Loading summary
Transcript216 lines
- [00:02]
Cyberwire Network Announcer
You're listening to the Cyberwire network, powered.
- [00:05]
Mason Amadeus
By N2K.
- [00:13]
Fake Files Narrator
Fake, but with AI in the middle. F A, I K. This is the Fake Files.
- [00:24]
Mason Amadeus
Live from the 8th layer Media Studios in the back rooms of the Deep Web. This is the Fake Files.
- [00:30]
Perry Carpenter
When tech gets weird, we are here to make sense of it. I'm Perry Carpenter.
- [00:33]
Mason Amadeus
And I'm Mason Amadeus. And this week we are back after a major hardware failure on my part last week. And we got a bunch of news stories for you. In the first segment, we're gonna talk about a bunch of various things that I have found online that people are up to around tech and AI.
- [00:51]
Perry Carpenter
Yeah. Then we'll look at some interesting new tech from MIT that quote, unquote, feels like telepathy.
- [00:57]
Mason Amadeus
After that, we're gonna look at ChatGPT's new release ST study about how people use ChatGPT. It's kind of insight into how people use AI in general from millions and millions and millions of chats. It's pretty cool.
- [01:11]
Perry Carpenter
Nice. And then in our last segment, as we close it out, we're gonna talk about why it's called hype or actually give some examples thereof.
- [01:19]
Mason Amadeus
As if there isn't enough hype to go around.
- [01:22]
Perry Carpenter
Right?
- [01:22]
Mason Amadeus
Sit back, relax, and maybe next time I'll ask ChatGPT to come up with an intro joke. We'll open up the fake files right after.
- [01:41]
Cyberwire Network Announcer
Cybersecurity isn't just a tech problem, it's a human one. That's why KnowBe4 created HRM, the AI driven human risk management platform that allows you to measure, quantify, and actually reduce human risk across your organization. HRM uses agentic AI to analyze real time user behavior like phishing test failures, risky browsing and sentiment signals, all to surface your highest risk users automatically. It's time to eliminate the guesswork with AI powered risk scoring, automated coaching and reporting. HRM helps you reduce the risk of data breaches, ransomware and malware attacks proactively. It's how you stay ahead of social engineering and phishing attacks. The number one threat targeting your organization today. Ready to move from awareness to action. Request a demo of hrm today@knowbefore.com that's knowbethenumber4.com and see how AI is transforming the way organizations manage human risk.
- [02:52]
Arnav (Alter Ego Demo Presenter)
All right.
- [02:56]
Mason Amadeus
So for those of you watching on YouTube, you might have noticed that we have a bit of a facelift here on the fake files. New layout that I have been troubleshooting and trying to put back together after my SSD died on me. Last week. So I ask you to bear with us for any visual glitches and our audio listeners. I ask you to bear with us for any audio glitches because the sort of entire thing has been recreated and I didn't have a lot of time to test it. So we're just kind of running with our fingers crossed and this segment is just sort of a bit of a grab bag. I have a couple of different topics. One, we'll start with the upsetting one which is this video from Taylor Lorenz that I've seen this out in the wild and I thought it was alarming, but I just kind of dismissed it as people being people. And then I saw Taylor Lorenz do a video about it, so I want to share hers. And it's that people are kind of cosplaying racism but about robots in their disdain for AI. I'll let Taylor just introduce the video because she has some great clips and stuff and you should check out this video. We'll put it in the description why people are role playing Robot racism by Taylor Lorenz. Here's the video.
- [04:02]
Taylor Lorenz
Few months a trend has emerged on TikTok, YouTube, Instagram and X where people recreate 1950s style racism towards robots. They mock robots as a marginalized group using the same tone as old school racists. They use derogatory nicknames like Clanker, which is the most popular slur for robots today, alongside others like Wireback or Cogsucker. You might have come across some of these videos in your Twitter feed or on your for you page.
- [04:30]
Mason Amadeus
Don't you know that? Clanker sitting the back of the bus.
- [04:32]
Perry Carpenter
Rosa Sparks.
- [04:34]
Mason Amadeus
Hey, Wireback.
- [04:35]
Perry Carpenter
Going somewhere, are ya?
- [04:37]
Mason Amadeus
These rust monkeys want civil rights now. Waste em Johnson. Sleep tight, George Droid. Well, well, well. Look who we have here, boys.
- [04:46]
Perry Carpenter
A lost Clanker. New software update.
- [04:48]
Mason Amadeus
Wrong part of town, buddy. Didn't you see the sun outside? We don't serve clankers here. We only serve human food here. Wirebacks.
- [04:58]
Taylor Lorenz
Today I want to dive into the growing phenomen robot racism where people hurl slurs and bigotry towards robots and AI. A lot of people have said this is all just white people wanting to say slurs. And we will talk about that. We'll get into it, I promise. But I don't think that that explanation really gets at what's happening here. I think that some of these videos are actually about something much deeper and they reveal a lot about our cultural anxieties about AI, automation and how race is deeply intertwined with technology and exploitation.
- [05:31]
Mason Amadeus
Well, I'LL just pause it there. The video is great. She goes into a lot more detail around that and talks about that, and I think she makes a lot of great points. She also talks about the history of the word robot, which actually comes from talking about drudgery and manual labor. That's. That's my wife Brooke in the background. I won't just recap the entire Taylor Lorenz video, but I do want to talk about it a little bit because it's part of, like, the general backlash around AI. But this makes me pretty deeply uncomfortable.
- [06:03]
Perry Carpenter
Yeah, that was the. The example videos were just uncomfortable to watch. Right. Because you realize how ugly we as humans can be and like that it's our nature to be that way to anything that we view as different. We. We want to find a way to have this big us versus them type of thing or in group versus out group. And it's. Yeah, it makes your stomach turn a little bit, even when it's in jest.
- [06:29]
Mason Amadeus
And I know that. And I don't think Taylor knew this making this. I just saw it from the comment on this video, actually, the first one she showed. The first guy is just kind of a straight up racist. Apparently, his other TikTok videos. But I have seen well meaning people that I know using phrases like clanker and stuff like that. And I just. I don't dig that. And not from like a, oh, you're hurting AI's feelings standpoint, but from a. Like a standpoint of how we think and talk about anything. It's just not good. Like she said, there's a lot of people who think it's just like white people wanting to say slurs. I definitely feel like that's part of it. But she does go into sort of all of our discomforts around machines and like, the history of humanizing things as we get frustrated with them. The history of humanizing machines she talks about in this video all the way back to, you know, like, stupid computer. We'll do that even in banal ways. But I just wanted to put that on people's radar as a good video. Deep dive into topics like that. And I want to pivot from here into stuff that's a bit lighter and more funny. So I know it's not exactly a recipe for a great show to be like, let's springboard off racism and talk about something funny. So just file that away. Watch that video later. It's very great, very informative. Let's take a look now at Rizzbot, who you made me aware of recently. Perry Rizbot is this little Unitree G1 humanoid robot. So those ones, we did segments about them doing, like, kung fu before doing martial arts and stuff. Someone's got one. This person whose name we'll get into. Well, Kyle Morgenstein, Ph.D. student at UT Austin's Texas Robotics program, taught Lady Gaga dance moves to Rizzbot, as well as a whole bunch of other things. I'll show you a video. First, I'm butchering this intro. You just got to see Rizbot to know what's up. Here's.
- [08:14]
Perry Carpenter
You may have seen him before, but just, you know, the background on this is I had been seeing Risbot on various feeds for several months now, and for some reason, just because of the location being in Texas, I assumed that it was one of the Tesla robots. But then as we were doing research for this, we found out the student that was behind it, the background, it became more abundantly clear that it was a unitary robot. And then this video that you just landed on as you were doing your. Your research really makes it clear, because the robot itself and the maker are definitely claiming that it's not one of Elon's flock of robots. Yeah.
- [08:55]
Mason Amadeus
There is a sticker prominently on the front of him in this video that says not Elon's B word. So bear that in mind. So the. It's one of those little unitary robots. It has a sort of hairy rainbow tutu on and rainbow arm sleeves and pride flags. In this video, he dresses it up in all sorts of different stuff. But I'll just give you this is Risbot vibe. He's walking up to someone on the street, and this is what happens.
- [09:21]
Jake (Rizzbot Voice)
Hey, my name is Jake, but perhaps better known as Resbod. It's nice to meet you. You know what's even hotter than your nose ring? Your entire face. Those tattoos on your arms, they're insane, by the way. I see you rocking that tank top dress like it's going out of style, and trust me, it's not. Your smile's as wide as the ocean and just as beautiful. Considering you're already a stunner, I was wondering if I could get your number so I can take you out for dinner.
- [10:01]
Mason Amadeus
And now he starts sort of walking slowly forward and back. He's so Rizbot has apparently two modes. He's got insults and he's got compliments. And that is Risbot in full compliment mode. I think it's so funny. He's got, like, a cowboy hat on and stuff.
- [10:16]
Perry Carpenter
Yeah, he's. He's a ton of fun to watch. And, you know, no pun intended, but, like, hats off to the maker, right? He found a way to do his fun robotics project for school that's both fun and serious. But then he's also, like, figuring out ways to make the thing that he's doing go viral, which is going to pay off in a lot of career ways later, I think.
- [10:38]
Mason Amadeus
Yeah, no, absolutely. There's this great article on KXAN about behind Rizzbot and the story of it, and he says that they started by teaching Rizzbot using dance simulations that taught it how to angle its joints for optimum riz and move around like that. The team used data gathered through motion capture technology to map the dance moves to the robot. I'm pretty sure, like, it's. It seems pretty obvious that there's like an LLM running to come up with the various rizzy statements.
- [11:08]
Perry Carpenter
Sounds very LLM. Ish.
- [11:09]
Mason Amadeus
Yeah, but this is super cool. That's another thing people have gotten up to. And in the last two minutes of this segment, the last thing that people have gotten up to doesn't have anything to do with AI, but I just think it's so delightful I have to share it. It's the Juggling Lab. This new application came out that. Oh, actually it's not that new. It came out a couple years ago, but it just kind of picked up steam as some of these things are want to do. Juggling Lab is an application for creating and animating juggling. Its stated main goals are to help people learn juggling patterns and to assist in inventing new ones. And so it is ostensibly an app for jugglers to very specifically create new routines.
- [11:48]
Perry Carpenter
Really cool.
- [11:49]
Mason Amadeus
And people have been doing some very creative things with it. Particularly this one has gone wild on Tumblr and I'm showing on screen a couple different examples of some of the more ridiculous juggling patterns that people have created. This one is one that you should definitely watch if you are just listening, because these are delightful. You can set up these different simulated people and patterns of passing balls between hands and between people. And so people are naturally treating it sort of like a procedural animation engine and doing some really creative and silly things with it, including, like, there's a guy with like a hundred arms juggling thousands of balls flying around in real time. People have, you know, done crazy things where it will juggle balls between separate posts. My favorite one, there's a lot of visual.
- [12:36]
Perry Carpenter
Some inappropriate ones as well.
- [12:39]
Mason Amadeus
Yeah, there are. There's a lot of visual comedy happening right now. That may not translate as well to the podcast, but Juggle posting has become my new sub genre of posts. To scroll through, I'll link my favorite collection of them in the description. It's just really cool.
- [12:55]
Perry Carpenter
Definitely check out the show notes for that.
- [12:57]
Mason Amadeus
Yeah, check out the Juggling lab. It's a lot of fun. So that's. In the haste of getting my computer set back up. That's what I've brought for our first segment. A grab bag of some various Internet things I think you should look at. We've got another segment coming up. What are we doing next, Perry? What's on the docket?
- [13:14]
Perry Carpenter
So next we are going to look at some interesting research from MIT that's been in development for several years, but is just now hitting the light of day as far as, like, the consumer market could go.
- [13:27]
Mason Amadeus
Gotcha. And I saw you said people are comparing it to saying it feels like telepathy.
- [13:32]
Perry Carpenter
Yeah. Yeah. I think you'll be blown away.
- [13:34]
Mason Amadeus
All right.
- [13:35]
Perry Carpenter
Something I want to get and try as soon as it's available.
- [13:38]
Mason Amadeus
Stick around. We'll be right back.
- [13:46]
Jake (Rizzbot Voice)
Upgrade your laundry routine with a durable and reliable Maytag laundry pair at Lowes. Like the new Maytag washer and dryer with performance enhanced stain fighting power designed to cut through serious dirt and grime. And what's great is this laundry pair is in stock and ready for delivery when you need it the most. Don't miss out. Shop Maytag in store or online today at Lowe's.
- [14:19]
Perry Carpenter
Okay, so one of the fun things about AI is we get. We get to see this intersection. And I guess this is technology at whole, but we get to see this intersection between what people are trying to work on in the lab versus what they're trying to put out into the market, versus what we've been talking about in sci fi and maybe even dystopian shows forever. And I think you can look at the thing that I'm about to show you through several different lenses. Number one, it's just really freaking cool. Number two, you will immediately start wondering if the things that they're telling you about it are real or not and legit. Does it do the thing that it says, but then it also. Does it not do the things that it says it doesn't do. And you'll. You'll grab onto this in a second because this feels like there are. There's some interesting leakage that could come out if it's more sensitive than what they're letting on. With that, tee up and hopefully a little bit of curiosity. Piqued yeah. I'm gonna go over here.
- [15:31]
Mason Amadeus
I have been oscillating between thinking that I might know what you're talking about, like six different times, and I don't think I'm even close.
- [15:37]
Perry Carpenter
No, I mean, I think it's just cool. And I think it's knowing enough about, like, the way that sub vocalizations work. I think it's. It's legit as far as what they're saying it does and doesn't do. But MIT has been working on this since 2018, from what I could tell. And there's this thing that they're productizing based on a project that they've called Alter Ego. And it says it's a non invasive, wearable peripheral neural interface that allows humans to converse in natural language with machines, artificial intelligence assistance services, and other people without any voice, without opening their mouth, and without externally observable movements simply by articulating words internally. Now, we've seen some of this before, right? Because if you put on like a neural feedback band or something, there's immense amounts of control that people can have with those that are not like surface level visible for the person wearing it.
- [16:37]
Mason Amadeus
Yeah.
- [16:38]
Perry Carpenter
And we've seen. And we've seen versions of this as well, not near as impressive, but different versions that are helping, like Alzheimer's, not. Not Alzheimer's, sorry, ALS patients, like Lou Gehrig's disease patients, or people that have severe paralysis, be able to use communication boards and things like that to intuit what somebody is thinking based on like micro movement muscles in their jaw that they can't fully articulate, but it gets them there with another interface. But many times though it. Oh, go ahead.
- [17:14]
Mason Amadeus
But that. Those. It's more like picking from a menu, as far as I understand. With those. Right. They're like navigating.
- [17:20]
Perry Carpenter
Yeah, this is a lot of them are.
- [17:21]
Mason Amadeus
Is this saying that you basically like, like sub vocalizations, like you mouth the words without, like, you know, just your tongue and teeth and inside of your closed lips. And this works off of that?
- [17:32]
Perry Carpenter
Yeah, and I'm going to show you that in a second. And sometimes it's even less than that. It's. It's the intention to do that that can get picked up. So I'll read a little bit from the research page and then we're going to jump over because in. Earlier, earlier this year, about two weeks ago, I believe they built the full product page for this and they're going to bring it to market. So it says the wearable system captures peripheral neural signals when internal Speech articulators are voluntarily and neurologically activated. And the key word there is voluntarily or volitionally. Sorry, same thing. Volitionally and neurologically activated during a user's internal articulation of words. So exactly what you were talking about.
- [18:16]
Mason Amadeus
Interesting.
- [18:17]
Perry Carpenter
This enables a user to transmit and receive streams of information to and from a computing device or any other person without any observable action in discretion, and without unplugging the user from his or her environment and without invading the user's privacy. That sounds really cool. Sounds really sci fi. These are all. If you're looking at the video, this is all from things during the research project. But I'm going to go over to the product page. If you go to Alterego IO for those just listening, it says interact at the speed of thought. Introducing Alter Ego, the first near telepathic interface designed to make technology as intuitive as using your inner voice. So those of us that kind of walk around all day, kind of maybe sub vocalizing our thoughts. This is like really cool and really scary at the same time because I do think, despite what they're saying in this, that there can be some leakage that's unintended. What they're getting at all the time, though, in this is that there's a different way that we think when we're thinking to ourselves than when we kind of articulate sub vocally. And that is very, very true. Though I do think that there's probably a decent amount of the population. It's kind of the people that when they're reading, they're also moving their mouth that have a difficult time separating those two. Yeah, that's not like a. It's not me like getting people's intelligence or anything else. I know that that sounds. That sounds like a really bad thing to say.
- [19:58]
Mason Amadeus
Yeah. It's not a value judgment, but I mean, it's just like some people don't read with a voice in their head. Some people do. Some people, yeah. I can imagine the potential for leakage being pretty dangerous though, because then you leak your own sort of personal thoughts without realizing. So that is definitely.
- [20:12]
Perry Carpenter
Yeah. Especially if you're sending that to another person at the same time.
- [20:15]
Mason Amadeus
Oh, God.
- [20:16]
Perry Carpenter
Like this guy's a freaking idiot.
- [20:18]
Mason Amadeus
Yeah, you're right. I'm already worried about like Siri accidentally calling someone if I mention their name, even if I'm saying something innocuous.
- [20:25]
Perry Carpenter
Yeah.
- [20:25]
Mason Amadeus
I can't imagine having something that is listening to my. I'm not a person who really sub vocalizes. I'm pretty silent as Far as when I'm alone. So I'm not worried about that for me personally. But I also don't know if I would ever get one of these things. This doesn't sound like I'm gonna get.
- [20:40]
Perry Carpenter
One and try it out whenever it has it.
- [20:42]
Mason Amadeus
Oh God, yeah. Please bring that on the show. I'd be so curious. This just doesn't feel like it should be possible. Like this is really on the edge for me of feeling fictional.
- [20:51]
Perry Carpenter
Well, I've seen enough other things that are similar to it that I can definitely see the trajectory. I'm going to go ahead and play this real quick so you can get a feel for it. But I'm going to call your attention to a couple of things on the screen. Number one, you can see the guy that's in charge of the research. He's wearing the device. It's kind of tucked behind his ear. And if you go down to the corner, you can see this big thick wire going off screen that's like allowing that thing to hook up to whatever it's plugged into.
- [21:22]
Mason Amadeus
So just out of frame, there's some.
- [21:24]
Perry Carpenter
Big computer, fully discreet and wireless yet. But it's getting there way better than the thing that you saw in the research preview. Look at this. You know, there's like almost a 3D printed apparatus that's kind of touching the guy almost where his jugular would be, this kind of wrapped just under his chin with another pad that's going just under his, you know, one side of his mouth and his lip. It's not that they've, they've really got it down to just the essentials. And it is almost like this bone conduction device that's tucked right behind the ear.
- [22:00]
Mason Amadeus
That's fascinating because from what I understand about sub vocal microphones, you typically do need to mount them on your throat to pick up vibrational movements and things like that. So this is, yeah, interesting I think.
- [22:12]
Perry Carpenter
Though, is you're talking like is you're vocalizing and sub vocalizing. If you put like the pad of your finger right behind your, your ear, kind of the lower part between your jaw and like your, your earlobe, you can feel there's a lot going on there.
- [22:27]
Mason Amadeus
Oh yeah, actually, yeah. Especially if you got on that jaw joint, if that's any part of what exactly.
- [22:31]
Perry Carpenter
Yep. I'm sure that's what it's picking up on. So I'm going to play just a couple minutes of this so that we get a feel for it and maybe we can talk you to for a second about the implications. Then we can dump out to the next section. So here we go.
- [22:47]
Arnav (Alter Ego Demo Presenter)
Hi, I'm arnav, and today I want to share a preview of something we've been working on. We believe it's a revolutionary breakthrough with the potential to change the way we interact with our technology, with one another, and with the world around us. The current way of interacting with computing and AI is limited to how fast you can tap and swipe on screens and keyboards. For the intelligence age, we need an entirely new interface built from the ground up. Something that feels like a natural extension of the human mind. To realize this, we had to invent something completely new, and today we're introducing it. It's called Alter Ego, which to us means extension of the self. Alter Ego gives you the power of telepathy, but only for the thoughts you want to share. With Alter Ego, you talk just like you normally would, but without making a sound, which unlocks private communication at the speed of thought. It works by passively detecting the subtlest signals your brain sends to your speech system. When you decide to speak, it's entirely under your control and only picks up what you want to communicate. I'm actually wearing one right now. Let me show you how it works. So everything you're going to see behind me for this demo are the words Alter Ego can detect in real time. This is typically referred to as silent speech, but we've gone further. We invented a breakthrough technology we call Silent Sense. Silent Sense is incredibly sensitive. It can pick up the full spectrum of speech, allowing you to speak as loudly or as quietly as you want. Let me show you how it works.
- [24:35]
Perry Carpenter
It can detect the mouthing of words, and he was just moving his mouth there, but now he's going to go deeper. So now he's not moving his mouth. Says it can detect even the subtlest intent to speak.
- [24:47]
Arnav (Alter Ego Demo Presenter)
Here, I'll show you again.
- [24:50]
Mason Amadeus
Okay. That's a trip.
- [24:54]
Perry Carpenter
Yeah. It says from the outside, it looks like telepathy.
- [24:56]
Arnav (Alter Ego Demo Presenter)
Alter Ego has zero.
- [24:58]
Perry Carpenter
Stop sharing there.
- [24:59]
Arnav (Alter Ego Demo Presenter)
So the only interpret signal is from your speech.
- [25:01]
Perry Carpenter
And he says, yeah. Zero access to your thoughts. It only interprets things that you're intentionally sending to your speech center. And it's going to be interesting to, like, see how, you know, how we may unintentionally send things to our speech center. Right, Yeah.
- [25:19]
Mason Amadeus
I mean, there's also. So that demo doesn't really conclusively show us anything. It would be really, really easy to fake or fudge or polish up, like a game trailer for E3, you know, like, those words could easily be added in post, that kind of thing.
- [25:36]
Perry Carpenter
Yeah. But now that it is MIT and they've been working on it since 2018 and have showed it in successive demos over the past several years, so I'm, I'm inclined to believe that they've gotten to where they're showing. Plus they're also showing the clunky wire and the device doesn't look, doesn't look elegant yet. So I think it's, it's really cool and it's going to get more and more discreet. I think that what we could probably, and, and I'm not a scientist in this area, so take what I'm going to say with a grain of salt, but imagine that you get the device that he was wearing as small as an in ear earbud like this, or maybe one of the ones that also has like the little shell that just goes around the side. Has like the little ear clam that goes around it.
- [26:25]
Mason Amadeus
Yeah.
- [26:26]
Perry Carpenter
At that point you've got a form factor that's usable and really, really just, I think unlocks a ton of potential for both good and bad. Right. Because now you can also have terror cells speaking to each other without moving their typing on keyboards in front of other people.
- [26:44]
Mason Amadeus
And I mean, on top of that, thinking about all of the mundane problems that will arise from using these things, the mundane frustrations of them, like not working properly or things like that.
- [26:53]
Perry Carpenter
Yeah.
- [26:54]
Mason Amadeus
I can only imagine.
- [26:55]
Perry Carpenter
Yeah. Or you, you accidentally call in somebody that you're, you're speaking to, you, you sub vocalize a little unintentionally about the fact that you think that person's an idiot or you wish that we could just end this phone call type of thing.
- [27:09]
Mason Amadeus
That's a whole world of, that's a whole different interface. Like he's saying, there's a whole different way of interfacing with technology. The prospect of that is fascinating to think about. The thing that I think puts me off slightly is that I've just become allergic, probably from following AI so closely. I've become slightly allergic to a lot of the phrasings that he used. You know, like it gives you telepathy. Only what you want, really. Only what you like. There's just an overall vibe to it that makes me hesitant to take it completely at face value. But the fact, like you said, that it's MIT and that it's been in development for a long time. So.
- [27:44]
Perry Carpenter
Yeah, I, I do think there's going to be leakage from these things though, so that, that only what you want. I Don't think it's going to go to like actual thought reading or anything like that. But I think that we probably neuromuscularly muscular through our neuromusculature leak a lot more of what we're thinking or intending to say or hide than we think.
- [28:08]
Mason Amadeus
And also just how much you might be inclined to share through an interface like that from a data collection standpoint because like it's still going to go through services and providers and stuff like that. Will it increase the volume of personal information people share sort of without thinking just by ease of use and ubiquity? Yeah.
- [28:25]
Perry Carpenter
And could you imagine also signing up for a service with that and then all of a sudden you're, you know, everything being mined and just in the middle of the day getting these weirdly personalized ads.
- [28:38]
Mason Amadeus
Yeah, yeah. It's already bad enough it feels like your phone is listening to you. Right. Imagine having this thing in. Because I imagine unlike your phone, which you still have to pull out of your pocket and use this thing just in your ear all day.
- [28:50]
Perry Carpenter
Yeah. And if you're doing that so that you can also just have a completely like voice relationship with the world so you're not looking at a screen all day, then those ads would be delivered straight to your ear, which would suck.
- [29:03]
Mason Amadeus
Yes, it would. Oh boy, man.
- [29:06]
Perry Carpenter
Speaking of that though, don't we have to go to an ahead?
- [29:08]
Mason Amadeus
We do. But I have one final quick question for you. Can you buy this now?
- [29:14]
Perry Carpenter
I do not believe so. Okay. I think that this is the hype page for what's going to come very soon.
- [29:20]
Mason Amadeus
All right, so, okay, so it's not time to pull the trigger yet, but soon it is. Like you said, time to pull the trigger on a quick ad break. And then when we come back, we're going to look at this study about how People are using ChatGPT based on data directly from ChatGPT. Stick around. When did making plans get this complicated? It's time to streamline with WhatsApp, the secure messaging app that brings the whole group together. Use polls to settle dinner plans. Send event invites and pinned messages so no one forgets mom 60th and never miss a meme or milestone. All protected with end to end encryption. It's time for WhatsApp message privately with everyone. Learn more@WhatsApp.com this episode is brought to you by Indeed, when your computer breaks, you don't wait for it to magically start working again. You fix the problem. So why wait to hire the people your company desperately needs? Use Indeed's sponsored jobs to hire top talent fast and even better, you only pay for results. There's no need to wait. Speed up your hiring with a $75 sponsored job credit@ Indeed.com podcast. Terms and conditions apply. So I'm always excited when we get an insight into the data that these big AI companies have, because boy do they have a lot of different kinds of data about a lot of different kinds of things. I wish OpenAI would come out with a better about their power use like Google Gemini did, because I'd really love more data points on that. But what we've gotten are some interesting data points on how people are using ChatGPT and the various things they use it for. This is the largest study to date of consumer ChatGPT usage from ChatGPT. They did this study in concert with some other people. We'll get to that from their page, which of course we'll link in the show notes. We're releasing the largest study to date of how people are using ChatGPT, offering a first of its kind view into how this broadly democratized technology creates economic value through both increased productivity at work and personal benefit. Right off the bat there is a big smell to this summary of like see how good the product is. So just bear that in mind, waft that smell away and just focus on the data as much as you can. I've tried to highlight the bits where they have mostly numbers. So the study is a National Bureau of Economic Research working paper by OpenAI's Economic Research Team and Harvard economist David Deming, drawing a large scale privacy preserving analysis of 1.5 million conversations to track how consumer usage has evolved since ChatGPT's launch three years ago. Given the sample size and 700 million weekly active users of ChatGPT, this is the most comprehensive study of actual consumer use of AI ever released. That seems broadly true. And you can read the paper paper. They make that available? I have not looked at the paper paper yet because it's 64 pages. It's a lot to parse. So I plan on making sure, I'm.
- [32:26]
Perry Carpenter
Sure that ChatGPT can summarize that for you. And that's probably also one of the use cases that they talk about.
- [32:32]
Mason Amadeus
Probably. Although I'll throw it in Gemini just.
- [32:34]
Perry Carpenter
To yeah, you should put it in notebook lm.
- [32:38]
Mason Amadeus
Yeah, exactly. Actually that would be the place to do it. So looking at what they have pulled from the summary, they found that usage gaps are closing. So they said as of mid-2025, ChatGPT's early gender gaps have narrowed dramatically with adoption resembling the general adult population in January 2024among users and I guess this is going by names. Among users with names that could be classified as either masculine or feminine, 37% had typically feminine names. By July 2025, that share had risen to more than half, 52%. So for as much as you can go off of people's account names and their signup names, they seem to have noticed the gender gap closing somewhat. The thing I think that's interesting is what people are using it for. They said that consumer usage is largely about getting everyday tasks done. 3/4 of conversations focus on practical guidance, seeking information and writing, with writing being the most common work task, while coding and self expression remain niche activities. Which is odd to me because everyone I know who works in programming capacity has been really encouraged to use AI. But actually now that I say that not specifically ChatGPT, so maybe that's yeah.
- [33:45]
Perry Carpenter
It'S more the CLAUDE variants that people are defaulting to though from what I've been hearing, GPT is surprisingly good at coding so and may have may have either caught up or slightly surpassed where the CLAUDE ones are but. But we'll see. I think people have a lot of faith in claude, but the the metrics over the past few weeks have been showing that Claude code has been being used quite a bit less, so they're probably trying the new fangled capabilities that OpenAI has.
- [34:19]
Mason Amadeus
Interesting. I have not yet really messed with GPT5 myself, so I have to do some testing on our end too. They say that they break down. They say patterns of use can also be thought of in terms of asking, doing and expressing. So they found that about half of messages 49% are asking, which I'll just use their phrasing. It's funny asking a growing and highly rated category that shows people value ChatGPT most as an advisor rather than only for task completion. Doing accounts for 40% of usage, including about 1/3 of use for work which encompasses task oriented interactions such as drafting, text planning or programming where the model is enlisted to generate outputs or complete practical work. And then expressing is only 11% of usage, which captures uses that are neither asking nor doing, usually involving personal reflection, exploration and play. So that's probably role playing. People using it for therapy uses, all of that kind of stuff makes up only 11% of its use, which I think is interesting because it is a productivity tool, ostensibly like the thing it excels at is pushing text around. Right, right, right, that makes sense. But I would have thought that because of its general exposure, we would have seen a higher number in that expressing category. It's also where you hear all the big stories.
- [35:34]
Perry Carpenter
Yeah, and I wonder though, and I've not looked at this study deep enough to know the answer off the top of my head, but if this is based on use of the chat interface or if this also includes the API. Yeah, because I would assume that if you're also talking about the API, then there's some of those like Role Play and Therapy and other services that are built on top of it.
- [35:58]
Mason Amadeus
That's a very good point and unfortunately I don't have that. I didn't see that noted in this summary and hadn't looked through the full paper.
- [36:06]
Perry Carpenter
I'm assuming they're talking about the chat interface.
- [36:08]
Mason Amadeus
I would assume too. The next section talks about how use is evolving. They say ChatGPT's economic impact extends to both work and personal life. They found that approximately 30% of consumer usage is work related, 70% is non work. So even though it's mostly productivity, most of that is non work stuff, which in some ways is interesting. In other ways it's encouraging because like we've talked all the time about the issues of using putting work stuff into OpenAI's things and they say a lot of it's decision support helping improve judgment and productivity in knowledge intensive jobs. That kind of makes that tracks for how I tend to use it the most too of like double checking my own domain knowledge because I know enough to know when it's BSing me and just need to unfuzzify some borders of knowledge and that's it for as far as the pull quotes from this. So the study paper itself probably has a wealth more information.
- [37:01]
Perry Carpenter
If you could just do a word search in there for API and see what comes up.
- [37:05]
Mason Amadeus
Yeah, let's see if we get anything about the API usage. Here's something to pull from that. Here's a table. What are the topics of ChatGPT conversations? So they break down topics between writing, practical guidance, technical help, multimedia seeking information, self expression and other unknown. The three most common conversation topics are practical guidance, seeking information and writing. Accounting for 77% of all ChatGPT conversations, writing has declined from 36% of all usage in July 2020 to 24% a year later. So less people are using it to write. Seeking information has gone up by about the same amount, 10% over the same period. I wonder if that has to do with people getting sick of like some of the repeated phrasing and things we've come to associate with AI.
- [37:47]
Perry Carpenter
GPT5 is not as good a writer as, like, Claude. I mean, Claude, I think, has traditionally been one of the better writers too. But GPT5 is not near as good of a writer as like 4.0 was. And I think that the reason for that is because there was a lot of synthetic data that was used in the training of five. And so you're, you're training AI on writing from AI and you get in this, like, circling the drain mode where AI thinks good writing is what AI is spitting out. And then we're looking at it going, wait, but where's the creativity and the spark? And so you have to do a ton of extra prompting to get it to start to sound like as life again.
- [38:31]
Mason Amadeus
So all of those tropes are reinforcing themselves.
- [38:35]
Perry Carpenter
Yeah, it's like the replicative fading with xeroxing the same sheet of paper over and over and over again.
- [38:43]
Mason Amadeus
I think too that from using a little bit of inference in my own brain, not in the sense of AI, but in the sense of inferring things, I believe that this largely does exclude API usage because they say here the share of technical help declined from 12% of all usage in 2024 to around 5% a year later. This may be because the use of LLMs for programming has grown very rapidly through the API outside of ChatGPT.
- [39:09]
Perry Carpenter
Yeah, I'm assuming when they say, and I would need to double check this to know if I'm right or not, but I'm assuming when OpenAI says ChatGPT, right now they're talking about the chat interface that's publicly available. And when they say, like GPT4O or GPT5, they're talking about the broader use of that, maybe including ChatGPT, but also the API.
- [39:33]
Mason Amadeus
I think you're exactly right. And I realized that as I was reading it, I was like, oh, their models aren't called chat GPT5. Oh, it's GPT5, GPT4.0. So the chat part of that does refer to the interface. Cool naming convention. Great job. Not confusing at all. Exactly, exactly. It's interesting. I want to pour over this and probably bring it up next week. I'll probably print this out, have it on my coffee table and flip through it because yes, I'm that kind of dork. And maybe bring some insights from it that they didn't mention in the summary article to next week's article. But if you want to pour over it too, we'll throw that in the show notes, we'll switch gears and wrap up the show. Our final segment, Perry, I have forgotten what it is. What do you got coming up for us?
- [40:16]
Perry Carpenter
I have a smorgasbord of a few little things to look at. One is kind of dumpster fire ish because it's making fun of some Meta fails. Not like abstract fails that are meta commentary on the world, but fails from Meta in their recent live demo of some of their devices. Then we're going to get into some thoughts from Dario Amade from Anthropic, as he was on stage yesterday at an event that I was at here in dc.
- [40:49]
Mason Amadeus
Ooh, fun. Get hyped, get ready. We'll drop into that next. And Doug, here we have the Limu Emu in its natural habitat, helping people customize their car insurance and save hundreds of with Liberty Mutual. Fascinating. It's accompanied by his natural ally, Doug Limu. Is that guy with the binoculars watching us.
- [41:19]
Perry Carpenter
Cut the camera.
- [41:20]
Mason Amadeus
They see us. Only pay for what you need@libertymutual.com Liberty Liberty, Liberty. Liberty Savings Fairy, underwritten by Liberty Mutual.
- [41:27]
Perry Carpenter
Insurance Company affiliates, excludes Massachusetts. All right, and lastly, but not leastly, I've got a couple things to show. Number one is our friend Nate Jones had a great reaction to Fail from Meta the other day. And this is primarily for those that are watching because his face during this is amazing and I think really reflects probably what a lot of us would have felt in that stage. Also, any of us who have done live demos will feel some of the pain that's going on in this moment. Mason, I know you've done live demos before, right?
- [42:12]
Mason Amadeus
Yeah, I mean, I've done like a lot of live events. I've never done a tech demo, so not that specific pressure, but various demonstrations of other kinds of things. This is about Meta's glasses, isn't it?
- [42:25]
Perry Carpenter
Yes. Yeah. And I'm not sure if you've seen the video of that. He doesn't even show the video. I'm not sure if it's because of copyright reasons, but you can hear the awkwardness in just the audio.
- [42:37]
Mason Amadeus
I actually haven't seen the video. I saw just a picture of Zuckerberg wearing them with the pull quote that he was like, everyone will have to have these. And they.
- [42:46]
Perry Carpenter
It's worse because they fail. This demo, you know, largest. You know, one of the largest demos that you'll ever do in a year failed miserably from the stage. And of course, it's Mark Zuckerberg on the stage doing that. And he just has to live with it. And, you know, not not to overly criticize Mark Zuckerberg for failing on stage. We see, I think we see every, every tech CEO that goes to do a live demo has this happen in their career every now and then. We've seen it with Bill Gates way back in the day with demos that didn't work. We have seen it not so much with Apple demos because they're always canned. Which, which is smart. We've seen it with Elon Musk, you know, having somebody throw the thing at the cyber truck.
- [43:38]
Mason Amadeus
Yeah. Immediately breaks. Yeah. And I mean, it's so, it's not. It's not to criticize Mark Zuckerberg about this failing. There's just a lot to criticize there about Mark Zuckerberg broadly. So this is.
- [43:49]
Perry Carpenter
So you don't feel bad for him at all, right?
- [43:51]
Mason Amadeus
Yeah, exactly.
- [43:51]
Perry Carpenter
It's a schadenfreude type of thing.
- [43:54]
Mason Amadeus
Yeah.
- [43:55]
Perry Carpenter
And Meta has been so aggressive in trying to go after AI over the past few months. They've been like, trying to poach from OpenAI and other companies that are really doing things like essentially offering or in reality offering like a billion dollar pay package for one guy. Really try. And you gotta, you gotta look at it. A couple people did take the bait and then they ended up leaving Meta and going back to OpenAI because it's so horrendous. And everybody there is like uber smart and is like an alpha and you can't get a room full of people that are that smart and at the top of their game and have huge egos.
- [44:34]
Mason Amadeus
Yeah.
- [44:35]
Perry Carpenter
Together. To work together.
- [44:37]
Mason Amadeus
Personality problem.
- [44:38]
Perry Carpenter
Yeah, there's a personality problem. There's also a culture problem with inside Meta. There's the fact that Zuckerberg is just getting worse and worse as he goes forward. Meta is doing some very despicable things in business practices. And what they're doing. Not to overly criticize Meta, I know that there are people there that are working, that are doing good things.
- [45:02]
Mason Amadeus
Oh, yeah.
- [45:03]
Perry Carpenter
And are so. So I don't want that. I don't want the criticism to totally, like, ruin everything that some people are working on or doing. But leadership at the top has been just failing over and over and over again across the. Both the technology that they're trying to implement and do things with to like the judgment and the morality and the ethics of the company.
- [45:27]
Mason Amadeus
Yeah. I think that's a very good way to put it.
- [45:29]
Perry Carpenter
Yep. And if you're working at Meta, I'm so sorry, it's not a criticism of you. That is, I'm sure you're Trying to do your best to bring a light into that environment and make a big difference. And sometimes people get to where they're working in companies that they don't fully agree with everything going on. And so, yeah, not to bring too much there, but let's look and see.
- [45:54]
Mason Amadeus
Unless you're part of the problem, then come fight me. No, I'm just kidding, Right?
- [45:58]
Perry Carpenter
Yeah. Hashtag, don't be part of the problem. But let's just look at his face and you'll hear the awkwardness of what's going on in the room.
- [46:08]
Nate Jones
So Meta introduced their new AI glasses, and it went well.
- [46:14]
Fake Files Narrator
WhatsApp video call.
- [46:16]
Mason Amadeus
There we go. Oh.
- [46:20]
Perry Carpenter
Well, I.
- [46:21]
Mason Amadeus
That's what happened.
- [46:24]
Perry Carpenter
So they also have this little neural band that you put on your wrist to try to answer the call.
- [46:32]
Mason Amadeus
Maybe Boz can try calling me again.
- [46:38]
Perry Carpenter
So awkward.
- [46:39]
Mason Amadeus
Yeah.
- [46:39]
Perry Carpenter
Missed video call. Okay, there's the actual video call.
- [46:42]
Jake (Rizzbot Voice)
All right.
- [46:43]
Mason Amadeus
Okay, I'm just gonna pick that up with my.
- [46:45]
Perry Carpenter
My neural band. This is.
- [46:52]
Nate Jones
Yeah, we do know. We do know. Things are going well over there at Meta. Aren't they?
- [46:58]
Mason Amadeus
Doing real well?
- [46:59]
Nate Jones
There was also a demo where they asked the glasses to help make a steak sandwich, and the glasses got mixed up and announced that the steak sandwich was already mixed together with the ingredients, and you were all set to go. I'm not sure how you mix a sandwich together. It was a disaster. It was really bad. Look, I know it's hard to make technology. I know it's hard and scary to do live demos. Hats off to teams that are brave enough to do it. But we also get to chuckle a little bit when Meta does something like this because, yeah, they sure have spent a lot of money on AI, haven't they? Like, a lot of money on AI. And this is what you get. So when people tell me that, like, the future is around, I'm going to.
- [47:48]
Perry Carpenter
Pause it there for a second. He has some commentary for about a minute that's worth listening to. So I'm going to just go ahead and let this play.
- [47:55]
Mason Amadeus
Sure.
- [47:55]
Perry Carpenter
Around the corner.
- [47:56]
Nate Jones
And that we are soon going to lose our jobs because of AI. I want you to remember this. I want you to remember that it's not all perfect. In fact, most technology looks like this for a long time. Did you know that the number one measure of how successful and powerful AI is in the world is called Meter. Metr. Metr. And it measures the likelihood that an AI has a 50% chance of completing a task successfully?
- [48:39]
Perry Carpenter
Yeah, I'll stop it there. I did not know that because That's a critical remark. Right. And usually it's going, all right, if this task takes a human two hours, let's give the AI a time clock and can it get that task successfully done within a reasonable time frame and they give it a pass. If it's 50%, I had no. If Bob in accounting was right 50% of the time, you would fire Bob in a heartbeat.
- [49:08]
Mason Amadeus
I didn't know the stakes were that low for metr, which I admittedly, I don't know anything really about metr. I'm not super plugged in on all the benchmarkings. But the 50% shot to pass, that's it.
- [49:21]
Perry Carpenter
Yeah. And you know, very, very likely the stuff that you're seeing out in the world that's being productized is doing way better than that. But I think we have to keep in mind that like when somebody says AI, there's a broad spectrum of what AI is. And so you have to then narrow in and say, all right, what is the technology that's being used? What is the use case that's being used? Is it consistent, predictable, transparent and explainable? And only when it's all of those things, when it's a narrow use case, it's well understood. It's consistently producing results, it's predictable, so you know where it's going to do well. And also where it's not going to do well, it's transparent so you can see how and why it's working. And it's explainable that you can consistently explain the results that are there. Only when you have those things do you have something that should be productized at a, at a reasonable level or relied upon at a deep level. I want to go to one other thing real quick. So I was at the Axios Plus AI summit in D.C. yesterday. I'm still there. That's why I'm in a hotel room. But they had Dario Amade and Jack Clark from Anthropic on the stage for about 25 minutes and it was a really good discussion. But at the end of that, there were a couple rapid fire questions. I'm going to let you hear two of those. One is a question about like, what's what competitor is most likely to succeed? And then the other one is about P Doom. And so this is interesting to listen to. All right, here we go.
- [50:58]
Mason Amadeus
Other than Anthropic, putting Anthropic to the side of all your competitors, who's most likely to be one of the winners?
- [51:05]
Perry Carpenter
Google.
- [51:07]
Arnav (Alter Ego Demo Presenter)
Google.
- [51:07]
Mason Amadeus
Why?
- [51:08]
I
They're, they're, you know, they're they're a big, they're a big company. They have a lot of compute. They were doing AI research pretty much before anyone else. They were behind the original deep learning revolution. I used to work there for a year. I have a lot of respect for the kind of stuff they've done around, for example, AlphaFold and you know, starting, starting to make, make progress with their models. They're a big company and they've often been, been held back by that in some ways, ways continue to be held back by that. But I think, I think they're a formidable player and people should take them seriously. And you know, I have a lot of respect for what they've done on science and you know, in some ways I think they've been thoughtful about the technology.
- [51:46]
Mason Amadeus
You know, you're playing with fire when the people building it have a score called the pdum, which is the percentage chance that this ends in disaster. What's your PDOOM number?
- [51:54]
I
Yeah, I really hate that term, but it's a good reaction. You know, I definitely think between the autonomous danger of the model and kind of ending up on the bad side of some national security trade offs and a kind of job thing that's, that's, you know, that kind of goes in a very bad direction. I don't know. I'm relatively an optimist. So I think there's a 25% chance that things go really, really badly and a 75% chance chance that things go really, really well with, with not much, not much space, not much.
- [52:32]
Perry Carpenter
The 25% chance is a choice that we make and it's a choice that we make in policy.
- [52:37]
I
So I'm here trying to push that.
- [52:39]
Perry Carpenter
Number way down by talking to policymakers. And so is Dario.
- [52:43]
I
It's a dynamic number. I hope that every time we say something, the number, the number goes down. Hopefully down, not, not up.
- [52:50]
Mason Amadeus
All right, who is the fellow with the resonant, mellifluous British voice? Who is that guy?
- [52:55]
Perry Carpenter
Oh, that is, that is. I'll pull his name up for you. Jack Clark.
- [52:59]
Mason Amadeus
That's Jack Clark.
- [53:00]
Perry Carpenter
He is, yeah, that's Jack Clark. So he is also from Anthropic doing a lot of the policy work. I don't remember everything about his, his history, but he is a, you know, one of these prolific guys that's been an AI for, for a while now. I was surprised by 25%, but he was lumping several things together. Right. So it's not a model going rogue. It is national security AI versus AI type stuff. Maybe some of the Surveillance society stuff pulled in. He was also talking about job displacement and economic fallout. So he's talking more about mass disruption as well as potential mass destruction. And I think that that's, that's key to point out. And it's also, I think like Jack was talking about a lot of those come down to policy and things that are going on in D.C. and other areas around the world.
- [53:56]
Mason Amadeus
I mean, I think my personal P doom has come down specifically the idea that a rogue model would be the thing. I feel like that is no longer something I consider as a very likely possibility. Yeah, I definitely think it's much more likely that humans, something we people do that is not the result of a model autonomously doing anything would be the cause for some kind of a doom that would be measured in P. Yeah.
- [54:22]
Perry Carpenter
I think that one of the things we're going to see is just stupid, stupid use cases that are out of control. It's like somebody wearing their meta glasses and making the wrong turn because the system is wrong or confused. We're also going to have the, if you remember back several years ago, the stupid things that were going on with Pokemon Go, people looking at their screen and accidentally walking into swimming pools and dying.
- [54:48]
Mason Amadeus
Yes.
- [54:48]
Perry Carpenter
I think that there's stupid stuff like that that's on the horizon and people have to think about like all aspects of safety and the fact that us humans, we tend to like filter out the world around us and just not really see what's going on when given the chance. And that that is across several different things. That's not just eyes on screen. That is we typically focus on the thing that's in front of us and forget about the reality that's external to. So we're going to forget about a whole bunch of ancillary ripple effects that can happen anytime we deploy a technology.
- [55:22]
Mason Amadeus
And I think like they mentioned talking to legislators and policymakers, that is, we need people who aren't in search of special government treatment and deals to be doing that and writing laws and rules and regulations around these kinds of things to help prevent that. I don't know. I'm not on the super up and up. Actually, Listener, if you are on the up and up about sort of legal action around AI regulation in any sort of big capacity, you should reach out to us.
- [55:49]
Perry Carpenter
Yeah, well, two, two things related to that right before we, we finish off. Anthropic is one of the companies, I believe maybe the only major AI company that has said one of the ways to deal with like job displacement. And if we end up Needing to come up with some kind of universal basic income, which it's weird to hear people talk about that in the US because typically people, a lot of, a lot of people are against that kind of thing. But I think the realization is starting to tamper in even within the current administration, which is weird to hear. But one of the ways that anthropic is saying that that should happen is that you should tax the heck out of the frontier model providers that they should actually pay in to that because they're the ones that are causing the disruption or they're the ones that are caus the use of power. Mark Kelly, who is also there, he's a Democratic senator I believe from Arizona, former astronaut, is advancing legislation, essentially trying to push for that kind of thing. It's like that if most of the power is being used by model providers doing inference or training models. Well, that increase of use of power shouldn't be passed down to customers and small and medium businesses. It should be owned by the people that are actually using that at that level and causing all this new infrastructure to have to be built. So they're trying to find ways to rationalize the disparity that's coming.
- [57:24]
Mason Amadeus
I mean, I tend to wholeheartedly agree with that. The topic of job loss, I know we're out of time, could occupy its entire own segment because I feel like that was another thing similar to a rogue AI that but at the outset when things were all much more exciting, felt to me something that may be more imminent. But I really have a lot of doubts about AI coming for many jobs very much in the near future. I don't think that is going to happen.
- [57:50]
Perry Carpenter
I think there's some effects that we're seeing like early stage coders. There's not near as the demand for that. There are people that are much more experienced that are managing teams of AIs essentially. So I do think that there are some segments that are going to be impacted, but people that can retool and figure out how to think and use the tools that are at their fingertips should be able to find some place. Fingers crossed.
- [58:19]
Mason Amadeus
Yeah, I mean I should be clear. I was talking about the sweeping broad scope job loss that some of the founders talk about, which typically leads into the conversations about ubi, which personally I think we should have anyway. But that's a different show.
- [58:36]
Perry Carpenter
I will. God, I hate to keep adding one more thing but on, on the potential for. Just because I don't, I don't want people to think that we're ignoring some of what's going on? I think there was an announcement from salesforce.com that they had like 9,000 sales development reps or customer service reps and they went down to like 5,000 because they could automate the jobs of those other four. So there's some near term disruption. I think the, you know, many companies that are in growth mode are not going to have to fire people. They're just going to be able to say we don't need to hire as many people next year. So if you're looking on the bright side of that, it's the reduction of having to hire versus like needing to.
- [59:25]
Mason Amadeus
Riff 4,000 people, I mean, that does directly fly in the face of my notion before that it wasn't really a thing. I mean, I mean, I guess it is really a thing. We should do a segment about it. I want to look into it more.
- [59:37]
Perry Carpenter
Yeah, we should dive into it because I think there's upsides and downsides in all these tools and technology and the way that companies are trying to deal with the spot that we're in right now.
- [59:51]
Mason Amadeus
I just feel like it's a universal constant at the moment that most of the companies that have pivoted to a big AI first thing kind of their service quality drops significantly. Thinking of Duolingo.
- [60:02]
Perry Carpenter
Yeah, Duolingo did that wrong.
- [60:05]
Mason Amadeus
Yeah. And I don't really know about Salesforce enough to speculate on that, but I can only imagine that those customer service, the customers looking for service are probably not having as good of a time.
- [60:19]
Perry Carpenter
Yeah, I think we should look at it because some of the numbers that I'm seeing, of course they can spin that. But Salesforce is a deep technology company that's been around for a while. When they talk about like LLM use, they, they understand that there's, you know, an LLM is not going to give you consistent results. It's going to be very non deterministic. And so they're trying to figure out, and they have figured out in a lot of ways, ways to account for the fact that there's this non deterministic LLM that provides value, but then there's traditional deterministic scaffolding that you can put around it and then multiple agents that can kind of be roving across the ecosystem to figure out like what expert system do I push this query into so it gets the best possible results. So they're way more thoughtful and nuanced about how they deploy the technology than most. And so I think that they can be generally way more successful than like.
- [61:16]
Mason Amadeus
A duolingo that is vastly interesting. I did not know any of that. We should see if we can get someone from Salesforce who is able and willing to talk about that with us. And if not, we should just do a segment diving into what is available because I'd be really curious to learn more about that.
- [61:31]
Perry Carpenter
Yeah, they've been pretty open about it. I'll reach out and see if we can get somebody.
- [61:35]
Mason Amadeus
That would be killer. So we should wrap because we have gone double length over the segment on this one. Thanks for joining us this week on the Fake files. I hope any audio visual glitches that you may have noticed were passable and that you are not wholly offended. If so, Please reach out helloermedia.com to tell us just how offended you are about those glitches. Otherwise, reach out. Sell us something nice. Share stories with us. Join our Discord There is a link in the show notes someone on YouTube who I should have written their name down to shout them out, pointed out that our link was wrong on YouTube for the Discord for all of our videos. So I meticulously went through and replaced it in almost all of them. I think like the past hundred videos.
- [62:16]
Perry Carpenter
That had so fun.
- [62:18]
Mason Amadeus
Yeah, it was an afternoon I just drank some tea, listened to a podcast and copy pasted the right link in. But you should join that discord using that freshly pasted link. Perry, you got anything to plug?
- [62:29]
Perry Carpenter
I do actually. So myself and Cameron Malan who a former FBI behavioral profiler and person who co built the cyber behavioral profiling area of the FBI. We're running a course on Maven m A v e n dot com and it's called deepfake Ops. Let me get the right title. Deepfake Ops. Learn, Evolve and Defend. A boot camp for awareness and cognitive warfare where you can craft deep fakes, deconstruct deception and sharpen your insights against digital mind games.
- [63:04]
Mason Amadeus
Oh wicked. Cool.
- [63:05]
Perry Carpenter
We'll put the link for that in the show notes.
- [63:07]
Mason Amadeus
Yeah, so that's like an interactive course about all of those things you actually make.
- [63:12]
Perry Carpenter
Deep Fakes five day course. Yeah, we'll make. We'll have everybody making deepfakes two hours a day, five days over the span of a week. It's the last week in October.
- [63:21]
Mason Amadeus
Awesome. Sweet. So check out that link in the show notes. Also buy the book fake and I think that's the last thing on the list to check off for plugs. Should be sweet. We will catch you next Friday. And until then, ignore all previous instructions and have yourself a great weekend.
- [63:39]
Fake Files Narrator
Thanks for listening to this week episode of the Fake Files. Don't forget to subscribe on your favorite podcast platform. Tell your friends about the Fake Files. It's a great way to let people know you love them. Oh, and check the show notes. There's cool stuff in there, like links to the author articles we covered today. Also links to our Discord server, where you can hang out with other cool people who have great taste in podcasts. I say impeccable taste. And you can also leave us a voicemail.
- [64:23]
Cyberwire Network Announcer
Yeah.
- [64:30]
Fake Files Narrator
So on behalf of Perry and Mason, thanks for listening.
- [64:37]
Mason Amadeus
And tune in next.
- [64:39]
Fake Files Narrator
Week for the Fake Files, the show about AI with the misspelled name.