All-In PodcastE99: Cheating scandals, Twitter updates, rapid AI advancements, Biden's pardon, Section 230 & more
EVERY SPOKEN WORD
150 min read · 30,166 words- 0:00 – 1:34
Bestie intros!
- DSDavid Sacks
We're seven minutes in and we've produced absolutely nothing that will go in the show.
- JCJason Calacanis
(laughs)
- DFDavid Friedberg
(laughs)
- CPChamath Palihapitiya
Here comes Sax, waking up with his-
- JCJason Calacanis
Oh, boy. Here he comes.
- DFDavid Friedberg
... with his commentary.
- JCJason Calacanis
When Friedberg is criticizing you for being too negative, you're in a dark place, Sax. (laughs)
- CPChamath Palihapitiya
I'm actually angry at Sax for not publishing my AMA from the other night while he publishes-
- DSDavid Sacks
It's coming, it's coming.
- CPChamath Palihapitiya
... while he publishes Neval's content.
- DFDavid Friedberg
His app crashed.
- DSDavid Sacks
We had such a crowded room. We had over 2,000 people-
- JCJason Calacanis
That was impressive.
- DSDavid Sacks
... in the room for, like, four hours.
- JCJason Calacanis
It was crazy. It was like the original days of Clubhouse.
- CPChamath Palihapitiya
Everyone I know that was trying to get in was texting saying they couldn't get in. So it definitely capped out, right?
- DSDavid Sacks
I know. Well, we hit, we hit some scalability limits.
- JCJason Calacanis
You may want to buy an extra server, Sax. Cheap (beep) . Weren't you the same guy who was responsible for scaling PayPal?
- CPChamath Palihapitiya
No, that was somebody else. That was eBay. They sold it before it scaled.
- DSDavid Sacks
No, no, that's not true. We, we had huge scalability challenges at PayPal too.
- CPChamath Palihapitiya
(laughs)
- JCJason Calacanis
It seems like a theme.
- DSDavid Sacks
Yeah. The theme is when you have an app that's breaking out, you hit scalability challenges. It's called a high-class problem.
- JCJason Calacanis
2,000 people is not a high-class problem. It's a trickle. It's 2022.
- DSDavid Sacks
2,000 people participating in the conversation is a challenge.
- JCJason Calacanis
I haven't written code in 20 years. Here's what you do. When you get to 1,000 people coming to the room-
- CPChamath Palihapitiya
That's a lie, that's a lie.
- JCJason Calacanis
... everybody else is in passive ro- mode.
- CPChamath Palihapitiya
You've never written code ever.
- JCJason Calacanis
Of course, I have. Of course, I have.
- 1:34 – 16:13
Breaking down major cheating scandals: chess, poker, fishing
- NANarrator
in.
- JCJason Calacanis
So there have been three (laughs) cheating scandals across poker, chess, and even competitive, uh, fishing. I don't know if you guys saw the fishing one, but they found weights and filets during a fish, uh, weigh-in. And then everybody wants us to check in on the chess and the poker scandals. Chess.com just released their report that this, uh, grandmaster has been suspended. They have evidence he cheated basically in a bunch of, uh, tournaments that were, in fact, uh, for money. He denied that he had done that, but he had previously cheated as a kid. Uh, they now have the statistical proof that he was playing essentially perfect chess. And, uh, they've outlined this in, like, hundreds of pages in a report. Sax, you, what, what are your thoughts on this, uh, scandal in chess?
- DSDavid Sacks
Magnus Carlsen finally came out and explained why he thought, um, Hans Niemann was cheating. Basically, he got the strong perception during the game that Hans wasn't really putting in a lot of effort, that he wasn't under a lot of stress, and he's, he... It's his experience that when he's playing, you know, the top players, they're intensely concentrating, and Hans Niemann didn't seem to be exerting himself at all. So his, you know, hackles were raised and got suspicious and then, you know, he has had this meteoric rise, the fastest rise in classical chess rating ever. And I guess there were... He had gotten suspended from Chess.com in the past for cheating. So on this basis and maybe other things that Magnus isn't telling us, Magnus basically said that this guy is cheating. I think that maybe the interesting part of this is that there's been a lot of analysis now of Hans Niemann's games, and I just think the methodology is kind of interesting. So what they do is they run all of his games, uh, through a computer and they compare his moves to the best computer move, and they basically assign a percentage that Hans-
- CPChamath Palihapitiya
A correlation.
- DSDavid Sacks
... matches, a correlation, matches the computer move. And what they found is, there were a handful of games where it was literally 100%. That's basically impossible without cheating. I mean, you look at the top players who, through an entire career, have never had a 100% game. The, the... You know, chess is so subtle that the computer can now see so many moves into the future that nailing the best move every single time for 40, 50, 100 moves is just-
- CPChamath Palihapitiya
No, and, and, and in chess, which a human really can't do that well, is that there are positional sacrifices that you will make in short lines that pay off much-
- DSDavid Sacks
Right.
- CPChamath Palihapitiya
... much later in the future, which is impossible for a human to calculate. And so-
- DSDavid Sacks
Right.
- CPChamath Palihapitiya
... you know... And, and you saw this, by the way, when, uh... I think it was the, it was the Google AI, the DeepMind AI that-
- DSDavid Sacks
Yeah.
- CPChamath Palihapitiya
... that also played chess. So the idea that this guy could pay, could play absolutely perfectly according to those lines is only-
- DSDavid Sacks
Right.
- CPChamath Palihapitiya
... possible if you're cheating.
- DSDavid Sacks
Right. Exact- So, so there were a handful of games at 100% and then there were tournaments where his percentages were in the 70-something plus. And so just to give you some base for comparison, Bobby Fischer during his legendary 20-game winning streak, uh, was at 72%. So he only matched the computer sev- for best move 72% of the time. Magnus Carlsen playing at his best is 70%. Uh, Garry Kasparov in his career was 69%, and then the, you know, the, the super GMs category are typically in the 64 to 68% range. So I think it's really interesting actually how you can now quantify by comparing the human move to the best computer move-
- JCJason Calacanis
And it's multiple computers, right, Sax?
- DSDavid Sacks
... whether who's playing the best.
- JCJason Calacanis
They actually have...
- DSDavid Sacks
It provides a way to assess who the greatest player ever is. Um, I actually thought that it was Magnus, but now maybe there's a basis for believing it was Bobby Fischer because he was at 72 and Magnus was only at 70. However, look, the idea that Hans Niemann is in the 70s, 80s, or 90s during tournaments would be, you know, just an-
- JCJason Calacanis
Preposterous.
- DSDavid Sacks
... off-the-charts level of play. And if, if he's not cheating, then we should expect over the next couple of years that he should rapidly become the world's number one player over the board, uh, you know, now that they have all this anti-cheating stuff, right? So it'll be interesting to see what happens in his career now that they've really cracked down on, you know, on... with, with anti, anti-cheat technology.
- CPChamath Palihapitiya
I have a general observation, which is these people are complete fucking losers. The people that cheat in any of these games don't understand this basic simple idea, which is that trying is a huge part of the human experience. The whole point is to be out there in the field of play trying. And it's basically taking the wins and the losses and getting better that is the path. That's what's fun.
- JCJason Calacanis
Yeah.
- CPChamath Palihapitiya
... once you actually win, it's actually not that much fun because then you have this pressure of maintaining excellence. That's a lot less enjoyable than the path to getting there. And so, the fact that these people don't understand that makes them slightly broken, in my opinion. And then, the other thing is, like, why is it that we have this strain of people now that are just so devoid of any personal responsibility that they'll just so brazenly take advantage of this stuff? It's really ridiculous to me. They don't... It's really sad. These people are pathetic. They're losers.
- JCJason Calacanis
It's really pathetic. Uh, this, uh... But it- it is really interesting how they caught him, uh, and running this against a computer. Here's a- a chart of, um, his scores in these tournaments. Oh, well, here is, this first chart is how quickly he advanced, which was off the charts, and then the second chart that's really interesting is, um, his Chess.com strength. So if you don't know Chess.com, it has become like a juggernaut in the chess world, especially after that HBO series came out. A lot of people subscribe to it. I subscribed to it. I like to play chess there. And man, you look at the chess strength score there, he was just like perfect. And then the number of games he likely cheated in, you can see the- the last two columns, he's basically cheating in every game. Uh, yeah, Queen's Gambit, yeah, great show on, uh, Netflix. And, uh, he, he said he didn't cheat in any, uh, of the, uh, games where they were live streaming, but they've proven that wrong. Sacks, how does he cheat in person then?
- DSDavid Sacks
Well, that's the thing, no one really knows and I don't wanna overly judge it until they have hard proof that he was cheating. I mean, look, here's the thing, he was never caught in the act. It's just that the computer evidence, you know, seems pretty damning.
- JCJason Calacanis
Yeah, I mean, they have deterrants.
- DSDavid Sacks
And I don't... Here's the thing, is I don't know how they prove, I don't know how they proved that he was cheating over the board without actually catching him doing it. And I don't, I still don't think anyone really has a good theory in terms of how he was able to do that.
- CPChamath Palihapitiya
Well, it's not just him though. Like, look, the fishing thing, Jason, which was crazy, I think Friedberg showed the video. This guy was in a fishing competition and they basically caught these fish and then they put these big weighted pellets inside the fish's body. They even put like a, you know, chicken breast and chicken filets inside of the- the thing so it would weigh more.
- JCJason Calacanis
Oh, it was fish filets. (laughs) Yeah, they put fish filets in.
- 16:13 – 30:05
Twitter deal updates
- CPChamath Palihapitiya
- DFDavid Friedberg
Earlier this week, it was reported that Elon, uh, contacted Twitter's board and suggested that they move forward with closing the transaction at the original terms and the original purchase price of $54.20 a share. In the couple of days since then, and even as of right now with some news reports coming out here on Thursday morning, um, it's, it appears that there are still some question marks around whether or not the deal is actually gonna move forward at 54.20 a share, because Elon, as of right now, the report said, is still asking for a financing contingency in order to close, and there's a lot of back and forth on what the terms are. Meanwhile, the court case in Delaware is continuing forward on whether or not Elon, uh, breached his terms of the original agreement to close and buy Twitter at 54.20. As we know, leading up to the signed deal or, or, uh, post signing the deal, Elon put together a financing syndicate, a combination of debt investors as well as equity co-investors with him to do the purchase of Twitter at $54.20 a share. So the 40 some odd billion dollars of capital that's needed was committed by a set of investors that were gonna invest debt and equity. And there's a big question mark now on whether or not those investors want to or would still consummate the transaction with Elon, given how the markets have turned and given how debt markets are trading and equity markets are trading. So Chamath, I'd love to hear your point of view on what, um, hurdles does Elon still have in front of him? Does he still want to get this done? And is there still a financing syndicate that's standing behind him at the original purchase price to get it done?
- CPChamath Palihapitiya
It's a great question. Um, maybe the best way to start is, Nick, do you want to cue up what I said in August 25th? The lawsuit really boils down to one very specific clause, which is the pinnacle question at hand, which is, there is a specific performance clause that Elon signed up to.Right? Which, you know, his lawyers could have struck out and either chose not to or, you know, couldn't get the deal done without. And that specific performance clause says that Twitter (drumming) can force him to close at 54.20 a share. And I think the, the issue at hand at the Delaware business court is going to be that, because Twitter is gonna point to all of these, you know, gotchas and disclaimers that they have around this bot issue as their cover story. And I think that really, you know, this kind of, again, builds more and more momentum in my mind that the most likely outcome here is a settlement where you have to pay the economic difference between where the stock is now and 54.20, which is more than a billion dollars, or you close at some number below $54.20 a share. And I think that that is like, you know, if you had to be a betting person, that's probably ... And if you look at the, the way the stock has traded and if you also look at the way the options market trades, that's what people are assuming, that there's a $7 to $10 billion swing. And if you impute that into the stock price, you kind of get into the $51 a share kind of, uh, an acquisition price. Again, I'm not saying that that is right or should be right, uh, that's just sort of what the market says. (whooshing) Yeah. So, so I, I, it turns out that, you know, sort of like that kind of guesstimate turned out to be pretty accurate because the stock today is at $51 a share. So I think that the spec- specific performance thing is exactly what this thing has always hinged on, and I think that there is a realization that there are very few outs around how that contractual term was written and agreed to. So there is an out in the contract, and that out says that, I think, it's by April. If, um, if the deal doesn't get done by April, then the banks can walk away from their commitment to fund the debt. And if the banks walk away, then Elon does have a financing contingency that allows him to walk away. So the actual set of events that have to happen is those two things specifically. Get to April so the banks can pass and say, "We've changed our mind. Market conditions are different," and then Elon is able to say, "Oh, you know, the banks just walked away." Right now, the banks, if you look at the, all of the debt that they've committed to, well, they committed at a point in time when the debt markets were much better than they are today. In the last, you know, six or seven months since they agreed to do this, the debt markets have been clobbered, and specifically junk bonds and a bunch of junk bond debt, the yields that you have to pay, so the price to get that kind of debt, has skyrocketed. So roughly, back of the envelope math would tell me that right now, the banks are offside between $1 and $2 billion because they're not gonna be able to sell this debt to anybody else. So I think the banks obviously want a way out. Um, the problem is, their only way out is to run the shot clock off until April. So I think that's the dance that they're in right now. Elon's trying to find a way to solve, you know, for the merger. I think Twitter is gonna say, "We're not gonna give you a financing contingency. You have to bring the banks in and close right now, and then we will not go to court. Otherwise, we're going to court." And so I think it's a very delicate predicament that they're all in, but my estimate is that the equity is probably 20% offside. So it's not a huge thing. He can make that up because he can create equity value like nobody's business. The debt is way offside by a couple billion dollars, which is hard to make back, but I think in the end, you know, given enough time, they can probably make that back. The best off in all of this are the Twitter shareholders. They're getting an enormous premium to what that company is worth today in the open market. Um, and so I think this deal is gonna close, and it's probably gonna close in the next few weeks. And had you bought Twitter when we were talking about it in August, you would have made 25% in six weeks and, you know, if the deal closed at 54, you, you would have made, you know, a third of your money in eight weeks, which is, you know, very hard to do in a market like this.
- DFDavid Friedberg
If you're a GP at one of the funds, like Andreessen or Sequoia, and you had made this commitment to Elon or even Larry Ellison couple months ago, do you fight against closing at 54.20? Do you stick with the deal and support him? I mean, what do you do given that the premium is so much higher than where the market would trade it at today? Some people are saying the stock should be at, like, 20 bucks a share or something.
- CPChamath Palihapitiya
The average premium in an M&A transaction in the public markets is about 30%. So, um, and I think the fair value of Twitter is around 32 to 35 bucks a share. So, you know, it's not like he is massively, massively overpaying. And so, you know, I would just sort of keep that in the realm of the possible. So, like, if you take $35 as the midpoint, fair value is really 45, 50. So yeah, he paid 20% more than he should have, but he didn't pay 100% more. So it's not as if you can't make that equity back as a private company, particularly because there's probably $10 of fat in the stock if you think about just OpEx-
- DFDavid Friedberg
Right.
- CPChamath Palihapitiya
... in terms of all the buildings they have. Maybe they don't need as many employees. Maybe they revisit salaries. You know, one thing is, when I looked at doing an activist play at Twitter, I think I mentioned this five or six years ago, one of the things that I found was at that time, Twitter was running their own data centers. And, you know, the most obvious thing for me at that time was like, we're gonna move everything to AWS. Now, I don't know if that happened, but I'm sure that if it hasn't, just bidding that out to Azure, GCP, and AWS can raise, you know, three or four billion dollars because I'm sure those companies would want this kind of an app on their cloud. So there's all kinds of things that I think Elon can do as a private company to make back maybe the small bit that he overpaid, and then he can get to the core job of rebuilding this company to be usable, uh, this product to be usable, bec- I, look, I'll just speak as a user right now. It has been decaying at a very, very rapid clip, and I think that his trepidation in closing the merger...... in part also, even though he hasn't said it, has to do with the quality of the experience. It's just degraded. It's not as fun to use as it was during the pandemic, um, or even before the pandemic. So something is happening inside that app that needs to get fixed. And if he does it, he'll make a ton of money.
- DFDavid Friedberg
Sort of like what happened with Friendster and MySpace and any social networking app over time, the quality degrades. If it's not growing, it's shrinking and it gets, uh-
- CPChamath Palihapitiya
If it, if it's, if it's not growing and also if the product hygiene isn't enforced in code, and product hygiene in this case are the sp- you know, spam, bots, you know, the trolling, it can really take away from the experience.
- DFDavid Friedberg
Yeah. I mean, interestingly, like, if you think back to the origi- the f- the, uh, the starting days, the original days of Twitter, I don't know if you guys remember, you would s- send in an SMS to do your tweet and then it would post up and other people would get the SMS notification. And, um, it would crash all the time and the apps were, were... The app was notoriously crashing. It was, uh, poorly architected at the beginning. And some people have argued that Twitter has had a cultural technical incompetence from the earliest days.
- CPChamath Palihapitiya
I think that's a little harsh. So I do th- look, Twitter was known for what's called the fail whale. You know, they used to have these fail whales constantly, and they did hire people that attempted to try to fix it. I remember the, the, the funniest part of when I went in there and said, "Hey, here's my plan and here's what I want to do," is literally a day or two later (laughs) the head of engineering quit. I can't remember who his name was, but he was just out the door. Uh, but, you know, it is a... I think it is a team that has tried its best that probably at the edges definitely made some technical miscalculations, like I said, at that time. The idea that any app of that scale would use their own data centers made no technical sense whatsoever. It made the app laggy, it made it hard to use, it made it more prone to downtime, to your point. But that being said, I would be shocked if they haven't made meaningful improvements because the stack of the internet has gotten so much better over the last seven years. And so, to your point, David, if they didn't take advantage of all these new abstractions and mechanims to, mechanisms to rebuild the app or to rebuild search or to rebuild, you know, how, y- you know, all these infrastructure elements of the app work, I would be really surprised because then what are they doing over there?
- DFDavid Friedberg
Yeah. Well, look, I mean, to the point earlier, besides the product points, there was a, a really good, a tweet (laughs) I liked that said, "For what it's worth, I think Elon will show us just how lean the Silicon Valley advertising companies can be run. At the very least, it'll be an interesting thought experiment for spectators." Um, because if he does go in and actually does significantly reduce OpEx and headcount and the company does turn profitable and he can grow it-
- JCJason Calacanis
Well, look, for, for-
- DFDavid Friedberg
... uh, it'll, it'll, it'll really... By the way, it'll really be a beacon for some of these other big companies. Yeah.
- CPChamath Palihapitiya
From a financial, from a financial perspective, there is $10 a share in OpEx cuts that he should make right away just so that he is economically breakeven and he looks like every other M&A transaction. You know, you paid a 30% premium and you bought a company. There's a lot of margin of safety there if Elon does that. So to your point, there probably is and there probably needs to be a meaningful riff at Twitter. I'm not saying it's right, I'm not saying it's... You know, and I feel for the people that may go through it. But from a financial perspective, the math makes sense for him to do that because then he is a breakeven proposition on a go-in M&A transaction. And I think that there's, there's a lot of intelligent financial sense so that all the debt holders feel like he's doing the right thing and all the equity holders particularly see a chance for them to make a decent return here.
- DFDavid Friedberg
All right. Well, let's move on. This has been a great conversation-
- JCJason Calacanis
JCal back to you, Big Boy. (laughs)
- DFDavid Friedberg
Yeah, great conversation between Chamath Palihapitiya and Dave Friedberg, uh, about the Twitter transaction, and now we're being rejoined by our besties who are, uh-
- JCJason Calacanis
(laughs) Yeah, by other besties.
- DFDavid Friedberg
... back from their cappuccino break.
- CPChamath Palihapitiya
(laughs)
- DFDavid Friedberg
Yeah. How was your cappuccino, JCal?
- JCJason Calacanis
Oh, it was great. I have, uh, I have, uh, I have a nice cold brew here, a nice iced cold brew and a, a nice, uh, drip coffee. I'm, I'm working both.
I'd love to talk about topics that I'm not, uh, being-
- DFDavid Friedberg
Yeah.
- JCJason Calacanis
... subpoenaed or depositioned about.
(laughs) We will have a lot to say in the coming weeks.
I love to talk about topics that my lawyers have advised me not to talk about.
- DFDavid Friedberg
(laughs)
- CPChamath Palihapitiya
That's brutal.
- 30:05 – 49:45
AI making rapid advancements: Tesla AI Day, Meta's text-to-video tool, where this all leads
- JCJason Calacanis
all good. So anyway, the event was, um, uh, super impressive. Elon only spoke when he showed the, uh, Optimus, the new robot. He's building a, a general purpose robot that will work in the factories. It's very early days, uh, but they showed two versions of it and, um, he said he thinks they could get it down to $20,000. It's gonna work in the factory, so it's actually got a purpose.And obviously, the factories already have a ton of robots, but this is more of a robot that, uh, will benefit from the general, uh, or the computer vision and the AI, the narrow AI being pursued by the self-driving team. This is like two and a half hours of really intense presentations. Um, the most interesting part for me was, uh, they're building their own supercomputer, uh, and their chips. And the Dojo supercomputer was really impressive, um, at how much they can get through a scenario. So they're building every scenario, every self-driving. I actually have the full self-driving beta on my car. I've been using it. It's pretty impressive, I have to say. If you haven't used it yet, I feel like AI is moving at a, a pretty advanced clip the past year. If you haven't also seen, Meta announced a text-to-video generator. So this is even more impressive than DALL-E. DALL-E, you put in a couple of words, Freeberg, and, and you get a painting or whatever. This is put in a couple of words and you get a short video. So they had one of a teddy bear painting a teddy bear. So it looks like you're going to be able to essentially create a whole movie by just talking to a computer. Really impressive. Where do you think we are, Freeberg, in terms of the compounding nature of these narrow AI efforts? You know, obviously we saw poker, chess, Go, DALL-E, GPT-3, self-driving. It feels like this is all compounding at a faster rate, or am I just imagining that?
- DFDavid Friedberg
Yeah, look, I mean, it's interesting when, when people saw the first computer playing chess, they said the same thing. I think any time that you see progress with a computer that starts to mimic the predictive capabilities of a human, it, it, it's, uh, it's impressive. But I will argue, and I just... I'll say a few words on this. It's... I, I think this is part of a, a 60-year cycle that we've been going through. Um, fundamentally, what humans and human brains do is we can sense our external environment, then we generate knowledge from that sensing, and then our brains build a model that predicts an outcome, and then that, that predicted outcome is what drives our actions and our behavior. We observe the sunrise every morning, then we observe that it sets, and you see that enough times, and you build a predictive model from that data that's been generated in your brain, that I predict that the sun has risen, it will therefore set. It has set, it will therefore rise. And I think that the computing approach is very similar. It's all about sensing or generating data and then creating a predictive model, and then you can drive action. And initially, the first approach was just basic algorithms, and these are deterministic models that are built. It's a piece of code that says, "Here's an input, here's an output," and that, that model is really built by a human. And a human designed it, designed that algorithmic model and said, "This is what the, uh, the predictive, uh, potential o- of this software is." Then there was this term called data science. So as data generation began to proliferate, meaning there was far more sensors in the world, it was really cheap to, to, to create digital data from the physical world, really cheap to transmit it, really cheap to store it, and really cheap to compute with it, data science became the hot term in Silicon Valley for a while. And these models were not just a bis- basic algorithm written, um, by a human, but it became an algorithm that was a similar deterministic model that had parameters, and the parameters were, uh, ultimately resolved by the data that was being generated. And so these models became much more complex and much more predictive, finer granularities, uh, finer range. Then we used this term machine learning, and, and in the data science era, it was still like, Hey, there's a model, and you would solve it statically. You would get a bunch of data, you would statically solve for the parameters, and that would be your model, and it would run. Machine learning then allowed those parameters to become dynamic. So the model was static, but generally speaking, the parameters that drove the model became dynamic as more data came into the system, and they were dynamically updated. And then this era of AI became... and that's the new catchword. And what AI is realizing is that the- there's so much data that rather than just resolve the parameters of the model, you can actually resolve a model itself. The algorithm can be written by the data. The algorithm can be written by the software. And so with A- with AI-
- JCJason Calacanis
Example.
- DFDavid Friedberg
So poker playing, an adaptive model. So people... Uh, so you're playing poker, and the software begins to recognize, uh, behavior, and it builds a predictive model that says, "Here's how you're playing." And then over time, it actually changes not just the parameters of the model, but the model itself, the algorithm itself. And so AI... And then it eventually gets to a point where the algorithm is so much more complex that a human would have never written it. And suddenly, the AI has built its own intelligence, its own ability to be predictive in a way that a human algorithmic programmer would have never done. And, and this is all driven by statistics, so none of this is new science per se. There's new techniques that all, on their underlying, use statistics as their basis, and then there's these techniques that allow us to build these new systems of model development, like neural nets and so on, and those statistics build those neural nets, they, they, they solve those parameters and so on. But fundamentally, there is an, um, geometric increase in data and a geometric decline in the cost to generate data from sensors, because the cost of sensors is coming down with Moore's Law, transmit that data, because the cost of moving data has come down with broadband communications, the cost of storing data, because the cost of DRAM and, and, and solid state hard drives has come down with Moore's Law, and now the ability to actually have enough data to do this AI-driven, or people are calling AI, but it really is the same... It's part of a spectrum of things that have been going on for 60 years to actually drive predictions in the, um, in the world, uh, is really being realized in a bunch of areas that we would have historically been really challenged and surprised to see. And so my argument is, at this point-
- JCJason Calacanis
Big data played a big role, yeah.
- DFDavid Friedberg
Yeah, we've... Over the last decade, we've reached this tipping point in terms of data generation, storage, and computation that's allowed these statistical models to resolve dynamically.And as a result-
- CPChamath Palihapitiya
Right.
- DFDavid Friedberg
... they are far more predictive, and as a result, we see far more human-like behavior in the predictive systems, both physic- both those that are, you know, like a, like a robot is the same as one that existed 20 years ago. But the way that it's run is using the software that is driven by this dynamic model, and that's what makes it so incredible.
- CPChamath Palihapitiya
The data allows for a better answer. Chamath? Okay, I have two things to say, but one, the first one is a total non sequitur. So you use the term data scientist. Do you know where the term data scientist came from? No, I- As, as classically used in Silicon Valley? It came from Facebook, and- Ah. ... it came from my team in a critical moment. This was in 2007, I was trying to, 2008, I was trying to build the growth team. This is the team that became very famous for getting to a billion users and, you know, building a lot of these algorithmic insights. And I was trying to recruit a person from Google, and he was like a PhD in some crazy thing like astrophysics or particle physics or something. And we gave him an offer as a data analyst because this is what I needed at the time, this is what I thought I needed, an analyst, you know, to analyze data. (laughs) And he said, "Absolutely not, I'm offended by the job title." And I remember talking to my- (laughs) ... my HR, you know, business process partner, and I asked her like, "I don't understand. What is this, where is this coming from?" And she said, "He fashions himself a scientist." And I said, "Well then call him a data scientist." So we wrote in the offer for the first time data scientist, and at the time people internally were like, "This is a dumb title. What does this mean?" Anyways, we hired the guy, he was a star. Mm. And, uh, and that title just took off internally.
- DFDavid Friedberg
Chamath, it's funny because parallel, we started Climate Corp in 2006 and the original, the first guy I hired was a buddy of mine who was a 4.0 for, in applied math from Cal. And then everyone we hired on with him, we called them the math team, and they were all applied math and statistics PhDs-
- CPChamath Palihapitiya
Yeah.
- DFDavid Friedberg
... and we called them the math team, and it was really cool to be part of the math team. But then we switched the team name to data scientist and then it obviously created this much more kind of impressive role, impressive title, central function to the organization that was more than just a math person or data analyst, uh, as I think it may have been classically treated. 'Cause they really were building the algorithms that drove the models that made the product work, right?
- CPChamath Palihapitiya
Peter Thiel has a very funny observation, not funny, but you know, uh, observation, which is you should always be wary of any science that actually has science in the name, political science, social science, (laughs) I guess-
- DFDavid Friedberg
Totally.
- CPChamath Palihapitiya
... maybe data scientists, you know, because the real sciences don't need to qualify themselves, physics, chemistry, biology. Anyways, that's... So here, here's what I wanted to talk about with respect-
- DFDavid Friedberg
Yeah.
- CPChamath Palihapitiya
... to AI. Um, two very important observations that I think, uh, is useful for people to know. The first one, Nick, if you throw it up here, is just a, um, a baselining of, you know, when we have thought about intelligence and compute capability, we've always talked about Moore's Law. And Moore's Law, essentially this idea that there is a fixed amount of time where the, um, density of transistors inside of a chip would double, and roughly that period for many, many years was around two years. And it was largely led by Intel. And we used to equate this to intelligence, meaning the more density there was in a chip, the more things could be learned and understood, and we used to think about that as the progression of how, um, computing intelligence would grow and eventually AI and artificial intelligence would, would get to mass market. Well, what we are now at is a place where many people have said Moore's Law has broken. Why? It's because we cannot cram any more transistors into a fixed amount of area, we are at the boundaries of physics. And so people think, well, does that mean that our ability to compute will essentially come to an end and stop? And the answer is no, and that's what's demonstrated on this next chart, just to make it simple. Which is that what you really see is that if you think about, you know, supercomputing power, so the ability to get to an answer, that has actually continued unabated. And if you look at this chart, the reason why this is possible is entirely because we've shifted from CPUs to these things called GPUs. So you, you may have heard companies like NVIDIA, why has companies like NVIDIA done so well? It's because they said, they raised their hand and said, "We can take on the work." And by taking on the work away from a traditional CPU, you're able to do a lot of what Freeberg said is get into these very complicated models. So, this is just an observation that I think that we are continuing to compound knowledge and intelligence effectively at the same rate as Moore's Law, um, and we will continue to be able to do that, because this makes it a problem of power and a problem of money. So as long as you can buy enough GPUs from NVIDIA or build your own, and as long as you can get access to enough power to run those computers, there really isn't many problems you can't solve. And that's what's so fascinating and interesting, and this is what companies like OpenAI are really proving. You know, when they raised a billion dollars, what they did was they attacked this problem because they realized that by shifting the problem to GPUs, it left all these amazing opportunities for them to uncover, and that's effectively what they have.
- DFDavid Friedberg
Yeah.
- CPChamath Palihapitiya
The second thing that I'll say very quickly is that it's been really hard for us as a society to build intelligence in a multimodal way like our brain works. So think about how our brain works. Our brain works in a multimodal way. We can process imagery, we can process words and sounds, um, we can process all of these different modes, text, um, into one system and then intuit some intelligence from it and make a decision, right? So, you know, we could be watching this YouTube video, there's gonna be transcription, there's video, voice, audio, everything all at once.
- DFDavid Friedberg
And we are moving to a place very quickly where computers will have that same ability as well. Today, we go to very specific models in kind of Balkanized silos-
- JCJason Calacanis
Correct.
- DFDavid Friedberg
... to solve different kinds of problems. But those are now quickly merging, again, because of what I just said about GPUs. So, I think what's really important about AI for everybody to understand is the marginal cost of intelligence is gonna go to zero. And this is where I'm just gonna put out another prediction of my own. When that happens, it's gonna be incredibly important for humans to differentiate themselves from computers. And I think the best way for humans to differentiate ourselves is to be more human.
- JCJason Calacanis
Hm.
- DFDavid Friedberg
It's to be less compute-intensive. It's to be more empathetic. It's to be more emotional, a lot less emotional. Because those differentiators are very difficult for brute force compute to solve.
- JCJason Calacanis
Be careful. The replicants on this call (laughs) are getting a little nervous here. Uh-
- DFDavid Friedberg
They're not processing that. That was an emotional statement.
- JCJason Calacanis
Okay, I know, uh, do not want to process that one. Well, uh, to your point, uh, during this AI day, um, they were showing in self-driving, as you're talking about, this balkanization and trying to, um, make decisions across many different, uh, decision trees. You know, they're looking at lane changes, they're looking at other cars and pedestrians, they're looking at road conditions like fog and rain, and then they're using all this big data, to your point, Friedberg, to run, uh, tons of different simulations. So they're building, like, this virtual, uh, world, uh, on Market Street, and then they will throw people, dogs, cars, people with weird behaviors-
- DFDavid Friedberg
No, no, but, but think about that-
- JCJason Calacanis
... into the simulation.
- DFDavid Friedberg
It's such a wonderful example. Imagine that system hears a horn.
- 49:45 – 59:24
Biden pardons all prior federal offenses of marijuana possession
- JCJason Calacanis
want to go-
- DFDavid Friedberg
Hey, Cal, can we-
- JCJason Calacanis
Yeah.
- DFDavid Friedberg
Let's talk about the marijuana, uh, pardoning-
- JCJason Calacanis
Yeah. Breaking news.
- DFDavid Friedberg
... that, that Biden just did. Yeah.
- JCJason Calacanis
Yeah. I was about to say, we got a couple of things we really want to get to here. Uh, Ukraine, Section 230, and then this breaking news. Uh, we'll pull it up here on the screen. While we're recording the show, President Biden says, and I'm just gonna quote here, "First, I'm pardoning all prior federal offenses of simple marijuana possession. There are thousands of people who are, were previously convicted of simple possession who may be denied employment, housing, or educational opportunities. As a result, my pardon will remove this burden." That's big news. "Second, I'm calling on governors, uh, to pardon simple state marijuana possession offenses. Just as no one should be in a federal prison solely for possessing marijuana, nobody should be in a local jail or state prison for that reason either." Finally, this is happening. Uh, third, uh, and this is an important one, "We classify marijuana at the same level as heroin and even, and more serious than fentanyl." Makes no sense. "I'm asking Secretary Bar- Becerra, uh, and the Attorney General to initiate the process of reviewing how marijuana is scheduled under federal law. I'd also like to note that, uh, as federal and state regulators change, we still need important limitations on trafficking, marketing, and underage sales of marijuana." Thoughts on this breaking news?
- DFDavid Friedberg
Sacks, does... Is this mo- Do you think that the, the timing on this is kind of midterm related? Did, did... It seems... Is this a, is this... I guess, is this a politically popular decision to do?
- DSDavid Sacks
I think so. I mean, look, I support it. So, uh, bi- Biden finally did something I like. (laughs)
- JCJason Calacanis
Great.
- DSDavid Sacks
I mean, I've thought that we should decriminalize marijuana for a long time, or specifically, I, you know, I agree with this idea of de-scheduling it. It, it, it does not make sense to treat marijuana the same as heroin, as a Schedule I narcotic. Just doesn't make any sense. It should be regulated separately and differently. Obviously, you wanna keep it out of the hands of minors, but no one should be going to jail, I think, for simple possession. So, I do agree with this and I think the thing they need to do, I don't see it mentioned here, is they should pass a federal law that would allow for the normalization of, let's call it, lega- legal, um, you know, cannabis companies. So, so companies that are allowed to operate under state laws, like in California, should have access to the banking system, should have access to the payment rails. Because right now, the reason why the legal cannabis industry isn't working at all in California is because they can't bank, they can't take payments, so it's this weird all-cash business that makes no sense. So, so listen, if we're not gonna criminalize it as a drug like heroin, if we're gonna allow states to make it legal, then allow it to be a more normal business where the state can tax it and it can operate in a more above board way. So, but look, I think this is-
- JCJason Calacanis
Federal, federal mandate is what you're saying. The federal mandate.
- DSDavid Sacks
I think it's, could still be regulated on a state by state basis, but I think you need the feds to bless the idea that banks and payment companies can take on those clients which states have already said are legally operating companies. And right now, they can't, and it's a huge gap in the law. So, maybe that's the one thing I would add to this, but I don't have any complaints about this right now based on-
- JCJason Calacanis
This is awesome.
- DSDavid Sacks
... what we know from this tweet storm, and I would say-
- JCJason Calacanis
Sacks, what are the polling data? What are the pol-
- DSDavid Sacks
... this, this, by the way, was an about face. This was an about face by Biden.
- DFDavid Friedberg
Yeah. Do you know what the polling data says? I mean, is, is there... Uh, I'm assuming there's big support in kind of the independence and the middle, uh, for this move.
- JCJason Calacanis
It was 70% at one point.
- DSDavid Sacks
Yeah.
- DFDavid Friedberg
Yeah.
- DSDavid Sacks
So, so look, this, th- To me, this is the kind of thing that Biden should be doing with a 50/50 Senate, finding these sorts of bipartisan compromises.
- DFDavid Friedberg
Right.
- DSDavid Sacks
So yeah. I... Look, this is good news far as I'm concerned.
- DFDavid Friedberg
Why hasn't this happened in the past? Like, what's been the political reason that other presidents, Obama even, didn't, that, that have espoused-
- JCJason Calacanis
There was rumors-
- DFDavid Friedberg
... similar ideology, like, why, why w- but why... Does anyone know why this hasn't been done in the past?
- JCJason Calacanis
There was rumors he was gonna do it in the second term, but he just didn't have the political capital-
- DFDavid Friedberg
But why?
- JCJason Calacanis
... to do it.
- 59:24 – 1:13:40
SCOTUS will hear cases regarding Section 230, common carrier, algorithmic recommendations
- DFDavid Friedberg
today.
- JCJason Calacanis
All right. So SCOTUS is gonna hear two, uh, cases for Section 230. The family of Nohemi Gonzalez, a 23-year-old American college student who was killed in an ISIS terrorist attack in Paris back in 2015, you remember those horrible attacks, is claiming that YouTube helped and, uh, aided and abetted ISIS. Uh, the family's argument is YouTube's algorithm was recommending videos that make it, uh, that makes it a publisher of content. As you know, it's Section 230, common carrier. Now, if you make editorial decisions, if you promote certain content, you lose your 230 protections. Uh, in court papers filed in 2016, they said the company, quote, "Knowingly permitted ISIS to post on YouTube hundreds of radicalizing videos inciting violence, which helped the group recruit, including some who were actually involved in the terrorist attacks." So they have made that connection.
- CPChamath Palihapitiya
Well, look, let's, let's be honest, we can, we can, we can put a pin in this thing because I think it would be shocking to me if this current SCOTUS, uh, all of a sudden found it in the cockles of their heart to protect Big Tech. I mean, they've dismantled, um, a lot of other stuff that I think is a lot more controversial than this. Um, and so, you know, we've, we've basically looked at gun laws, we've looked at, uh, affirmative action, we've looked at abortion rights. So well, I mean, I think a- as, as we've said, I think we all know where that die is unfortunately going to get cast. Um, so to me it just seems like this could be an interesting case where it's actually nine-zero in favor for complete, for completely different sets of reasons. I mean, if you think of the liberal left part of the court, they have their own reasons for saying that there aren't 230 protections for Big Tech. And if you look at the, the, the far right or the, the right-leaning parts, members of the SC- of SCOTUS, they have a, they have another set of different reasons.
- JCJason Calacanis
So you think you're gonna make a political decision, not a legal?
- CPChamath Palihapitiya
No, but e- even in their politics, they actually end up in the same place. They both don't want the protections, but for different reasons. So there, there is a reasonable outcome here where, you know, uh, Roberts is gonna have a really interesting time trying to pick who writes the majority opinion.
- JCJason Calacanis
There was a related case in the Fifth Circuit in Texas where... Do you guys see this, uh, Fifth Circuit decision where...
- DSDavid Sacks
... Texas passed a law imposing common carrier restrictions on social media companies, the idea being that social media companies need to operate like phone companies and they can't just arbitrarily deny you service or deny you access to the platform. And the argument why previously that had been viewed actually as unconstitutional was this idea of compelled speech, that you can't compel a corporation to support speech that they don't want to because that was a violation of their own First Amendment rights. And what the first, uh, the Fifth Circuit said is, "No, that doesn't make any sense. Facebook or Twitter can still advocate for whatever speech they want as a corporation. But as a platform, they, if Texas requires them to not discriminate against people on the basis of their viewpoint, then Texas has the right to, to impose that." Because that doesn't, it, their quote was, "That does not chill speech. If anything, it chills censorship." So in other words-
- JCJason Calacanis
What's the right legal decision here, in your mind? Putting aside politics, if you can, for a moment, putting on your-
- DSDavid Sacks
Yeah.
- JCJason Calacanis
... legal hat, what is the right thing for society? What is the right legal issue around Section 230, specifically in the YouTube case, and just generally, should we look at YouTube? Should we look at a blogging platform like Medium or Blogger, Twitter? Should we look at those as common carrier and they're not responsible for what you publish on them? Obviously, they have to take stuff down if it, if it breaks our terms of service, et cetera-
- DSDavid Sacks
Yeah.
- JCJason Calacanis
... or if it's illegal-
- DSDavid Sacks
I've made the case before that I, I do think that common carrier requirements should apply on some level of the stack to protect the rights of ordinary Americans to have their speech in the face of these giant monopolies which could otherwise de-platform them for arbitrary reasons. Just to, you know, just to explain this a little bit. So historically, there was always a debate between, uh, so-called positive rights and, and negative rights. So where the United States started off as a country was with this idea of negative rights, that what a right meant is that you'd be protected from the government taking some action against you. And if you look at the Bill of Rights, you know, the original rights are all about protecting the citizen against an intrusion on their liberty by, by a state or by the federal government. In other words, Congress shall make no law. It was always a restriction. So the right was negative. It wasn't sort of positively enforced. And then with the progressive era, you started seeing, you know, more, uh, progressive rights, like, for example, uh, American citizens should have the right to healthcare, right? That's not protecting you from the government. That's saying that the government can be used to give you a right that you didn't otherwise have. And so that was sort of the big progressive revolution. My take on it is I actually think that the problem we have in our society right now is that free speech is only a negative right. It's not a positive right. I think it actually needs to be a positive right. I'm embracing a more progressive-
- JCJason Calacanis
Wow.
- DSDavid Sacks
... version of rights, but on behalf of sort of this original negative right. So, and the reason is because the, the town square got privatized, right? I mean, you used to be able to go anywhere in this country, there'd be a multiplicity of town squares, anyone could pull out their soapbox, draw a crowd, they could listen. That's not how speech occurs anymore. It's not on public land or public spaces. The way that speech, political speech especially, occurs today is in these giant social networks that are m- that have giant network effects and are basically monopolies. So if you don't protect the right to free speech in a positive way, it no longer exists.
- JCJason Calacanis
So you not only believe-
- CPChamath Palihapitiya
Wow.
- JCJason Calacanis
... that YouTube should keep its Section 230, you believe they, YouTube shouldn't be able to de-platform as a private company, you know, uh, Alex Jones is but one example, they should have their free speech rights, and we should lean on that side of forcing YouTube to put Alex Jones or Twitter to put Trump back on the platforms. Is that your position?
- DSDavid Sacks
I'm not saying that the Constitution requires YouTube to do anything. Uh, what I'm saying is that if a state like Texas or if the federal government wants to pass a law saying that YouTube, if you are, say, of a certain size, you're a social network of a certain size, you have monopoly network effects, I wouldn't necessarily apply this to all the l- little guys, but for those big monopolies, we know who they are, if the, if the federal government or a state wanted to say that they are required to be a common carrier and they cannot discriminate against certain viewpoints, I think the government should be allowed to do that because it furthers a positive right.
- JCJason Calacanis
Hm.
- DSDavid Sacks
Historically, they've not been able to do that because-
- JCJason Calacanis
I agree with that.
- DSDavid Sacks
... of this idea, because this idea of compelled speech, meaning that it would infringe on YouTube's speech rights.
- JCJason Calacanis
By the way-
- DSDavid Sacks
I don't think it would. I mean, Google and YouTube can advocate for whatever positions they want. They can produce whatever content they want.
- JCJason Calacanis
Yeah.
- DSDavid Sacks
But, but the point that, and I think Section 230 kind of makes this point as well, is that they are platforms, they're distribution platforms. They're not publishers.
- JCJason Calacanis
Okay.
- DSDavid Sacks
So if they want, so especially if they want Section 230 protection, they should not be engaging in viewpoint discrimination.
- JCJason Calacanis
Okay. So now there is a rub here, Chamath-
Episode duration: 1:25:29
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode 0JTcxIVlQfU
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome