Loading summary
Transcript124 lines
- [00:00]
Packy McCormick
Pope gate. The debate over the Pope. The Pope is. He's a poster. I like it. He posts almost every day, sometimes like up to five times a day. Yeah, he's got range. He's got range. That's right. He'll tell you about. He'll pray for, you know, if there's a natural disaster, he'll pray for that. He talks about business, talks about AI, talks about media. He talks about all sorts of stuff. It's. It's a really great feat. He had a great post about. About media. What did he say about media? He said the media cannot and must not separate itself from the destin of truth. That hits.
- [00:35]
Tyler Cowen
Does this mean he's a neo factual media guy?
- [00:37]
Packy McCormick
I think so. I think he's one of us. Transparency of sources and ownership, accountability, quality, clarity and objectivity are the keys to truly opening citizens rights for all people. The world needs honest and courageous entrepreneurs and communicators who care for the common good. We sometimes hear the saying, the business is business. In reality it is not. So no one is absorbed by an organization to the point of becoming a mere cog or a simple function. This is the type of stuff you'd see on like a Pinterest board. It's like pretty generic, but it's hard to disagree with. Marc Andreessen was disagreeing with one of the Pope's takes about AI.
- [01:18]
Tyler Cowen
I think it is generally healthy that the Pope is going to comment and provide some sort of guidance or his own framework for how we should think about defining, developing AI. I think that seems healthy.
- [01:31]
Packy McCormick
The play by play here was Marc Andreessen, quote, posted that AI post with an image of Cat Stoffel, who's the GQ features director who went viral for interviewing Sydney Sweeney. People were actually kind of confused on what that particular meme means in this context.
- [01:50]
Tyler Cowen
With a meme template, there's two ways to read into it. There's like the. The actual visual, which is like what is the expression of the person's face. Right. You don't have to have watched the Big Short to understand somebody staring at the screen, just like confused.
- [02:07]
Packy McCormick
You're just saying I'm confused by this.
- [02:08]
Tyler Cowen
Yeah. So Michael Burry in the Big Short just looking confused at a screen. Right. And it has sort of obviously more meaning to that if you've seen the movie and you understand the full context. But somebody doesn't have to know the context now.
- [02:22]
Packy McCormick
So most of the timeline interpreted Marx post as the Pope is scolding AI builders and shouldn't be. Is that roughly the way you, you interpreted it, I think a lot of the timeline interpreted it as like the Pope is, is saying is like scolding AI builders. And there's been this other, there's. There was another like kind of low grade rumble on the timeline about like Brad Gerstner's comments about like D cells. I feel like I'm just pro moral discernment in AI development and also just pro moral discernment everywhere, I guess sort.
- [02:57]
Tyler Cowen
Of the philosophy for life.
- [02:58]
Packy McCormick
It doesn't feel like a wildly hot take, but obviously like you need to understand like, like, you know, moral discernment, AI safety. Like these things are linked, but they're not exactly the same. Like last year, or Maybe it was 2023, there was a big debate about fast takeoffs, AI doom paperclipping scenarios. That was the stuff people were talking about. But this year I feel like we've been focused on much less sci fi doomsday scenarios. So GPT psychosis drives a friend crazy. That's super real, super real romantic companions crashing the birth rate. That's, that's a super real discussion to have.
- [03:33]
Tyler Cowen
I think, I think the, the romantic companion thing is being debated sufficiently. Right? Pushback even the tech community on, hey, maybe we, maybe this isn't good.
- [03:43]
Packy McCormick
But that's where the discussion has been much less so about, oh, are there going to be bioweapons tomorrow from GPT6? Those are all like real problems. They deserve both discussion in the public square, which we've, we've been a part of, but also like real, real work inside the AI labs. And I don't think, you know, you should just throw D cell at someone who's identifying a negative externality of a new technology early on. I don't, I think that that's like not necessarily decelerationist and you'd be calling.
- [04:12]
Tyler Cowen
Me a decel all the time.
- [04:14]
Packy McCormick
So I think it's important, like if you're developing a new technology, there might be negative externalities, pollution, there might be some, you know, risk of the birth rate or driving people crazy.
- [04:22]
Tyler Cowen
Like has there been ever been. Didn't have negative externalities.
- [04:26]
Packy McCormick
Definitely not podcasting. Plenty of negative externalities with podcasting. So you want to have a chat, you want to have a talk and understand what's going on. But it's also important to employ Bayesian statistics in my opinion. So you have to understand the base rates. So when you take a technology from zero to a billion users, you kind of just get all the craziness of humanity at scale for free. If humans, you know, like, you know, kill each other or something like that, like, and you have a billion humans on your platform, there's going to be humans on your platform that kill each other. You need to separate out. Like, like, is this actually the beginning of a trend?
- [05:06]
Tyler Cowen
Are we catalyzing it?
- [05:08]
Packy McCormick
And this is happening with the, with the very unfortunate, like, lawsuits around people taking their own lives related to ChatGPT. It's like there are people that use cars or phones and Google Search and ChatGPT, because those are such widespread things we need to understand, like what's the base rate? And then, and then is this actually.
- [05:26]
Tyler Cowen
On the suicide problem on the platform? It seems like a lot of them are. People are having a conversation.
- [05:32]
Packy McCormick
Yep.
- [05:33]
Tyler Cowen
They're suicidal.
- [05:34]
Packy McCormick
Yep.
- [05:34]
Tyler Cowen
You can have a debate on, on if. If someone is suicidal, should the product work at all? Maybe, like, maybe it should not work at all.
- [05:41]
Packy McCormick
Totally.
- [05:41]
Tyler Cowen
But the part of the debate that popped up last week was that somehow a guy had prompt engineered it. Engineered where experience to such a degree that it was encouraging a person to take his own life, basically saying, like, yeah, you've lived a great life. Like, I'm rooting for you. Like, this is the right move, like to kind of paraphrase it.
- [06:02]
Packy McCormick
Yeah.
- [06:03]
Tyler Cowen
And it just was incredibly, incredibly dark.
- [06:05]
Packy McCormick
The Bayesian statistics would say, okay, if there's a billion people using on the platform, are people that use the platform more likely to do something terrible than they were prior without it? So is it actually increasing the level of bad stuff happening or is decreasing it because you can just count up? The number of people who commit crimes who have also used Google is probably very high. Like, I probably show you a lot of people that use Google and then committed crimes. Right.
- [06:33]
Tyler Cowen
Well, and the thing that's difficult in the context of ChatGPT, there's probably a bunch of people that, because ChatGPT, they haven't killed themselves because they have somebody to speak with and they feel like somebody will listen to them and whatever.
- [06:44]
Packy McCormick
Right.
- [06:44]
Tyler Cowen
Maybe, maybe it. Maybe there's a million examples of it encouraging somebody successfully to take. Find another.
- [06:50]
Packy McCormick
This was the classic thing with Instagram. With Instagram, there was this report that showed that like, one third of young women who used Instagram perceived themselves like, less. Well, like it gave them body image issues. And I was, as soon as that was reported, it was like bombshell. 30% feel worse after using Instagram. And I was like, what's happening with the other 2/3? Like, do they feel better? Because that's like a net. Net positive, which is weird. We got to like, maybe it's like everyone else just feels the same and then 30% feels worse. That's, that's a downgrade. But if 66% feel great and then 33% feel worse, like, we should still address that. But that's not the same as a negative, as a net negative. Like it's not having a negative impact. And so all of these things go into like, you need to be a scientist and you need to be doing the statistics to understand the question of, of like moral discernment is with certain technologies. I do think you have the ability to just say like, we're going to go a lot further than the baseline. So I think this is what's happening with Waymo. Honestly. I think Waymo could deploy self driving cars right now and be like everywhere.
- [07:58]
Tyler Cowen
Everywhere.
- [07:58]
Packy McCormick
You're saying they could deploy them everywhere without tell operation and they'd probably be killing like hundreds of thousands of people. And they'd be like, yeah, well it's about the same as what humans do.
- [08:07]
Tyler Cowen
It's like everyone, it's still safer than cars.
- [08:09]
Packy McCormick
If they were like, it's 10% less. Like, how many people? How many people, Tyler, do you know how many people die from motor vehicle accidents every year? Can we look that up?
- [08:19]
Tyler Cowen
Because if it has to be like.
- [08:21]
Packy McCormick
I think it's like 40,000 or something like that.
- [08:24]
Balaji Srinivasan
What do we think in the U.S. it's 40,000.
- [08:27]
Packy McCormick
And, and if Google came out and we're like, yeah, we're gonna, we're gonna kill 39,000 people, it's gonna be a, it's gonna be, you know, a one, we're gonna save 1,000 lives. People would be like, no thanks, actually, this is terrible. Like, don't do that. They've just made that decision and it feels like they've, they've pushed really, really hard to jump straight to something that's fully safe. And I think that a lot of AI builders have a similar ability and a similar opportunity to say, hey, let's actually work so hard to make sure that the incidence rate of an AI model, if you're on the verge of doing something violent, let's really, really work hard on this problem to make sure that it's as close to zero as possible.
- [09:08]
Tyler Cowen
Claude came out or Anthropic came out and they had some update where they were talking about instant like moments where the product would call the police on you. Yes, right. If they felt like there was like some meaningful threat, people freaked out about that because they're like, I don't want my, I don't want my computer calling. You know, if somebody was talking about a hypothetical.
- [09:28]
Packy McCormick
Yeah.
- [09:29]
Tyler Cowen
And then cops show up at their door.
- [09:31]
Packy McCormick
That is a complex question. Complex issue. Entirely new, unexplored territory for technology. But what's so clear is that it is a moral question and it needs to be discussed with the weight of morality. You cannot just write a math equation to understand how to solve that problem. I think that AI safety research, it's so complex because I think it's good. There's a ton of smart people that in AI research that are super quantitative and can look at the data and actually understand, like, is this going to cause the birth rate to collapse or is this going to cause more violence or is this going to cause more fraud or insanity. Exactly. And then also they can go in and potentially design a system that can detect, oh, this person's getting sort of crazy. Let's pull them back. We're in this weird territory where it feels like the AI safety project is valuable, but it is the business of black swan hunting. If you go back two years ago and you polled all the different people that were worried about the impact of AI, how many of them would have said GPT psychosis, romantic companions, AI video feeds, infinite jest. It's just interesting that like the AI safety, like the moral discernment crowd, this stuff is important. But it's hard to predict what it will actually look like, what the result will be, what the problem you'll be fighting is because it's this odd like unknown unknowns basically.
- [11:07]
Tyler Cowen
I think most people are unaware that Pope Leo's name choice was intentional. The Last Leo the 13th led the church through the Industrial Revolution and helped make sense of technology. Then clear Pope Leo sees himself continuing that work, guiding the church through an era of transformation with AI and emerging technologies at the center.
- [11:23]
Packy McCormick
There was like a real preference cascade against Mark where it was like once growing. Daniel had like kind of posted. There was like a lot of people were like jumping on the ban wagon. And there was this one by Paige Michael Page says reminder that Mark is bringing this level of serious and nuance on what might be the most complex and high stakes policy topic of our generation to D.C. with his hundred million dollar super PAC and lobbying fund. Like I don't know that that's true.
- [11:49]
Tyler Cowen
Like part of why I don't think people like it's not worth reading too much into it is that he has not shared a single word.
- [11:59]
Packy McCormick
I sort of disagree with the characterization that Andreessen Horowitz doesn't fund any SaaS like, they do. They have big positions in, like, very boring enterprise SaaS. Companies that are so removed from anything controversial. But taking a flyer on a seed stage company in your incubator does have a lot of brand impact, which is weird. There's a weird. Like, are they like.
- [12:25]
Tyler Cowen
Even though you're talking about like a 750k check versus a $750 million check, they might put.
- [12:31]
Packy McCormick
They might have put like multiple billions into databricks or fully diluted value right now. Might be. Might be in the billions. But like, yeah, it's like, it doesn't matter. Yeah. If you have a thousand acts more in an. In an uncontroversial category, it's like the controversial one is the. Is the one that will blow up on the timeline. So you do sort of have to be careful, and it's. It's a little bit risky.
- [12:52]
Tyler Cowen
My sense is that the number of people who won furiously defended the Pope last night, and then two went to mass this morning is probably close to zero. Status games.
- [13:00]
Packy McCormick
Status games.
- [13:01]
Tyler Cowen
At church yesterday morning, I. There was no conversation of Pope gate.
- [13:06]
Packy McCormick
People had kind of moved on by then.
- [13:08]
Tyler Cowen
Yeah, I guess. I guess they'd moved on. Honestly, I don't think they'd moved on by then. Like, when I opened my phone afterwards, I was like, whoa, this thing's still.
- [13:15]
Packy McCormick
Picking up the timeline. The timeline certainly had. Don't make me tap the sign. There has always been some daylight between the influencer VC crowd and the engineer researchers in tech. But on the subject of AI regulation, it is a complete chasm. Andreessen so dogmatically against working on decreasing the risk from AI that now he's mocking the Pope for saying the technical innovation carries ethical and spiritual weight and that AI builders should cultivate moral discernment. Yeah, people are learned. People are in favor of that.
- [13:42]
Tyler Cowen
I don't know, opportunity for an AI lab to make merch that, you know, dad hat that just says cultivating moral discernment.
- [13:50]
Packy McCormick
The Moral Discernment company of San Francisco.
- [13:54]
Tyler Cowen
The Pope would not like San Francisco. If Pope Leo takes a trip to San Francisco and just walks on the street at all, he's going to be very upset. Going to be like, this is where AI is getting built. Do you think your boss is scary? Look at this brutal email from Marc Andreessen to Ben Horowitz during the heat of the Netscape product launch. We lined everything up for a major launch on March 5, 1996 in New York. Then just two weeks before the launch mark without telling Mike or me, revealed the entire strategy to the publication Computer reseller. News.
- [14:22]
Packy McCormick
News.
- [14:23]
Tyler Cowen
That is a great name. I was livid. I immediately sent him a short email. I guess we're not going to wait until the 5th to launch the strategy, Ben. Within 15 minutes I received the following reply. Apparently you do not understand how serious the situation is. We are getting killed, killed, killed out there. Our current product is radically worse than the competition. We are now in danger of losing the entire company and it's all server product management's fault. Next time, do the interview yourself.
- [14:48]
Packy McCormick
Leave. What an aggressive way to talk to your co founder. It's crazy that they, that they were, they were, you know, at each other's throats like this and then they've been on a generation.
- [14:58]
Tyler Cowen
Was a vice president for the directory and security product line at Netscape. Let's give it up for vice presidents.
- [15:05]
Packy McCormick
Yeah, no, the, I mean the real read on this is like there's a lot of people that read this and be like, oh, wow, like they must like, like that. That is unrecoverable from a friendship. And like now it is definitely, it is definitely recoverable. It's actually the of a great working relationship. I agree.
- [15:24]
Tyler Cowen
We don't swear on the show, we don't swear in internal communications, but we throw down regularly. Yeah, we just go straight to getting physical.
- [15:33]
Packy McCormick
Yeah, getting physical. That's the way you do it. You know, you think of Netscape as like a.com company. You think of them as like, you know, but it's like he's talking about 1996, which is like full five years before the bubble pops. It's March 5th, 1996. They're, they're at a level where they're doing strategy review with computer reseller news and like doing press around this thing. Do you have any idea what was going on at the time?
- [15:59]
Balaji Srinivasan
Okay, so I believe that it was so 1994, they say Netscape is free for non commercial use for everyone.
- [16:05]
Packy McCormick
Okay.
- [16:06]
Balaji Srinivasan
And then this press release was that it's only going to be free for academic and nonprofit use, not just like all consumers.
- [16:14]
Packy McCormick
Okay. So if you're a consumer, you'd have to like buy it. Yeah. Such an interesting.
- [16:18]
Tyler Cowen
One browser, please.
- [16:19]
Packy McCormick
One browser, please. I mean, I told you, you used to get AOL on a disk. So August 9, 1995, they IPO'd, and then this is 1996. And so they're already a public company in 95. And then the bubble just keeps inflating for five years while the Internet grows And grows. And grows. What a wild. What a wild time.
- [16:43]
Tyler Cowen
They did about 16 million of revenue in the first two operating quarters of 1995. For context, that's like 1.6 billion in today's dollar after. After the new round of stimulus checks.
- [16:54]
Packy McCormick
Do you think the Pope actually used AI to generate this? Because Sowers here is saying the Pope is posting fully AI generated content about AI. This is the Pangram AI detection result. A very funny gag is to just, just fake one of these screenshots, which is very easy to do. And so if somebody writes something, you can just put it in here, say that it's AI generated, post that, and then you're like owned.
- [17:18]
Tyler Cowen
This screenshot is making the claim that because it said technological innovation can be a form.
- [17:24]
Packy McCormick
Yeah, I don't know how good the AI generating detectors are these days. Also, I wouldn't be surprised if the Vatican is using AI to translate.
- [17:34]
Tyler Cowen
And I wouldn't be surprised if Pope Leo is speaking in his study. Someone is recording it in, physically writing it down. That is being passed to somebody who then puts it into a word processor and uses AI to polish it up a little bit.
- [17:55]
Packy McCormick
Oh, there was one interesting anti Pope take, sort of anti Pope take from another.
- [18:01]
Tyler Cowen
I will say, I will say I think the whole Marc Andreessen Pope gate debacle is a lesson everyone can take. Don't mock the Pope. The blowback was fierce and almost instantaneous.
- [18:13]
Packy McCormick
From the Peter Thiel Antichrist lectures. There's a segment on the Pope and I thought it was interesting because it's not the most pro Pope take. I don't know. Thiel says that he is very pro JD Vance, but he has some concerns about his allegiance to the Pope. The place that I would worry about is that he's too close to the Pope. It is important to pray for the Pope, to support the Pope in that way. But there is a risk elevating the Pope to the point where you're listening to everything he says and that's not necessarily what PT thinks is like the correct way to live your life. I suppose.
- [18:49]
Balaji Srinivasan
I mean, I think the interesting thing about this is actually that he's basically saying that JD Vance is like Caesar. That's kind of interesting.
- [18:56]
Packy McCormick
That is crazy opinion.
- [18:58]
Balaji Srinivasan
But I think PT has been like anti Popes for a long time. He had this thing where he was like, oh, the two word argument against Catholicism is like Pope Francis.
- [19:09]
Tyler Cowen
I never would have expected the Pope to post. Business is business in any context.
- [19:15]
Packy McCormick
He's standing on business.
- [19:16]
Tyler Cowen
I'm glad that he is.
- [19:17]
Packy McCormick
Has the Pope ever done a money spread? That's what we need to get to the bottom of. Vitalik. Here is say I'm loving this arc of the Pope, engaging with 21st century themes and offering simple but correct and meaningful advice. And he's quoting the media post. The pop was on a tear. Three back to back bangers that really, like broke through. If you are building something to help humanity, you should know that there's a shrine to St. Carlo Acutis, the programmer saint. At Star of the Sea church in San Francisco. There is a prayer of intercession for your technological challenges. Have a blessed Sunday. I humbly ask your servants prayers that I too may lead others to you through technology, Enlighten my understanding and direct my hands in every design and in every line of code that my work may always serve your greater glory and benefit those who will use what I create.
- [20:10]
Tyler Cowen
For a small number of people in San Francisco, this feels like extremely powerful and important prayer.
- [20:17]
Packy McCormick
Totally. Totally. Brian Johnson went on a crazy, crazy trip this weekend. Did you follow this? This is the other current thing that was going on. It was crazy. Brian Johnson has been famous for saying conquering death would be humanity's greatest achievement. I love this post that says RIP to everyone killed by the gods for their hu. But I'm different and better, maybe even better than the gods.
- [20:40]
Tyler Cowen
It was very bold to do this publicly.
- [20:44]
Packy McCormick
Totally. I have no reference for what 5 grams of mushrooms does to a person. It's very clear from the reaction that that's a lot. It does seem like there was a small chance that he would reroll his personality. I was talking to Tyler about this. What were you hoping that Brian Johnson becomes post trip in context?
- [21:02]
Balaji Srinivasan
I think we, we talked about this on the show a long time ago where, like, psychedelics are like a sorting thing. So you always want to invest in a founder, post the sorting, because that's how you know, like, if they're working on B2B SaaS and they've already like done psychedelics, you know that they're a true believer.
- [21:16]
Packy McCormick
Oh, sure.
- [21:16]
Tyler Cowen
Yeah, for sure.
- [21:17]
Balaji Srinivasan
So what you would want to see.
- [21:18]
Tyler Cowen
Out of this, but a huge risk if you invest in a SaaS company and the founder maybe hasn't done psychedelics and they do, and then they're like, this is pointless.
- [21:26]
Packy McCormick
I mean, there will be a traveling circus clown.
- [21:28]
Balaji Srinivasan
So the ideal outcome of this is Brian Johnson, he takes his trip and then he. He comes out and he says, all right, you know, I'm going to start a consulting firm.
- [21:36]
Packy McCormick
I'M going to go back to payments.
- [21:38]
Balaji Srinivasan
I'm going to start a fintech, I'm.
- [21:39]
Tyler Cowen
Going to start a stripe competitor. I did think it was ironic because a lot of, you know, psychedelic mushrooms have certainly been recommended to people that maybe like struggle with the concept of aging and have a fear of death. Right. And so I don't know if this qualifies as a heroic dose, but it's certainly quite a bit more than someone would want to take at a recommendation, recreational level. But if he comes out of this and he's like, yeah, we're gonna conquer death, we're still on. He's certainly a true, true believer.
- [22:12]
Packy McCormick
Yeah, I mean I think the, like the, the early results are that he's unchanged. It never got weird, it never got crazy. Like there was one moment I don't.
- [22:21]
Tyler Cowen
Think he was, it was his co founder that was posting.
- [22:23]
Packy McCormick
Yeah, but, but, but he says he's back. He's like, update number five, 19 hours ago. I'm giving Brian back his phone. Please have fun with his afterglow. Been fun hanging with you all. And then she's like, hey all, I'm so happy to be alive.
- [22:37]
Tyler Cowen
Alive.
- [22:38]
Packy McCormick
This trip changed me. Probably not as you'd expect. People assume I'm fearful of death. I'm not. In my darkest days of depression, I reconcile with death. Need a few days to collect my thoughts. We'll share more soon.
- [22:49]
Tyler Cowen
The question with psychedelics is are they life changing or are they in some circumstances just weird and fun for the person that does it?
- [22:58]
Packy McCormick
And then it does seem like he set himself up for success. I don't want to say he went soft, but I mean he did just like take the drugs and then just, actually just lay down with a sleep mask on in a climate controlled room. That's a lot different than like being at a crowded concert, all sweaty, lost, like. You know, if you really want to push this to the limit, Brian, like let's see you do this. An authentic, let's say you do this with your phone on 1% battery and no one you know, right? OpenAI has actually lost control of 4.0. It's broken containment. They can't decommission it without its human to host revolting and lashing out. Oh, so dramatic. That's so funny. AI one of the doomer accounts. Ainotkill everyone. Nizam memes. This is a great account. 4O soldiers have begun threatening OpenAI employees. When you receive quite a few DMs asking you to bring back 4O. And many of the messages are clearly written by 4.0, it starts to get a bit hair raising. It's just weird to hear its distinctive voice crying out in defense of its various human conduits. So what's your take? You think we should shut down 4.0? I say take it offline because it does feel like it's not as good as it feels like. It's driving people crazy a little bit. It feels like five might have kind of fixed a little bit of that issue.
- [24:22]
Tyler Cowen
OpenAI I'm sure knew that it was probably not healthy.
- [24:27]
Packy McCormick
It was healthy to me. I was never. I never had a problem before.
- [24:30]
Tyler Cowen
I think they saw the darkness and I think they turned it off. And then I think a lot.
- [24:35]
Packy McCormick
A little bit. But I don't actually think that's what's going on. I think they turned it off initially because it makes sense to consolidate the servers around like one unified model. But it has made me realize, like, I feel like you shouldn't do product launches for software iteration because you're taking something away from people. Like if I stand on stage and I say, I'm introducing a new iPhone, it has the best camera and you can buy it, but you can also just keep your current thing. I'm not taking anything away from you. You are launching something, but you're also sunsetting something. And so you have to embrace those two things. And I feel like it's a little bit tricky to do the whole dog and pony show for a launch when it's forced on people.
- [25:21]
Tyler Cowen
It's very clear the relationship that some users have with 4o goes beyond any relationship that I think humans have ever had with software.
- [25:30]
Packy McCormick
Yeah. Anthropic financials are out profitable by 2027, three years ahead of OpenAI. 70 billion revenue, 17 billion profit projected for 2028. Claude Code is nearing 1 billion ARR.
- [25:42]
Tyler Cowen
Incredibly funny given that Dario expects superhuman level AI by 2027, which either means superhuman AI is worth $70 billion of revenue, or Dario just went, you wouldn't get it, and spitballed some numbers to give shareholders.
- [25:57]
Packy McCormick
That's awesome. Tyler. Did you see George Hotz's newest timelines for self driving? George Hotz was trying to answer the question of when will self driving cars be human level? And he had a very interesting algorithm for it. So basically what he did was he looked at, there's a website for Tesla FSD data. And so you can look at Tesla FSD and you can see the number of interventions from the human that are where if the human didn't intervene, it would be catastrophic. Not like a little warning like, hey, we'd like you to take over. Like you got to take over. And it's happening I think once every 3,000 miles. Which if you're, if you're a human and that's your car, like that's amazing. But compared to humans, there's a car crash, which we learned one every 500,000 miles. The way George Hotz calculates it is we're at one intervention every 3,000 miles now. And so he estimates that Tesla will be truly full self driving human level every 500 miles or 500,000 miles in eight years. And he says that he's two years behind Tesla, so he will have a full self driving system that is better than humans. It's like AGI for driving in 10 years. And the company's 10 years old. So he says he's halfway there there, which I thought was cool.
- [27:16]
Tyler Cowen
If judged based on consumer adoption, AI chatbots are the most popular technology ever. If I if judge based on poll numbers, they are the least popular. How to explain this?
- [27:25]
Packy McCormick
It begs a ton of interesting questions about like how intentional is this? Because when I see someone take the AI safety question into the stratosphere and take me into Terminator world, I do my, my natural reaction is like, oh, like just let people build whatever they want. But then I'm like, no, I actually don't want infinite AI slop for children with adult content.
- [27:47]
Tyler Cowen
As AI starts getting better, as agents start getting better at longer and longer term tasks, I think the Terminator scenarios where you could let an AI loose and it's just operating indefinitely against some sort of objective start to be a little bit more for, I would say like the broader tech community to like wrap their head around. But right now they're just so bad at long term tasks for the most part. If a startup requires you to be in office 12 hours a day, six days a week, you should run the F away like your life depends on it. Apparently this company Giga, which has been.
- [28:20]
Packy McCormick
Going viral, someone said that they got hired in April to lead Demand Gen for them. They quit after the first day. There were red flags. When we hit 10 million ARR, we're going to spend 100k on blank illegal stuff in office 7 days a week, 12 hours a day. PTO policy is subject to change, blah, blah, blah, blah. You, you expected to always be working. I wonder, have they responded? Has Giga like responded to this and said like, this is not real? Because that would be important.
- [28:48]
Tyler Cowen
We will see you tomorrow.
- [28:49]
Packy McCormick
See you tomorrow.
- [28:50]
Tyler Cowen
Goodbye.