- [162:15]
A
Well, you saw that Kohler's making a toilet with a camera in it.
- [162:20]
B
Oh, did I?
- [162:22]
C
She prayed it wouldn't be on the show. Prayed you wouldn't bring it up.
- [162:25]
A
Not gonna bring it up. That. Not gonna be.
- [162:28]
C
Well, meanwhile, there's also the next one.
- [162:29]
B
Leave Our Bottoms Alone. Why do we need to be recording any?
- [162:34]
C
It's a show title.
- [162:36]
A
Leave My Bottom alone.
- [162:39]
C
There's a smell net, is a large scale data set for real world smell recognition.
- [162:49]
A
I think it's time for us on that note to do our picks of the week.
- [162:56]
B
That's not related to nose.
- [162:57]
A
Poor Paris. He has to get to work on her costume. It's not easy finding logs in New York City.
- [163:03]
B
I really do. I've got to glue together. I've got to glue together a lot of Styrofoam and then.
- [163:10]
A
Are you gonna make your own?
- [163:11]
B
Yeah, I'm making my own log. My pick of the week is.
- [163:16]
A
Can you use fondant?
- [163:19]
B
Ostensibly.
- [163:19]
A
I probably could text.
- [163:22]
B
So I'd Recommend this week do a little photo shoot with your friends. I. The other week, a photographer named Robbie something Osberger was in town and me and my friends did a. My friend Maddie and I did a photo shoot, the results of which are posted.
- [163:40]
A
I saw these pictures and I was puzzled. Can I tell you? I was. I was puzzled. I don't. I didn't. I didn't really know what to think of this.
- [163:49]
B
It's just sometimes you've got to wear a suit with your friends and take some silly photos. The one.
- [163:54]
A
So this guy bring the backdrop or brought the backdrop.
- [163:57]
B
He brought all the props. I just showed up in a suit and I had a fun little time in 30 minutes. And now I've got frankly these photos down below that I'm going to put on my website for sources to contact me, which is me holding an old phone.
- [164:11]
A
That's great. Great.
- [164:12]
C
It's the Paris 2026.
- [164:14]
A
Although you look a little. You look a little bit like crunk.
- [164:17]
C
That's true.
- [164:18]
B
I'm a little crunk.
- [164:18]
C
Like, I think we have a Paris calendar. But you know, you're right.
- [164:21]
A
There could be a pretty good.
- [164:23]
B
Right.
- [164:23]
A
There's a calendar. Did you thrift that suit?
- [164:26]
B
No, I've had that suit. I bought that suit new many years ago.
- [164:31]
C
Wow.
- [164:32]
B
I know.
- [164:33]
A
Everybody should have a plaid.
- [164:34]
C
She moved.
- [164:35]
B
He's coming to Portland. He's coming. Coming to la. You know, you can do his pet photography. Check them out.
- [164:41]
C
I like.
- [164:42]
A
Dude, I like the. The quote that goes with this. We saw you from across the business conference and really liked your vibe.
- [164:50]
B
You got it. My other actual pick this week is I saw Bonia and it. It's interesting. It's a Yorgos Lanthimos. I am probably.
- [165:02]
A
Oh, this is the one where Emma Stone is bald.
- [165:04]
C
Yes.
- [165:04]
B
Well, this is the one where two conspiracy nuts abduct the CEO Emma Stone of a biotech firm that's also kind of Amazon like and they forcibly shave her head because they believe she's an alien and her hair is used to communicate with aliens. And it gets crazier from that.
- [165:24]
A
Are they Qanon types?
- [165:27]
B
I believe at some point they might name check Qanon. He goes gone through all the different. Different conspiracy theories.
- [165:34]
A
See, I liked poor things. I. I like Yorgos Lanthimos's work. I. He's a little challenging.
- [165:41]
B
Yeah, no, it's. I would recommend it. It was a very. I was gonna say fun film, but also my. My take on fun films may not be everybody's take. Have you seen some people said it was filled with despair.
- [165:55]
C
Yeah.
- [165:56]
B
I said afterward was like I gotta watch more Yorgos. And I watched the fun favorite.
- [166:00]
A
Well, I was going to ask you have you seen the favorite? Because that is an awesome.
- [166:04]
B
I watched that also. I watched that. This. I watched that like two days ago.
- [166:07]
A
Isn't that. Oh man, I could watch that again and again. And the lobster was interesting.
- [166:15]
B
I need to watch that was.
- [166:16]
A
I think it's one of his first movies. It's a little weird.
- [166:20]
B
All of his films are a little weird. Yeah, he's a weird guy.
- [166:23]
A
Yeah. But the favorites, the least weird and I think quite good.
- [166:28]
C
She's worked with him how many times?
- [166:30]
A
I don't know. She did poor things with him.
- [166:32]
B
And yeah, also in the favorite also. And I believe one or two others. Yeah.
- [166:38]
C
Yeah.
- [166:40]
B
They're a great pair. And it's also just fun to see someone get their head shaved live on camera. Famously. This is Begonia is the movie that an early screening they did last week. They didn't Oops. All bald screening where you could only you could get a free ticket to the screening and only bald people were allow allowed to be.
- [166:58]
A
I'm not shaving my head again. Don't ask.
- [167:00]
B
Once they in for people to come in. They literally had a barber on deck for people to get their heads shaved. But too many people wanted to get their heads shaved to see Begonia.
- [167:11]
A
Just one point. They had movie tickets.
- [167:13]
B
They had to cut it off and start handing out bald caps. Imagine being the last person to shave your head before the bald cap.
- [167:20]
A
I could have worn a bald cap.
- [167:22]
C
Is there a picture picture of the audience? There should be.
- [167:24]
B
Yes there is. And apparently they all went wild during the head shaving scene.
- [167:32]
C
Brilliant.
- [167:32]
A
Okay, I'm looking for Begonia bald caps. Not everyone had to go bald.
- [167:37]
B
Yeah. Wait, no, just search Begonia all bald screening. I'm sure the photo will come up. Oh yeah, here it is.
- [167:47]
A
Yeah, some of them did not get the best shave jobs. I had that problem too because we had a, a an amateur shave my head at the event and I actually had to go to a bar. I did the same thing to you. I shaved your beard and it did a terrible job. Yeah, you did. And look at all these baldies. All right.
- [168:05]
B
Makes the hair stand out.
- [168:07]
A
Jeff Jarvis, what's your pick of the week, sir?
- [168:09]
C
Well, we could celebrate 25 years of Google advertising, but I somehow think that that's probably Woohoo.
- [168:15]
A
What a party. Why would they celebrate? Really? Is that the word you want to use, Google?
- [168:21]
C
Well, it's been Good for them.
- [168:23]
A
Actually, some of their ads have been pretty clever. Are they talking about their television ads? No, they're talking about Google advertising. Yeah, the thing they're getting sued for by everybody and their brother because they have a monopoly. Okay, fine.
- [168:38]
C
So then next, the. The Washington Post had a TikTok guy, Jorgensen, who left, and their TikTok traffic at the Post has plummeted.
- [168:48]
B
Yeah, because Dave was basically the face of their tick tock. He also ushered in a new era of like, brand tick tocks for like, news companies.
- [168:58]
A
So then that's really interesting. Boy, that's. That shows that these tick tock social media people are not fungible. No, there's something going on there. But he made a star of himself, right?
- [169:12]
C
Full head. So Paris, how high res was your TV that you spent a fortune buying on Leo's device?
- [169:20]
A
She's got a 4k 4k here.
- [169:23]
C
Well, turns out research shows that doesn't matter. We've reached peak screen long ago. That research shows that a 4K or 8K screen offers no distinguishable benefit over a 2K screen.
- [169:37]
A
So here's the weasel words in an average living room where you're sitting like 15ft away.
- [169:43]
B
My living. My living room. Not average. I also did Leo's trick and measured the distance between my teeth.
- [169:49]
A
Exactly.
- [169:50]
B
I'm gonna be watching all of my nickvember movies in beautiful form.
- [169:55]
A
You're so. You're so lucky. No, they're right. If you're not close enough, you won't be able to distinguish the difference or the size.
- [170:03]
C
But it kind of says we've reached peak screen panel.
- [170:06]
A
Well, in the size of the panel.
- [170:08]
C
But.
- [170:08]
A
And I agree, there's no point to go at 8k. 4k is like pretty darn good, right? Maybe I. Maybe not.
- [170:15]
C
Maybe there are there. Are there some. Trying to sell you 12k ones now. Well, there isn't even. There isn't even yet.
- [170:22]
A
Yeah, the human eye can resolve more detail than commonly thought. The average, 94 pixels per something for grayscale images, 89 for red, white and green images. Yellow and violet, lower 53 ppd pixels per. What is it? Inch. What does the D stand for? Pixels per degree. O degree. See, this is why the size of screening. The distance from the screen is relevant.
- [170:54]
C
Right, okay, Yep.
- [170:59]
A
Yeah, I mean, that makes sense. But. But if. But if. But how close are you to your beautiful 4K screen you measured?
- [171:05]
C
Paris, how close are you?
- [171:07]
B
I don't know. They're really saying whatever the right distance is for the size of my.
- [171:13]
A
Listen to this. If. If someone Already has a 4k 44 inch tv. Who has a 44 inch tv?
- [171:20]
B
I had a 44 inch tv until you convinced me to make poor choices.
- [171:24]
A
And watches it from about two and a half meters away. That's already more detail than the eye can see. Right. They don't need an up 8k. Well, thank you for. Thank you for that, bozo.
- [171:35]
C
There's a handy display resolution calculator.
- [171:38]
A
Well, that's handy. It's free.
- [171:40]
C
I think you earn more points in AGI.
- [171:45]
A
Actually, this is from the University of Cambridge. So you know they're the. There. It's got to be. This would be useful for you, Paris, just to see. But all this will tell you is how close you should sit. But you're what, how far? Five feet, six feet away. Probably not more than that, right?
- [172:00]
B
I couldn't tell you. I have no understanding of space.
- [172:03]
A
You don't. You don't sit like you're not watching the TV for back here, are you?
- [172:09]
B
No, I mean. Well, I will say I got really close to the TV whenever I was watching Twin Peaks. Yeah, Season three part.
- [172:18]
A
And aren't you glad you got excited? The pixels. You were looking at the log and saying.
- [172:23]
B
I was.
- [172:24]
A
How do I make that?
- [172:24]
B
I gotta make that.
- [172:26]
C
How do I make you?
- [172:27]
A
Now what else besides carrying a log around? I think it's a pretty simple costume.
- [172:31]
C
Yeah, what else?
- [172:32]
B
I mean, yeah, listen, I got. I got her glasses, I got a skirt, I got a button up shirt that's the same one as hers. A little sweater. I'm gonna make her little brooch as well tonight.
- [172:46]
C
Now, how many of your friends have watched Twin Peaks and will get it?
- [172:50]
B
Not enough for the amount of effort I'm putting it. But I've done a like cardboard.
- [172:57]
A
You already have the glasses.
- [172:59]
B
I mean, they're not. I. I bought some red ones online, just.
- [173:04]
A
Oh, you got serious.
- [173:05]
B
Oh, she does like five bucks.
- [173:08]
A
What is the. I know. She's a lady who carries around a log. Is there anything else to say about her?
- [173:14]
B
There's so much to say. There's so much to say.
- [173:18]
C
Actually.
- [173:21]
B
Spoilers for Twin Peaks, I guess. Skip forward.
- [173:24]
A
She's important. She's important.
- [173:26]
B
She receives like, kind of like divine prophecies from her log. Her log talks to her. And her log also contains the spirit of her husband, the fireman who died on her wedding. Their wedding night, because he went into the mysterious woods of Twin Peaks weeks. And then the next day, while searching for his remains, she found this log.
- [173:49]
A
Wow.
- [173:50]
B
The log has a message for you. As. As she famously says, you know what I think?
- [173:56]
A
I think you should put a little speaker in that log. That's maybe brilliant. Have a little, have some mofflin sounds.
- [174:04]
C
I think muffin sounds. Yes, have it. Have a chirp as you. As you pet it.
- [174:09]
B
She does so. One of my favorite little like vestigial parts of Twin Peaks is in. God noises are back. I'm so sorry to curse.
- [174:21]
A
But we're gonna have to bleep that.
- [174:23]
B
We will, we will. And I think that really just goes to show how taken aback I was by this.
- [174:32]
C
Chamber B says hey.
- [174:36]
B
One of my favorite vestigial parts of Twin Peaks is whenever it was re shown, I believe on like Showtime or something. Afterwards, Margaret Lanterman, the log lady, recorded these intros that played at the beginning of every episode where she delivers a straight to camera monologue, sometimes a message from the log, sometimes just musings on nothing at all that vaguely relate to the episode. It's phenomenal.
- [175:00]
A
Well, sounds like the perfect costume for a. A young person named Paris Martineau. You'll find, I think only people who.
- [175:06]
C
Were born in the 1950s will get.
- [175:08]
A
It, but other than that, you will find Paris's log lady pictures on her blue sky. Right, that's what you'll post.
- [175:16]
B
Paris nyc.
- [175:17]
A
Harris nyc. That's where you find all of the weird pictures. Paris post. She's also on Instagram. What's your Instagram handle?
- [175:27]
B
Paris Martineau. Don't be weird.
- [175:30]
A
Don't be weird. Yeah, I got rid of my Instagram. I deleted it from all possible. It was the only thing I was using it for is to keep up with Saul Hank. And I don't need to anymore. I just watch Seth Meyers. Jeff Jarvis is of course the author of fabulous books like the Gutenberg Parenthesis and magazine. His newest is coming soon. Hot type. I cannot recommend it.
- [175:54]
D
More high.
- [175:55]
A
It's fantastic. Another great book from the Jarvis printing machine. You know the one thing missing from that book? I think pictures of you in oddball costumes on the COVID That's the one thing I'm telling you. Sells books. Yeah.
- [176:14]
C
Sells books means the mad printer.
- [176:18]
A
Yep. Actually by at some point. Yeah. See I ended up having my own Leoville Press.
- [176:29]
C
That's cool.
- [176:29]
A
Yeah.
- [176:30]
C
Your own imprint.
- [176:31]
A
My own imprint didn't make me any more money.
- [176:36]
B
Let's all end the show with a little bit of spooky noises. You guys can contribute the way you see fish. Yes.
- [176:43]
A
Did I show you my costume? I gotta show you my costume.
- [176:47]
B
You guys gotta do some like Chain rattling noises. We can make a soundscape.
- [176:52]
A
I'll be wearing this. I don't know where I'm gonna wear this because it's a little bit inconvenient. You bump into people a lot. But I have the chicken.
- [177:01]
B
Oh, you're a chicken jockey. Do you even understand what that's a reference to?
- [177:09]
A
It's to a guy riding a chicken. What? Oh, no. I know this because I don't play.
- [177:16]
B
Minecraft, but it's a Minecraft reference.
- [177:19]
A
Is it really became so.
- [177:20]
B
Yes, it's like, I guess like a rare.
- [177:24]
A
They have chickens somehow.
- [177:26]
B
No, but specifically, it's Chicken Jockey. It's like a guy riding a chicken. And it became a big meme, and it caused panic in the streets because when children went to see the Minecraft movie and chicken jockey appeared, they like, like, threw a bunch of stuff at the screen. They had, like, shut down movie theaters.
- [177:47]
C
Crazy catchphrase. Took over multiplexes.
- [177:50]
B
Yes. No, it was. It was a problem.
- [177:52]
A
Well, I don't think that's who I am.
- [177:54]
C
They're gonna be throwing things at you.
- [177:56]
A
I don't think I'm chicken jockey. I think I'm just a guy.
- [178:01]
B
Oh, my God. The live action version of it is terrifying looking.
- [178:05]
A
Oh, well, that movie was fairly terrifying, to be honest.
- [178:11]
C
When Jack Black yells that in a Minecraft movie, young audiences respond raucously.
- [178:16]
B
That's the most New York Times way to describe 676-7677 as they're all saying.
- [178:23]
A
I missed a bed. I was 67 last year, and I could have really had something, something going. There it is. There's the chicken, There's Jack Black. There's somebody else.
- [178:36]
C
Okay, this definitely gets us taken down.
- [178:39]
A
Okay, so we're not going to show it. We'll just leave it as an exercise.
- [178:44]
C
Bonito protects us from ourselves.
- [178:45]
A
I would really like to see the Minecraft movie with a bunch of young people. Is that Jason Momoa? All right, thank you, everybody. Thank you, everybody. We do this show every Wednesday about right after Windows Weekly, about 2pm Pacific, 5pm Eastern. That's now going to be 2200 UTC because we go. Yay. Finally, we go to daylight, to standard time.
- [179:12]
C
Yay for me.
- [179:14]
A
Yeah, it's good for Bonito. Because your clock does not change, right, Benito?
- [179:19]
C
I get an hour.
- [179:20]
A
You get an extra hour of sleep. Next week, we're going to talk about Reflection AI. We're going to talk about post training with Jeremy Berman. In fact, Reflection AI has raised $2 billion. $2 billion to be America's open frontier. AI lab. Kind of a la deep seq. So we'll talk to the man behind reflection AI next week on OpenAI. I mean, intelligent machines. We're not called OpenAI.
- [179:52]
B
That's somebody else on OpenAI right here.
- [179:55]
A
If only I could get that money. Thank you for joining us. We are. I mentioned the time we're on live because you can watch us live. We're on YouTube, Twitch, Facebook, LinkedIn, X.com and Kick. You can also watch if you're in the club. In the club. Twit, Discord.
- [180:10]
C
And if you follow me on Twitter or Facebook or LinkedIn, you get it there too.
- [180:16]
A
He does a Simon, you follow me.
- [180:18]
B
You won't get it there.
- [180:20]
A
Yeah, there's blissful silence, except for the occasional muffling sounds.
- [180:24]
C
She doesn't want to tell her friends what she actually does.
- [180:28]
B
They don't need to know this because then sometimes I'll get texting. Like, I watched the podcast.
- [180:32]
D
Oh.
- [180:32]
B
I'm like, oh, boy.
- [180:33]
A
I always. When somebody says, should I watch your shows? I always say, no.
- [180:37]
B
Same. I'm like, it's for the people who are watching it right now.
- [180:41]
A
It's not for sure. If you know, you know, that's it. On demand versions of the show available audio or video or both at Twitter TV. If you. There's a YouTube channel dedicated to it. Of course you can subscribe in your favorite podcast client and get it automatically. That's the easiest way to do it. Leave us a 5 star review. Have we not had any good 5 stars lately, Paris, or just you haven't checked?
- [181:05]
B
We have. Yeah. Great one from victorwinn.com.
- [181:09]
A
Thank you.
- [181:10]
B
Victor describes himself as a weekly listener. They say this is the one podcast I always look forward to hearing on release day. I'm glad that they shifted their focus to AI when it was Twig. The show was mostly a second twit. I'm curious to know whether Paris receives hazard pay for putting up with Leo and Jesse. Hey, me too.
- [181:31]
A
And if you watch live, you can hear Paris swear like a sailor. That's another reason I'm keeping it really under wraps. He's keeping it real.
- [181:40]
C
There was only one the real Paris stray.
- [181:44]
A
Oh, it's perfect. Thank you, everybody. Have a wonderful week. Have a great Halloween. A safe Halloween. We will see you back here next week on Intelligent Machines. Bye. Bye. I'm not a human being.
- [181:57]
B
Not into this animal scene. I'm an intelligent machine.
- [182:04]
A
Morning, Zoe. Got donuts.
- [182:06]
B
Jeff Bridges, why are you still living above our garage?
- [182:10]
C
Well, I dig the mattress and I.
- [182:12]
A
Want to be in a T mobile.
- [182:13]
C
Commercial like you teach me.
- [182:15]
B
So Dana oh no, I'm not really prepared.
- [182:19]
A
I couldn't possibly at t mobile get.
- [182:21]
B
The new iPhone 17 Pro on them. It's designed to be the most powerful iPhone yet and has the ultimate pro camera system.
- [182:28]
C
Wow, impressive. Let me try.
- [182:30]
A
T Mobile is the best place to get iPhone 17 Pro because they've got the best network.
- [182:35]
C
Nice. Jeffrey, you heard them.
- [182:37]
A
T Mobile is the best place to.
- [182:39]
C
Get the new iPhone 17 Pro on.
- [182:41]
A
Us with eligible traded in any condition. So what are we having for lunch?
- [182:46]
B
Dude, my work here is done.
- [182:48]
A
The 24 month bill credits on experience beyond for well qualified customers plus tax and 35 device connection charge credit send and balance due if you pay off earlier. Cancel Finance agreement. IPhone 17 Pro 256 gigs $1099.99 and.
- [182:57]
C
New line minimum 100 plus a month.
- [182:58]
A
Plan with auto PayPal taxes and fees required. Best mobile network in the US based on analysis by Ooklove Speed Test Intelligence data 1H 2025 visit t mobile.com Marketing is hard, but I'll tell you a little secret. It doesn't have to be. Let me point something out. You're listening to a podcast right now and it's great. You love the host. You seek it out and download it. You listen to it while driving, working out, cooking, even going to the bathroom. Podcasts are a pretty close companion. And this is a podcast ad. Did I get your attention? You can reach great listeners like yourself with podcast advertising from Libsyn Ads. Choose from hundreds of top podcasts offering host endorsements or run a pre produced ad like this one across thousands of shows. To reach your target audience in their favorite podcasts with Libsyn Ads, go to libsynads. Com. That's L I B S Y N Ads. Com. Today.