TRANSCRIPTIE 🔗
WEBVTT
1
00:00:00.060 --> 00:00:05.460
Something about AI that causes the people in the field, I would say, to become both excessively utopian and excessively apocalyptic.
2
00:00:05.580 --> 00:00:13.080
Having said that, I think what's actually happened is an enormous amount of technical progress that built up over time. And like for, for example, we now know the neural network is the correct architecture.
3
00:00:13.280 --> 00:00:17.700
And I, I will tell you, like there was a 60-year run where that was like a f- you know- Yeah... even 70 years where that was controversial.
4
00:00:17.920 --> 00:00:25.340
And so, so the way I think about what's happening is basically I think, I think about basically the, the, the period we're in right now is it's, I call it 80-year overnight success. [laughs] Right?
5
00:00:26.120 --> 00:00:32.600
Which is like, it's an overnight success 'cause it's like bam, you know, ChatGPT hits, and then, and then o1 hits, and then, you know, OpenClaw hits.
6
00:00:32.640 --> 00:00:43.080
And like, you know, these are open- these are, these are like overnight, like radical overnight transformative successes, but they're drawing on an 80-year sort of wellspring backlog, [laughs] you know, of, of, of, of ideas and thinking.
7
00:00:43.440 --> 00:00:51.240
It's not just that it's all brand new, it's that it's an unlock of all of these decades of like very serious hardcore research. If I were 18, like this is 100-- this is what I would be spending all of my time on.
8
00:00:51.280 --> 00:00:58.440
This is like such an incredible conceptual breakthrough. Before we get into today's episode, I just have a small message for listeners. Thank you.
9
00:00:58.600 --> 00:01:07.120
We would not be able to bring you the AI engineering, science, and entertainment content that you so clearly want if you didn't choose to also click in and tune into our content.
10
00:01:07.260 --> 00:01:17.060
We've been approached by sponsors on an almost daily basis, but fortunately, enough of you actually subscribe to us to keep all this sustainable without ads, and we wanna keep it that way.
11
00:01:17.440 --> 00:01:24.340
But I just have one favor to ask all of you. The single most powerful, completely free thing you can do is to click that subscribe button.
12
00:01:24.500 --> 00:01:32.880
It's the only thing I'll ever ask of you, and it means absolutely everything to me and my team that works so hard to bring The Late In Space to you each and every week.
13
00:01:33.400 --> 00:01:46.740
If you do it, I promise you, we'll never stop working to make this show even better. Now let's get into it. [upbeat music] Hey, everyone, welcome to the Late In Space podcast.
14
00:01:46.900 --> 00:01:58.460
This is Alessio, founder of Kernel Labs, and I'm joined by Swyx, editor of Late In Space. Hello, and we're in a16z with A, uh, Marco Iacone. Welcome. Yes. Yes. A and what, half of 16? [laughs] Something like that. A one.
15
00:01:58.520 --> 00:02:04.780
Exactly.
Load full transcript
Exactly. Uh, a- apparently this is the, the final few days in your, your current office. You're moving across the road. Uh, we're...
16
00:02:04.820 --> 00:02:11.060
Yeah, we have a little bit of some, we have some projects underway, but yeah- Yeah... this is... Actually, this is the original. We're in actually the original office. We're in the, we're in the, we're in the- Nice...
17
00:02:11.260 --> 00:02:21.360
we're in the whole thing. It's beautiful. Yeah. Great. Thank you. So I have to come out. Uh, this is a, you know, I wanted to pick a spicy start. [laughs] In October 2022, I just made friends with Rune, and- [laughs]...
18
00:02:21.360 --> 00:02:35.600
uh, I wanted to give him something to sort of be spicy about, and I said, uh, uh, "It'll never not be funny that a16z was constantly going, 'The future is where the smart people choose to spend their time,' and then going deep into crypto and not in AI," and that was in October 20, 2022.
19
00:02:36.000 --> 00:02:46.060
And Rune says there was an internal meeting in a16z to reorient around GenAI. Obviously, you have, but was there a meeting? What, what was that? [laughs] I mean, I don't... Look, I've been doing AI since the late '80s.
20
00:02:46.280 --> 00:02:52.220
Yeah. So I, I don't know, like all the... As far as I'm concerned, this stuff is all Johnny-come-lately. Yeah. You know, I mean, look, we've been doing AI our entire existence.
21
00:02:52.320 --> 00:02:59.500
I mean, we've been doing AI, machine learning deep, you know, deep... We've been doing this stuff w- w- way from the beginning. Obviously. A- AI is just core to computer science.
22
00:02:59.660 --> 00:03:06.340
I, I, I actually view them as like quite, uh, quite continuous. Um, you know, Ben and I both have computer science degrees. Um, you know, we, we both...
23
00:03:06.380 --> 00:03:11.900
Ben, Ben and I actually both are, we're old enough to remember the actual AI boom in the 1980s. Yeah. There was like a, there was a big AI boom at the time.
24
00:03:12.160 --> 00:03:21.080
Um, and there was a, there's glittered names like Expert Systems, um, and an era of like LISP and LISP machines. Um, I, I coded in LISP. I was coding in LISP in, in 1989- Yeah...
25
00:03:21.100 --> 00:03:28.620
where that was the, the language of the AI future. Um, yeah, so this is something that we're like completely, you know, completely comfortable with and been doing the whole time and are very enthusiastic about.
26
00:03:29.120 --> 00:03:37.480
Is there a strong like this time is different? Because, uh, my closest an- analog was 2016, '17. There was also an AI boom. [laughs] And it petered out very, very quickly.
27
00:03:37.980 --> 00:03:44.680
Um- Well- It just, it just in terms of investing... sort of. Sort of. Yeah. Investment- Yeah, yeah, yeah... investment excitement. Although that's really when the, the, the Nvidia phenomenon really...
28
00:03:44.700 --> 00:03:53.880
It, it, it was, it, I would say it was in that period when it was very clear that at, at the time, it, it, the vocabulary was more machine learning, but it, it was very clear at that time that machine learning was hitting some sort of takeoff point.
29
00:03:54.160 --> 00:03:54.340
Yeah.
30
00:03:54.380 --> 00:04:02.100
Well, and as you guys, you guys have talked about this at length on your, on your thing, but, you know, if you really track what happened, I think the real story is it was, it was the AlexNet, uh, basically breakthrough- Yes...
31
00:04:02.140 --> 00:04:09.620
in like 2013. That was the, that was the real knee in the curve. Um, and then it was obviously the transformer breakthrough in '17. Yeah. Um, and then everything that followed.
32
00:04:09.660 --> 00:04:19.560
But, but, you know, look, machine learning, you know, there were, you know, look, uh, I mean, look, I've been working, you know, I've been working with, at the, one of my, you know, kind of projects working with Facebook since 2004, um, and on the board since 2007.
33
00:04:19.579 --> 00:04:31.080
And of course, you know, they, they started using machine learning very early, um, and you know, have used it basically, you know, for like 20 years for, you know, content, you know, feed optimization and advertising optimization and obviously many, you know, financial services.
34
00:04:31.120 --> 00:04:36.920
You know, many, many, many companies, many different sectors have been doing this. And so it's like one of these things. It's like it's not a sing- it's not a single thing.
35
00:04:36.960 --> 00:04:43.820
Like it's, it's like l- it's like layers, right? Yeah. Um, and, and the layers arrive at different paces, and, but they kind of build up. Yeah. Uh, they kind of build up over time.
36
00:04:43.840 --> 00:04:49.040
And then, and then, yeah, and then look, in retrospect, it was 2017 was kind of the, you know, the key, the key point with the trans- transformer.
37
00:04:49.080 --> 00:04:56.500
And then, and then as you guys know, there was this really weird like four-year period where it's like the t- the transformer existed, and then it was just like- Let's go. [laughs] Yeah.
38
00:04:56.680 --> 00:05:05.380
Well, but, but it was just- But, but between 2020-- but between 2017 and 2021, I, I mean, that was the era of which like companies like Google had internal chatbots, but they weren't letting anybody use them. Yeah.
39
00:05:05.560 --> 00:05:11.600
Right? And then, you know, and then OpenAI developed ChatGPT, or GPT-2, and then they told everybody, "This is way too dangerous to deploy, right?" Yeah. Yeah.
40
00:05:11.620 --> 00:05:19.320
"We're," you know, "we can't possibly let normal people, normal people use this thing." And then do, you guys, I'm sure, remember AI Dungeon. Um- Mm-hmm. So the o- there was like a year- That was old one. Yeah...
41
00:05:19.340 --> 00:05:26.560
where like the only way for a normal person to use GPT-3 was in, in AI Dungeon. Yeah. And so you, you... We would do this. You'd go in there, and
you'd pretend to play Dungeons & Dragons.
42
00:05:26.580 --> 00:05:28.600
In reality, you're just trying to talk to, talk to GPT.
43
00:05:29.200 --> 00:05:35.920
And so there was this, you know, there was this long, you know, and the, I, you know, the big c- big companies, you know, big companies are cautious, and you know, the big companies were cautious. It, it...
44
00:05:36.360 --> 00:05:44.040
By the way, it took OpenAI, you know, they, they, they talk about this. It took OpenAI time to actually adjust, you know, kind of re- redirect their research path. Yeah. I, I think, uh- So-... it was at Rosewood, right?
45
00:05:44.460 --> 00:05:54.580
Uh, that the dinner that founded OpenAI was right there. Right. Right. But that, that dinner would've taken place in 20- '18 The, the formation of OpenAI? Uh-huh As late as 2018? I, I... Sorry, uh- Okay...
46
00:05:54.620 --> 00:06:01.980
no, I'm, I'm, I'm wrong. Okay. It should be 20-- Yeah, they just celebrated their 10-year anniversary, so it, it is- Okay... 2025. Yeah. That's it, so 2015. Yeah, 2015. Yeah, 2015.
47
00:06:02.520 --> 00:06:13.540
But then, uh, um, Alec Radford did GPT 1 in- Mm-hmm... what? Probably- 17, 18. Yeah... 18, 17, 18. So it, it... Yeah, for the... And then, and then they didn't really... And then GPT 3 was what? 2020, 20- 2020- 2020...
48
00:06:13.900 --> 00:06:21.880
because that became co-pilot immediately. Yeah... 2021. Even OpenAI, which has been, you know, the leader of l- of this thing in the last decade, you know, e- even they had to adapt and, and, and lean into the new thing.
49
00:06:21.920 --> 00:06:33.080
And so, um, yeah, I, I think it's just this process of basically w- sort of wave after wave, layer after layer, you know, building on itself, and then you kind of get these catalytic moments where the, where the whole thing pops, and, and obviously that's what's happening now.
50
00:06:33.680 --> 00:06:42.200
Is it useful to think about will there be an AI winter? 'Cause there's always these patterns. Like, is this endless summer? It's something I constantly think about because do I get...
51
00:06:42.420 --> 00:06:53.520
Do I just, like, just get endlessly hyped and just trust that I will only be early and never wrong? Or [laughs] Right... well, are we, uh... Will there be a winter? [laughs] So there's something about...
52
00:06:54.000 --> 00:07:00.600
Say the following, there's something about AI that has led to this repeated pattern. Um, and, and, a- and y- you guys know this, but- It's a w- summer, winter, summer, winter...
53
00:07:00.660 --> 00:07:10.000
summer, winter, summer, winter, and it goes back 80 years. Yeah. 80 years. Uh, so the original neural network paper was 1943- Mm-hmm... right? Which is, which is amazing, uh, that it was, it was far back that long.
54
00:07:10.040 --> 00:07:18.140
And then there was... You... I don't know if you guys have ever talked about this on your show, but there was this, uh... There was a big, uh... There was an AGI conference at Dartmouth University in 1950- '55... '55.
55
00:07:18.280 --> 00:07:26.660
Yeah. And they got an NSF grant to, uh, for the, all the AI experts at the time to spend the summer together, and they figured if they had 10 weeks together, they could get AGI [laughs]... out the other end.
56
00:07:26.680 --> 00:07:31.160
And they got their... By the way, they got the grant, they got the 10 weeks, and then, you know, 1955- Yeah... you know, no, no AGI.
57
00:07:32.160 --> 00:07:35.300
And like I said, I, I lived through the '80s version of this where there was a big, a big boom and a crash.
58
00:07:35.340 --> 00:07:42.280
And so, so there is this thing and there, there is something about AI that causes the people in the field, I would say, to become both excessively utopian and excessively apocalyptic.
59
00:07:42.760 --> 00:07:46.860
Um, and, and it's probably on both sides of, like, the, the, the boom bust cycle you, you kinda see that play out.
60
00:07:47.600 --> 00:07:54.760
Having said that, I think what's actually happened is, like, just at, in, you know, and we now know in retrospect, like, an enormous amount of technical progress that built up over time.
61
00:07:54.800 --> 00:08:00.920
And like, for, for example, we now know the neural network is the correct architecture. And I, I will tell you, like, there was a 60-year run where that was like a, you know- Yeah...
62
00:08:00.940 --> 00:08:08.140
or even 70 years where that was controversial and, and we now know that that's the case. And so we, we now... You know, everything we're building on today sort of derives from the original idea in 1943.
63
00:08:08.740 --> 00:08:15.120
And so, so in retrospect, we, we now know that, like, these, these guys are right. They, you know, they would get the timing wrong, and they thought, you know, capabilities would arrive faster, or they were...
64
00:08:15.140 --> 00:08:18.160
It could be turned into businesses sooner or whatever. But, like, they were fundamentally...
65
00:08:18.180 --> 00:08:25.140
The, the scientists who worked on this over the course of decades were fundamentally correct about what they were doing and, and the, and the payoff from, from, from all their work is happening now.
66
00:08:25.780 --> 00:08:33.200
And so, so the way I think about what's happening is basically I think, I think about basically the, the, the period we're in right now is it's, I call it 80-year overnight success. Right.
67
00:08:33.200 --> 00:08:41.500
[laughs] Which is like, it's an overnight success 'cause it's like bam, you know, ChatGPT hits and then, and then o1 hits and then, you know, OpenClaw hits and, like, you know, these are open...
68
00:08:41.540 --> 00:08:50.940
These are, these are like overnight, like radical overnight transformative successes, but they're drawing on an 80-year sort of wellspring backlog [laughs] you know, of, of, of, of ideas and thinking.
69
00:08:51.300 --> 00:08:57.320
It's not just that it's all brand
new, it's that it's an unlock of all of these decades of like very serious hardcore research, um, and thinking.
70
00:08:57.380 --> 00:09:03.260
I mean, look, there were AI researchers who spent their entire lives, [laughs] they got their PhD, they, they worked for...
71
00:09:03.320 --> 00:09:09.880
They've researched for 40 years, they retired, in a lot of cases they passed away and they never actually saw it work. Yeah. So sad. It is, it is sad. It is sad.
72
00:09:09.900 --> 00:09:15.540
And I know some- I think Jack Hinton was like the last guy. [laughs] Yeah, yeah. Well, there were guys, I think it was a guy, Alan Newell. I mean, there's tons of... John McCarthy.
73
00:09:15.540 --> 00:09:21.240
[laughs] You know, John McCarthy was like one of the inventors of the field. He's one of the guys who organized the Dartmouth conference and, you know, he taught at Stanford for 40 years and passed- Wow...
74
00:09:21.280 --> 00:09:27.740
you know, passed away, I don't know, whatever, 10, 10 years ago or something. Never, never actually got, got to see it happen. But like it, it is amazing in retrospect.
75
00:09:27.780 --> 00:09:35.240
Like these guys were incredibly smart and they worked really hard and they were correct. So anyway, so then it's like, okay, you know, as they say, history doesn't repeat but it rhymes.
76
00:09:35.600 --> 00:09:39.140
It's like, okay, does that mean that there's gonna be another like, you know, basically boom bust cycle?
77
00:09:39.180 --> 00:09:47.540
And I, I will tell you like, look, like i- in a sense like, yes, everything goes through cycles and, you know, people get overly enthusiastic and overly depressed and there's, there's a time, there's a timeless to that.
78
00:09:48.100 --> 00:09:54.580
Having said that, there's just no question. Um, [laughs] so the four mo- the four most dangerous words in investing- This time is different... are this time is different.
79
00:09:54.840 --> 00:10:02.660
Do you know the 12 most dangerous words i- in, in investing? No. The four most dan- four most dangerous words- [laughs]... in investing are this time is different. Yeah. Um, the 12 most dangerous words.
80
00:10:03.200 --> 00:10:11.039
And so like I'll tell you what's different, like s- now it's working. Like [laughs] like there's just no... I mean, look, there's just no question. And by the way, I'll, I'll just give you guys my take.
81
00:10:11.099 --> 00:10:18.520
Like LL- LLMs, like from, from basically the ChatGPT moment through to spring of '25, I think you could still...
82
00:10:18.740 --> 00:10:34.700
I think well-intentioned, well-informed skeptics could still say, "Oh, this is just pattern completion," and, "Oh, these things don't really understand what they're doing," and, you know, "The hallucination rates are way too high," and, you know, "This is gonna be great for creative writing and creating, you know, Shakespearean so- sonnets and, you know, as, as rap lyrics or whatever.
83
00:10:34.740 --> 00:10:43.780
Like it's gonna be great at all that stuff, but we're not gonna be able to harness this to make this relevant in, you know, coding or in medicine or in law or in, you know, f- you know, kind of fields that, you know, kind of really, really matter."
84
00:10:44.280 --> 00:10:46.140
And I think basically it was the reasoning breakthrough.
85
00:10:46.400 --> 00:10:52.680
It was o1 and then R1 that basically answered that question and basically said, "Oh no, we're gonna be able to actually turn this into something that's gonna work in the real world."
86
00:10:52.840 --> 00:10:58.800
And, and then obviously the coding breakthrough over the, over... Basically the coding breakthrough that kind of catalyzed over the holiday break was kind of the third step in that. Mm-hmm.
87
00:10:58.820 --> 00:11:07.260
We were just like, all right, if, if, you know, if Linus Torvalds is saying that AI coding is now better than he is, like- [laughs]... like that's, that's never happened before. That's the benchmark. Yeah.
88
00:11:07.340 --> 00:11:09.300
That's never happened before, and so now we know that it's- Mm...
89
00:11:09.340 --> 00:11:15.320
it's gonna sweep through coding and, and then, and then we, we know, you know, we know that if it's gonna work in coding, it's gonna work in everything else, right? It's just...
90
00:11:15.480 --> 00:11:21.520
Th- then 'cause that's, that's like, that's like, that's like the hardest... In many ways, that's the hardest example and now everything else is gonna be a, a derivative of that.
91
00:11:22.080 --> 00:11:27.620
And then on top of that, we just got the agent breakthrough wi- you know, with OpenClaw, which is fantastic, which is amazing and incredibly powerful.
92
00:11:27.660 --> 00:11:32.700
And then we just got the, the, um, the auto research, um, you know, the, the self-improvement. You know, we're now into the self-improvement breakthrough.
93
00:11:32.720 --> 00:11:43.740
And so the, so the way I think about it is we've had four fundamental breakthroughs in functionality, LLMs, reasoning, uh, agents, um, and then, uh, and, and then now, uh, RSI. Um, and, and they're all actually working.
94
00:11:44.040 --> 00:11:49.000
Um, and so I'm, I'm just as you like- [laughs]... as you can tell, I'm jumping out of my shoes. Like, like this is, like this is it.
95
00:11:49.040 --> 00:11:54.740
Like th- this is the culmination of 80 years' worth of, worth of work and this is the time it's becoming real. Yeah. Amazing. Yeah. I, I'm completely convinced.
96
00:11:55.100 --> 00:12:03.520
I think the anxiety that people feel is like during the transistor era, you had Moore's law and it's like, all right, we understand why these things are getting better. We understand the physics of it. Yeah.
97
00:12:03.560 --> 00:12:13.040
With AI, it's, it's so jagged in like the jumps. Yeah. Where li- like you said, it's like in three months you have like this huge jump like- And people are like, "Well, this can't keep
happening," right?
98
00:12:13.160 --> 00:12:18.840
But then it keeps happening. It'll keep happening. And so, like, how do you think about also timelines of, like, what's worth building?
99
00:12:18.940 --> 00:12:27.660
I think we always have this question with guests, which is like, you know, should you spend time building harness- Right... for a model versus, like, the next model just gonna do it one shot in the latest base? Right.
100
00:12:28.020 --> 00:12:33.560
And how does that inform, like, how you think about the shape of the technology? You know, you talk about how it's a new computing platform.
101
00:12:34.160 --> 00:12:41.280
If you have a computing platform that, like, every six months it, like, drastically changes- Yeah... in what it looks like, it's hard to build companies on top of it. Yeah. Yeah, so, so a couple things.
102
00:12:41.320 --> 00:12:44.500
So one is, like, look, the, the M- Moore's law was what we now call a scaling law.
103
00:12:44.620 --> 00:12:51.680
Like when Moore's law was a scaling law, and for your younger viewers, Moore, Moore's law was every chip, chip- chips either get twice as powerful or twice as cheap every, every 18 months.
104
00:12:52.260 --> 00:12:58.580
And that, and that, and that, you know, that it's gotten more complicated in the last few years, but like that, that was, like, the 50-year trajectory of, of, of, of the computer industry.
105
00:12:59.080 --> 00:13:06.160
And then, and then by the way, and that's what took the mainframe computer from a $25 million current dollar thing into, you know, the phone in your pocket being, you know, a million times more powerful than that.
106
00:13:06.200 --> 00:13:09.460
Like that, you know, for, for 500 bucks. And so that, that was a scaling law.
107
00:13:09.480 --> 00:13:16.640
And then, and then, and then key to any scaling law, including Moore's law and the AI scaling laws is, you know, they're not really laws, right? They're, they're, they're, they're predictions.
108
00:13:17.220 --> 00:13:20.200
But when they work, they become self-fulfilling predictions because they, they, they- Right...
109
00:13:20.420 --> 00:13:25.720
they, they set a benchmark and, and then the entire industry, right, all the smart people in the industry kind of work to make sure that, that, that actually happens.
110
00:13:26.380 --> 00:13:32.540
And so they, they kind of motivate the breakthroughs that are required to, to keep that going. And in, and in, and in chips, that was a 50-year, you know, that was a 50-year run, right?
111
00:13:32.580 --> 00:13:40.620
And it w- it was amazing, and it's still happening in, in, in some areas of, uh, of chips. I think the same thing is happening with the, the core scaling laws, the core scaling laws in, in, in AI.
112
00:13:41.040 --> 00:13:47.980
You know, they're not, they're not really laws, but, like, they, they are basically, they're predictions, and then they're motivating catalysts for the research work that is required to be...
113
00:13:48.020 --> 00:13:53.100
And, and, and by the way, also the investment, uh, dollars, um, uh, you know, required to basically keep, you know, keep the curves going.
114
00:13:53.720 --> 00:13:59.560
And, and look, it, it's, it's gonna be complicated, and it's gonna be variable, and there, you know, there are gonna be walls that are gonna look like they're fast approaching, and then they're gonna be- Mm-hmm...
115
00:13:59.580 --> 00:14:04.620
you know, engineers are gonna get to work, and they're, you know, figure out a way to punch through the walls. And obviously that's, you know, that's been happening a lot.
116
00:14:05.080 --> 00:14:10.840
You know, and then look, there's gonna be times when it looks like the walls have, you know, the, the, the laws have petered out, and then they're gonna, they're gonna pick up again and surge.
117
00:14:10.880 --> 00:14:15.300
And then, and then, and then it, it appears what's happening to the AI is there's now multiple, you know, multiple scaling laws.
118
00:14:15.980 --> 00:14:21.820
Um, there's multiple areas of improvement, and, and I think, you know, I don't know how many more there are already yet to be discovered, but there are probably some more that we don't know about yet.
119
00:14:22.420 --> 00:14:28.580
You know, they, like, for example, there's probably some scaling law around, um, world models and robotics that we don't fully under- you know, kind of acquisition of data- Mm...
120
00:14:28.640 --> 00:14:34.840
at scale in the real world that we don't fully understand yet, so that, that, that one will probably kick in at some point here. There's a bunch of really smart people working on that.
121
00:14:34.940 --> 00:14:43.200
Um, and so yeah, I, I think the expectation is that, that, you know, the, these scaling laws generally are gonna continue. Yeah, the, the pace of improvement will continue to move really fast. Um,
122
00:14:44.120 --> 00:14:48.220
to, to your question on, like, what to build, so, uh, I'm a complete believer that the scaling laws are gonna continue.
123
00:14:48.280 --> 00:14:52.300
I'm a complete believer the capabilities are gonna keep getting amazing, um, you know, le- leaps and bounds.
124
00:14:52.360 --> 00:15:06.940
Uh, the part where I kind of part ways a little bit with how, what I would describe as the AI purists, um, you know, which is, which I would characterize as, like, the people who are m- in many ways the smartest people in the field, but also the people who spend their entire life, like, in a lab, um, and have, have, I would say, have very little experience in the outside world.
125
00:15:07.100 --> 00:15:07.180
Mm-hmm.
126
00:15:07.180 --> 00:15:20.540
Um, the, the, the nuance I would offer is the outside world of eight billion people and institutions and governments and companies and economic systems and social systems is really complicated, um, and um, and doesn't, you know, it, it...
127
00:15:21.380 --> 00:15:26.160
Eight billion
people making collective decisions on planet Earth is not a simple process of, like, just, like...
128
00:15:27.260 --> 00:15:35.840
Uh, you see this happening now, it's like a bunch of the AI CEOs have this thing, which is just like, well, there's just this, they just all have this kind of thing when they talk in public where they're just like, "Well, there's this, these obvious set of things that society needs to do."
129
00:15:36.040 --> 00:15:43.600
Mm-hmm. And then they're like, "Society's not doing any of those things," right? And it's like, how can society not... You know, what- whatever their theory is, how can society not see X, Y, Z? Mm-hmm.
130
00:15:43.640 --> 00:15:46.380
And the answer is, well, society is, number one, there's no single society.
131
00:15:46.420 --> 00:15:51.440
It's like eight billion people, and they, like, all have a voice, and they all have a vote, like, at the end of the day of how they, they react to change.
132
00:15:51.480 --> 00:16:00.080
And then, you know, it just, like, it's just human reality is just really complicated and messy. Um, and, and, and so the specific answer to your question is, like, as usual, it depends.
133
00:16:00.100 --> 00:16:06.080
[laughs] Um, you know, it, it depends. Look, p- there's no question people are gonna like... There's no question there are gonna be companies, it's already happening.
134
00:16:06.400 --> 00:16:11.180
There are companies that think that they're building value on top of the models, and they're just gonna get blitzed by the, by the next model. There's no question that's happening.
135
00:16:11.580 --> 00:16:20.920
But I think there's no question also that just the process of adaptation of any technology into the real, in- into the real messy world of humanity is, is just going to be messy and complicated.
136
00:16:20.960 --> 00:16:22.440
It's, it's not going to be simple and straightforward.
137
00:16:22.460 --> 00:16:33.340
It's gonna be messy and complicated, and there are gonna be a lot of companies and a lot of products, um, uh, and in, and in fact, entire industries that are gonna get built that, that, to basically actually help all of this technology actually reach real people.
138
00:16:33.900 --> 00:16:40.740
The amount of capital going into these companies, I mean, Dario talked about it on the Door Cash podcast, and Door Cash was like, "Why don't you just buy 10 X more GPUs?"
139
00:16:40.780 --> 00:16:45.840
And he's like, "Because I'm gonna go bankrupt if the model doesn't exactly hit the, the performance level."
140
00:16:45.860 --> 00:16:56.460
[laughs] How do you think about that also as a risk on, you know, you guys are investors in OpenAI and Thinking Machines and World Apps. It seems like we're leveraging the scaling laws at a pretty high rate. Right.
141
00:16:56.500 --> 00:17:04.120
Like, how comfortable, I guess, do you feel with the downside scenario? Like, and say, like, things peter out, you think you can kind of like- Okay...
142
00:17:04.140 --> 00:17:07.980
restructure, uh, these build outs and, uh, you know, capital investments? Yeah.
143
00:17:08.020 --> 00:17:17.020
So I should start by saying, so I lived through the dot-com crash, um, and I can tell you stories for hours about the dot-com crash, and it was horrible. No, it was awful. It was, it was, it was apocalyptic.
144
00:17:17.180 --> 00:17:21.880
Uh, by the way, the, a lot of the dot-com crash was actually, at the time, it was actually a telecom crash. It was a bandwidth crash.
145
00:17:22.120 --> 00:17:29.100
Like, the, the thing that actually crashed that wiped out all the money was the tele- the telecom companies. Global crossing. Global, global, yeah. So I'm from Singapore, and- Oh...
146
00:17:29.120 --> 00:17:32.180
they, they laid so much cable o- over, over our oceans.
147
00:17:32.420 --> 00:17:40.840
Actually, there [laughs] was a scaling law in the dot-com era, and it was literally the c- the US Commerce Department put out a report in 1996, and they said internet traffic was doubling every quarter.
148
00:17:41.660 --> 00:17:46.560
Um, and, and actually in 1995 and 1996, internet traffic actually did double every quarter, and so that became the scaling law.
149
00:17:46.580 --> 00:17:52.460
And so what all these telecom entrepreneurs did was they went out and they raised money to build fiber- Mm... anticipating that the demand for bandwidth was gonna keep doubling every quarter.
150
00:17:53.260 --> 00:17:58.740
Doubling every quarter though is like, you know, grains of chess on the chessboard. Like, at some point, the numbers- Right... become extremely large, right?
151
00:17:58.780 --> 00:18:06.320
And, and, and it really, and really what happened was the internet c- the internet, by the way, continuously kept growing basically since inception. It is, you know, it's, it's continuously grown. It's never shrunk.
152
00:18:06.840 --> 00:18:12.720
And it's growing really fast compared to anything else- Mm-hmm... you know, in, in, in human history. But it wasn't doubling every quarter as of 1998, 1999. Right.
153
00:18:12.740 --> 00:18:20.300
And so there was this gap in the expectation of what they thought was a scaling law versus reality, and that's actually what caused the dot-com crash, which was the, it, they, they way over...
154
00:18:20.540 --> 00:18:23.300
Companies like Global Crossing way overbuilt fiber, which is sort of the...
155
00:18:23.340 --> 00:18:30.220
It, and by the way, fiber telecom equipment, you know, so all the, all the networking gear, you know, and then, and then by the way, the actual physical data centers.
156
00:18:30.320 --> 00:18:38.880
Like, that was the beginning of the, of the, of the data center build and then, and then data center overbuild. And so you had that, but it was, it was literally, I think it was, like, $2 trillion got wiped out, right?
157
00:18:38.959 --> 00:18:41.720
It was like- Jesus Christ... it, it, it was like a big... It was...
158
00:18:41.740
--> 00:18:50.580
And, and by the way, the other, the other subtlety in it was the internet companies themselves n- never really had any debt, 'cause tech- tech companies generally don't run on debt. But the telecom companies run on debt.
159
00:18:50.620 --> 00:18:57.720
Physical infrastructure companies run on debt. And so the companies like Global Crossing not just raised a lot of equity, they also raised a lot of debt, so they're highly levered. And so then you just do the thing.
160
00:18:57.760 --> 00:19:05.800
It's just like, okay, you have a highly levered thing where you're, you're just over- Mm-hmm... you're overbuilding capacity. Demand is growing, but not as fast as you hoped, and then boom, bankrupt, right?
161
00:19:05.840 --> 00:19:12.960
And, and then it, and then it's like they say about the hotel industry, which is it's always the third owner of a hotel that makes money. [laughs] Right? It, it has to go bankrupt twice, right?
162
00:19:13.000 --> 00:19:17.800
You have to wash out all of the over-optimistic exuberance before it gets to actually a stable state, and then it makes money.
163
00:19:18.340 --> 00:19:26.080
So by the way, all of those data centers and all of those, all the fiber that, uh- They're in use... it, it's all in use today [laughs]- Yeah... but 25 years later. But it, it took...
164
00:19:26.160 --> 00:19:34.720
And actually the elapsed time was it took 15 years. It took 15 years from 2000 to 2015 to actually fill, fill up all that capacity. The cautionary warning is the, the overbuild can happen.
165
00:19:35.060 --> 00:19:50.160
Um, and, and, and, and, you know, you, you get into this thing where basically everybody, everybody who basically has any sort of institutional capital is like, "Wow, it's just I, I don't know how to invest in these crazy software things, but for sure I can put, build data centers, and for sure I can buy GPUs and I can deploy, you know, compute grids and, a- a- and all these things."
166
00:19:50.700 --> 00:19:58.760
Um, and, and so, y- you know, y- if you're a pessimist, you can look at this and you can say, "Wow, this is, like, really set up to be able to basically replicate, you know, what we went through, what we went through in 2000."
167
00:19:58.860 --> 00:20:04.900
Obviously that would be bad. The counterargument, which is the one I, I agree with, which is the counterargu- on the other side, is a couple things.
168
00:20:05.120 --> 00:20:11.400
One is the companies that are investing all the m- the, the companies that are investing the money are, like, the bluest chip of companies. And so ba- back, back in the, in the dot...
169
00:20:11.580 --> 00:20:20.360
Like, Global Crossing was like an, it was like an entrepreneur, it was like a, a new venture. But, like, the money that's being deployed now at scale is Microsoft and, you know, and Amazon and Google. Facebook. Right.
170
00:20:20.400 --> 00:20:26.220
And Facebook and Nvidia and, you know, these, the, the, these- Yeah... and, and now, you know, by the way, OpenAI and Anthropic, which are now at, like, you know- Mm-hmm...
171
00:20:26.240 --> 00:20:35.960
really serious size, um, you know, as companies with, you know, very serious revenue. These are very large scale companies with, like, lots, lots of cash, lots of debt capacity that they've, they've never used.
172
00:20:35.980 --> 00:20:39.180
And so th- this is institutional in a way that that really wasn't at the time.
173
00:20:39.740 --> 00:20:46.260
And then the other is, at least for now, every dollar that's being put into anything that results in a running GPU is being turned into revenue right away.
174
00:20:46.900 --> 00:20:49.740
Like, so, a- and you guys know this, like everybody's starved for capacity.
175
00:20:49.800 --> 00:20:56.000
Everybody's starved for compute capacity, and then, you know, all the associated things, memory and, and, and interconnect and everything else, um, data center space.
176
00:20:56.040 --> 00:20:59.520
And so e- every dollar right now that's being put in the ground is turning into revenue.
177
00:20:59.540 --> 00:21:09.920
And, and it, and in fact, I actually think there's an interesting thing happening, which is because everybody's starved for capacity, the models that we actually have that we can use today are inferior versions of what we would have if not for the supply constraints.
178
00:21:10.600 --> 00:21:17.260
Um- It's true... if, if, right. Suppose a hypothetical universe in which GPUs were 10 times cheaper and 10 times more plentiful- Mm-hmm... the models would be much better- Of course...
179
00:21:17.260 --> 00:21:24.480
'cause you would just allocate a lot more money to training, and you'd just build better models, and they would be better. Um, and so we're, we're actually getting the sandbag version of the technology. [laughs] Yeah.
180
00:21:24.500 --> 00:21:35.020
No, everything we use is quantized- Right... because the, the labs have to keep the, the full versions. Right. Like... [laughs] We're not even getting the good stuff. Yeah. But, but getting the good stuff is, is just...
181
00:21:35.060 --> 00:21:46.760
E- even if technical progress stops, once there's, like, a much bigger build of, like, GPU manufacturing capacity and memory, you know, all, all the things that have to happen in the course of the next 5 or 10 years, once it happens, even the current technology is gonna get, gonna get much better.
182
00:21:47.240 --> 00:21:52.020
And then as you know, like, there's just, like, a million ways to use this stuff. Like, there's just, like, a million use cases for this. Mm-hmm.
183
00:21:52.040 --> 00:21:56.440
Like, it, it, you know, this isn't just sending packets across a, a thing whatever and hoping that people find something to do with it.
184
00:21:56.480 --> 00:22:03.700
This is just like, oh, we apply intelligence into every domain of human activity- Yeah... and then it works, like, incredibly well. Yeah. Um, here's what I know. Here's
what I know.
185
00:22:04.240 --> 00:22:12.260
Um, in the next three or four year, it's like somewhere between three or four years out, basically everything is selling out. So, like, the, the entire supply chain is, is, is, is sold out or, or, or selling out.
186
00:22:12.300 --> 00:22:17.600
And so there, there's no, like... It, the, I really we're just gonna have, like, chronic supply shortage for, you know, for years to come.
187
00:22:18.180 --> 00:22:26.920
Um, there's going to be a response from the market that's gonna result in an enormous, you know, it's happening now, an enormous flood of investment in a new fab capacity and ev- you know, every- everything else to be able to do that.
188
00:22:26.960 --> 00:22:31.220
At some point, the supply chain constraints will unlock, you know, at least to some degree.
189
00:22:31.280 --> 00:22:38.020
That will be another accelerant to industry growth when that happens, 'cause the products will get better and everything will get cheaper. Um, and so, so I know that's gonna happen.
190
00:22:38.120 --> 00:22:42.280
I know that, you know, the deployments, you know, the, the actual use cases are, like, really compelling.
191
00:22:42.320 --> 00:22:49.980
And then, like I said, you know, with reasoning and agents and so forth, like I know they're just gonna get, like, much, much better from here. And so I, I know the capabilities are, like, really real and serious.
192
00:22:50.100 --> 00:22:56.300
I also know that the technical progress is not going to stop. It, it, it is accel- it is, is accelerating. Like, the breakthroughs are, are tremendous.
193
00:22:56.440 --> 00:22:58.560
I mean, even just month over month, the breakthroughs are really dramatic.
194
00:22:59.220 --> 00:23:08.700
And so, you know, I think if you were a cynic, and there, there are cynics, you can look at 2000 and you can find echoes, but I can't even imagine betting on that this is gonna, like, somehow disappoint in, you know, at least for years to come.
195
00:23:09.420 --> 00:23:17.140
I think it would be essentially suicidal to make that bet. Yeah. Um, it was, uh, Michael Burry. Uh, uh- That's an interesting guy, huh? We'll, we'll pi- yeah. We'll pick on a guy. We'll pick, let's pick on one guy. Yes.
196
00:23:17.160 --> 00:23:22.280
[laughs] We'll pick... Well, 'cause he did. He, he came out with a... Was it, was the- He doesn't mind. Yeah, yeah... it was the Nvidia short, right? Yeah, yeah, yeah. He came out with the Nvidia short.
197
00:23:22.640 --> 00:23:35.740
And then d- you guys probably talked about this, which is the, the analysis now that, like, the current models are getting better faster at such a rate that if you are running an Nvid- if you're running an Nvidia inference chip today that's three years old, you're making more money on it today than you did three years ago.
198
00:23:35.840 --> 00:23:43.400
Mm-hmm. Yep. Because the pace of improvement of the software is, is faster than the, than the depreciation cycle of the chip. And then my understanding is Google is running... I don't think, I don't know exactly what...
199
00:23:43.800 --> 00:23:51.280
I, these are rumors that I've heard, or maybe it's public, but, um, I think Google's running very old TPUs very profitably. For inference, yeah. And very profit- Absolutely... and, and very profitably. Yeah.
200
00:23:51.380 --> 00:23:57.080
Um, and so, so it actually turns out, as far as I can tell, it's actually the opposite of the Burry thesis is actually... He was actually 180 degrees wrong.
201
00:23:57.140 --> 00:24:02.820
It's actually the, the, the, the old Nvidia chips are getting more valuable, which is something that's, like, literally never happened before.
202
00:24:02.940 --> 00:24:08.340
[laughs] Like, it's never been the case that you have an older model chip that becomes more valuable, not less valuable. Yeah.
203
00:24:08.360 --> 00:24:15.840
And, and again, that's an expression of the just in- ferocious pace of software progress, ferocious pace of capability payoff- Yeah... uh, that you're getting on the other side of this.
204
00:24:15.880 --> 00:24:26.528
And so I just The idea of betting against that, like- Yeah. One, one of my- It's like an invitation to get your face ripped off. [laughs] One of my early hits was, like, modeling the lifespan of the H100 and H200s.
205
00:24:26.808 --> 00:24:35.448
And, and going like, you know, usually they have eyes like four to seven years, and it was, you know, maybe you sort of realistically haircut it down to two to three. Yeah. But actually it's going up and not down.
206
00:24:35.448 --> 00:24:42.408
[laughs] Yeah. And, and, uh, that's-- I mean, that's, I think that's the dream. Uh, we are finding utilization, and I think utilization solves all problems. Like you can- Yeah...
207
00:24:42.428 --> 00:24:52.108
you can find use, use cases for even, like the poor... Like, even memory we're having a shortage, right? And, and e- even like the, the shittier versions of, of memory that we do have, we are finding use cases for it.
208
00:24:52.348 --> 00:25:00.668
Right. So, like, that's great. Yeah. [laughs] How, how important is open source AI and kind of like edge inference in a world in which you have three years of supply crunch? Like, do you think in the...
209
00:25:00.708 --> 00:25:08.008
Like, you know, if you fast-forward like five years, like how do you think about inference, uh, in the data center versus at the edge? Well, so just to start, yeah.
210
00:25:08.068 --> 00:25:12.448
So I think, I think open source is very important for a bunch of reasons. I think edge, edge inference is very important for a bunch of reasons.
211
00:25:12.568 --> 00:25:17.008
I, I think just practically speaking, if we're just gonna have fundamental constr- uh, co- so supply crunches for the next...
212
00:25:17.068 --> 00:25:20.868
I mean, you, you guys know, if you just project forward demand over the next three years- Right. Yeah...
213
00:25:20.908 -->
00:25:29.108
rel- relative to supply, one of the dismaying predictions you can do is what's gonna, what, what's gonna happen to the cost of, of inference in the core, uh, over the next three years, and like it may rise dramatically, right?
214
00:25:29.148 --> 00:25:35.028
Like, so, so what i- and then as, as you know, like the, the mo- the big model competitors are subsidizing heavily right now. Right. Right. Yeah. And so, so what's the...
215
00:25:35.528 --> 00:25:43.248
What will be the average person's, you know, per day, per month token cost, you know, three years from now to do all the things that they want to do? And I, I don't know what it's gonna be.
216
00:25:43.308 --> 00:25:52.008
I mean, I have-- you guys probably have friends. I have friends today who are paying $1,000 a day for OpenClau- for cloud tokens to run OpenClau, right? And so, okay, $30,000 a month, right? Damn.
217
00:25:52.008 --> 00:25:55.888
And, and by the way, tho- those friends have like 1,000 more ideas of the things that they want their cloud to do, right? Yeah.
218
00:25:55.928 --> 00:26:03.588
And so you, you could imagine there, there's like latent demand of up to, I don't know, five or $10,000 a day of, of, of tokens for a fully deployed, you know, person- personal agent.
219
00:26:04.208 --> 00:26:09.968
Uh, and obviously consumers can't pay that, right? And so, so... But it gives you a sense of the fu- of the fu- of the future scope of demand, right?
220
00:26:10.008 --> 00:26:18.748
And so, so even, even if there's a 10X improvement in price performance, that still, you know, goes to $100 a day, which is still way beyond what people can pay. Mm-hmm. So there's just gonna be like ferocious demand.
221
00:26:18.788 --> 00:26:24.108
By the way, the agent thing, the other interesting thing is I think the agent thing... So up until now, a lot of the constraints have been G- GPU constraints.
222
00:26:24.168 --> 00:26:32.588
I think the agent thing now also translates into CPU constraints, right? [laughs] CPU and memory, yes. CPU and memory, right? And so like the entire chip ecosystem is just gonna get- Wait, wait for network constraints.
223
00:26:32.648 --> 00:26:38.228
That, that will be the killer. [laughs] It's all bottlenecked potentially for years. And so, so I, I think that Brad... And, and I think it's actually possible.
224
00:26:38.628 --> 00:26:44.948
I mean, generally inference costs are gonna keep coming down. But I think the-- let's put it this way, the rate of decline I think may level out here for a bit- Yeah... because of these comp- supply constraints.
225
00:26:45.348 --> 00:26:49.268
And then at some point, maybe the lab stops subsidizing so much, and that, that, that again will be a, be an issue.
226
00:26:49.848 --> 00:26:55.648
And so there's just gonna be so much more demand for inference than, than can be satisfied, um, you know, kind of w- with the centralized model.
227
00:26:55.708 --> 00:27:02.808
And then, and then you know, you guys know this, but like all the just the dramatic, I mean, just the dramatic innovations that have happened in the Apple silicon to be able to do, uh, inferences is quite amazing.
228
00:27:03.208 --> 00:27:06.648
The level of effort being put, like the open source guys are putting incredible effort into getting, you know...
229
00:27:07.088 --> 00:27:15.128
Th- th- this recurring pattern where the big model will never run on a PC, and then six months later- Mm-hmm. Right... it runs on a PC, right? It's like amazing. And there's very smart people working on that.
230
00:27:15.308 --> 00:27:19.308
So there's all that. And then look, there's also, you know, there's also like other, there's other motivators.
231
00:27:19.348 --> 00:27:23.648
There's other motivators, which is just like, okay, how much trust are the big centralized model providers, you know...
232
00:27:24.348 --> 00:27:32.508
H- how much trust are they building in the market versus, you know, how much are, you know, at least for in certain cases with some people for certain use cases, people being like, "Well, I'm not willing to just like turn everything over."
233
00:27:33.068 --> 00:27:40.868
Mm-hmm. So there, there, there's all the trust issues. Um, by the way, there's also just like straight up price optimization. There's many uses of AI where you don't need Einstein in the cloud. Right.
234
00:27:40.888 --> 00:27:49.588
You just need like a, a, a, a smart local model. There's also performance issues where you wanna hi- you know, you, you want, you know... You're gonna want your door knob to have an AI model in it- Right...
235
00:27:49.608 --> 00:27:57.408
you know, to be able to, you know, do, um, you know, to be able to do access control. Um, obviously, like everything with a chip is gonna have an AI model in it- Mm-hmm... and, and a lot of those are gonna be local.
236
00:27:58.068 --> 00:28:03.748
Um, and so yeah, no, like I think, I think you're gonna have ti... And then you're gonna... By the way, also wearable devices, you know, you don't wanna do a complete round trip.
237
00:28:03.788 --> 00:28:10.388
You want, you know, your, whatever your smart devices are, you want it to be like super low latency. Yeah. Go ahead. The question, do we care who makes it? Yeah.
238
00:28:10.428 --> 00:28:21.388
One of the biggest news this week was the collapse of AI2, the Allen Institute. Mm-hmm. One of the actual American open source model labs. Yeah. Um, and, uh, I'm not o- that optimistic on, on American open source. Yeah.
239
00:28:21.408 --> 00:28:29.968
Like, you, you guys invested in Mistral, and Mistral is doing extremely well. Uh, outside of China, that's about it. Yeah, we'll see. [laughs] We'll see. I... Look, I... Number one, I do think we care.
240
00:28:30.028 --> 00:28:38.228
I, I do think we, I do think we care who makes it. Um, I would say this, the, the, the previous presidential administration wanted to kill it in the US. Oh, yeah. Like they, they, they wanted to
drown it in the bathtub.
241
00:28:38.908 --> 00:28:45.688
Um, and so they wanted to kill it. So at least we have a government now that actually, like actually wants it, wants it to happen. And you're on the council at PCAST? Yeah, it's in the new... In PCAST, yeah.
242
00:28:45.728 --> 00:28:54.208
So the, the, you know, this admin- for whatever other political issues people have, which are many, you know, this administration has, I think, a very enlightened view, and in particular an enlightened view on AI, and in particular on open source AI.
243
00:28:54.828 --> 00:29:02.528
Uh, and so they're very supportive. Um, my read is the Chi- the Chinese have a very sp- the various Chinese companies have a very specific reason to do open source, which is they, they, they don't...
244
00:29:02.748 --> 00:29:10.008
Fundamentally, they don't think they can sell commercial, uh, AI outside of China right now, and or at least specifically not, not in the US for a combination of reasons.
245
00:29:10.048 --> 00:29:18.288
And so they, they kind of view, I think, open source AI as a bit of a loss leader against basically domestic, uh, you know, paid, paid services, and then kind of, a- you know, kind of anc- ancillary products.
246
00:29:18.788 --> 00:29:24.528
You know, they're, they're very excited about it. By the way, I think it's great. I think it's great that they're doing it. Um, you know, I think DeepSeek was like a gift to the world.
247
00:29:24.828 --> 00:29:34.048
Um, I think the great thing about open source, open source, the, the, the impact of open source is felt two ways. One is you, you get the software for free, but the other is you get to learn how it works, right?
248
00:29:34.088 --> 00:29:40.608
And so like- The paper... the paper. The paper, and, and the code. Right, and the code. And so like for example, I thought this was amazing.
249
00:29:40.648 --> 00:29:44.828
So OpenAI comes out with o1, and it's an amazing technical breakthrough, and it's just like absolutely fantastic.
250
00:29:45.188 --> 00:29:49.664
But of course, they don't explain how it works in detail, and then of course they hide the, they hide the reasoning traces. Right?
251
00:29:49.704 --> 00:29:56.244
And, and then, and then, and then everybody's like, "Okay, this is great, but like who's gonna be able to replicate this? Are other people gonna be able to do this? You know, is there secret sauce in there?"
252
00:29:56.264 --> 00:30:04.024
And then R1 comes out, and it's just like there's the code and there's the paper. [laughs] And now the whole world knows how to do it. And then, you know, three months later, every other AI model is, is adding reasoning.
253
00:30:04.044 --> 00:30:06.024
And so, so you get this kind of double...
254
00:30:06.064 --> 00:30:15.004
Like, even if the Chinese models themselves are not the models that get used, the education that's taken place to the rest of the world, the information diffusion, you know, is incredibly powerful. So that
happens.
255
00:30:15.044 --> 00:30:22.404
And then I don't know. We'll, we'll see. You know, there are a bunch of American, you know, open source, you know, AI, uh, model companies. I mean, look, there's gonna be tremendous... You know, there already is.
256
00:30:22.444 --> 00:30:27.184
There's, you know, there's gonna be treme- there's tremendous competition, uh, among the primary model companies.
257
00:30:27.264 --> 00:30:35.624
You know, there's, depending on how you count, there's like four or five, you know, big co-model companies now that are, you know, kind of neck and neck, uh, in different ways.
258
00:30:35.864 --> 00:30:44.004
Um, uh, you know, and, and, and, um, you know, and then obviously, both, both X and then Meta where I'm involved are, you know, both have huge, you know, huge attempts to, you know, kind of, to kind of k- leapfrog our way.
259
00:30:44.484 --> 00:30:49.904
And then you've got, you know, a whole fleet of startups, new companies, including a whole bunch that we're backing that are, you know, k- trying to k- come out with different approaches.
260
00:30:49.944 --> 00:30:58.544
And then you've got whatever it is. I don't know. How, how many, how many like mainline foundation model companies are there in China at this point? It's probably six- There's five tigers- Five... is what they call it.
261
00:30:58.744 --> 00:31:05.584
Yeah. Uh, Qwen is in- in questionable because- Right... there's change in leadership. Right. Yeah. But that, does that include... That, that includes, like, Moonshot? Yes. Okay, yeah.
262
00:31:05.604 --> 00:31:17.324
Kimi, DeepSeek, um, uh, ZAI- Right... um, Qwen, o1 is in there. Right. And then, um, and ByteDance. And then you can see- ByteDance would be like the next tier. ByteDance. They, they weren't as prominent. They weren't.
263
00:31:17.324 --> 00:31:20.784
They don't have a leading model yet. But now, you know, it's... Yeah, but they're, you know, at least, you know, C- Yeah... CDance is very inspiring- Yeah...
264
00:31:20.784 --> 00:31:23.904
and presumably they have more stuff coming, and Tencent probably has more stuff coming- Yeah... and, and so forth.
265
00:31:23.944 --> 00:31:29.344
And so, so, and so like, look, here, here would be a thing you could anticipate, which is there are not, these markets, there are not going to be...
266
00:31:29.404 --> 00:31:36.824
Between the US and China right now, there's like a dozen primary foundation model companies that are, like, at scale, at, at some level of, like, critical mass. It's not gonna be a dozen in three years, right?
267
00:31:36.944 --> 00:31:42.764
Like, it, just because these industries don't bear- Yeah... a dozen. It's, it's gonna be three, you know, there's gonna be three or four big winners, or maybe one or two big winners.
268
00:31:43.184 --> 00:31:48.904
And so there's gonna be like a whole bunch of those guys that are gonna have to figure out alternate strategies. Um, and I think, like, open source is one of those
strategies.
269
00:31:48.924 --> 00:31:56.964
And so I, I think you could see, like, a whole... I, I, I think the question's like who's gonna do open source. I think that could change really fast. I, I think that, that, that's a very dynamic thing.
270
00:31:57.004 --> 00:32:01.164
I think it's very hard to predict what happens. And, and I think it's very important. NVIDIA's doing a lot.
271
00:32:01.244 --> 00:32:08.504
You mentioned- Well, I was gonna say, well, exactly, and then you got NVIDIA, and then, and then, you know, just again in industrial, you know, there's, there's an old thing in business strategy, which is called, uh, commoditize- Commoditize your complements...
272
00:32:08.524 --> 00:32:12.964
commoditize the complement. And it's right. And so if you're Jensen, it's just kind of obvious, of course you want to commoditize the software. Yeah.
273
00:32:13.004 --> 00:32:20.064
And he's, and to his enormous credit, he's putting enormous resources behind that. And so maybe it, maybe it's literally NVIDIA, and I think that would be great. Yeah. Yeah.
274
00:32:20.164 --> 00:32:31.704
Uh, narrative violation to European projects, uh, in the building. [laughs] Uh... Bam. I, I'm hosting my, uh, Europe, uh, conference soon, and I got both of them. They got us. They got us, Mark. You're finished. Okay.
275
00:32:31.724 --> 00:32:37.504
They got us. Well, wait a minute. Where was Peter? So where was Steinberger when he did all this? He was in Vienna. Austria. Yeah, yeah, yeah. He was in what? He was in Vienna. Oh, he was in Vienna. Yeah.
276
00:32:37.524 --> 00:32:44.444
And then where is he now? Uh, he's moving to SF. [laughs] Okay. Okay. All right. Okay. There we go. And then, yeah, the Pi guy, right, the Pi guys are European. Yeah, they're also there, yeah.
277
00:32:44.444 --> 00:32:52.064
Uh, they have some buddies in Austria. Ma- Mario is also there, yeah. Right. And are they... Yeah, they haven't announced yet any sort of change- Yeah... change to... Or have they? No, they're, they have a company there.
278
00:32:52.344 --> 00:33:02.784
Okay, got it. In Austria. Okay, good. Yeah, yeah. Good. Um... Yeah. Good. Anyways, I think Pi and OpenClou have very important software things. Yeah. And, and I just wanted you to just go off on [laughs] what you think.
279
00:33:03.064 --> 00:33:10.404
Yeah. So I think in co- the, the combination of the two of them, I think, is one of the 10 most important- Yeah... software breakthroughs. OpenClou got all the attention, but- Right... talk about Pi.
280
00:33:10.424 --> 00:33:12.964
Pi, Pi is kind of the... Yeah. Pi's, Pi's kind of the architectural breakthrough.
281
00:33:13.504 --> 00:33:19.864
For those of us who are older, there was this whole thing that was very important in the world of software, basically from like 1970 to, I don't know.
282
00:33:19.904 --> 00:33:27.164
It, it still is very important, but like 19, from 1970 through to, like, basically the creation of Linux, which is basically
this, this thing we used to call, like, the Unix mindset. Like, so, so...
283
00:33:27.544 --> 00:33:29.224
'Cause there were all these different, you know, theories.
284
00:33:29.264 --> 00:33:34.504
There are all these different operating systems and mainframes and, and then, you know, all these Windows and Mac and all these things, and then there was this, uh, uh...
285
00:33:34.524 --> 00:33:40.224
But kind of behind it all was this idea of kind of the Unix mindset. And the Unix mindset was this thing where basically you don't have these, like...
286
00:33:40.764 --> 00:33:45.644
Like, in the old days, like, like, the operating system that, like, made the computer industry really work, like in the 1960s- Mm-hmm...
287
00:33:45.704 --> 00:33:53.284
was this thing called OS360, which was this big operating system that IBM developed that was supposed to basically run everything, and it was this, like, giant monolithic architecture in the sky.
288
00:33:53.344 --> 00:34:00.064
It was like a, you know, it was like a giant castle, um, of software. And, and by the way, it worked really well, and they were very successful with it, but, like, it was this huge castle in the sky.
289
00:34:00.104 --> 00:34:06.784
But it was this thing that was almost unapproachable, which was like you had to be kind of inside IBM or very close to IBM, and you had to really understand every aspect of how the system worked.
290
00:34:06.823 --> 00:34:17.084
And then the, the Unix guys originally out of AT&T and then out, out of Berkeley, um, you know, came out and they said, "No, let's have a completely different architecture, and the way architecture is gonna work is we're gonna have, we're gonna have a, a prompt and a, and a, and a shell.
291
00:34:17.504 --> 00:34:22.684
And then, and then we're gonna... All the functionality is gonna be in the form of these discrete modules, and then you're gonna be able to chain the modules together." Mm-hmm. Yeah.
292
00:34:22.704 --> 00:34:26.864
And so, like, the, the, the op- it's almost like the operating, its operating system itself is gonna be a programming language.
293
00:34:27.524 --> 00:34:38.864
Um, and then that led, led to the, the, the sort of centrality of the shell, um, and then that led to sort of, uh, you know, b- basically chaining the other Unix tools, and then that led to the emergence of these, these scripting languages like Perl, where you, you could basically kind of very easily do this.
294
00:34:39.504 --> 00:34:45.344
And then the shells got more sophisticated. And then, and then, and then look, like, you know, that, that, that... Number one, that worked, and that, that was the world I grew up in.
295
00:34:45.384 --> 00:34:53.224
Like, I was, I was a Unix guy, you know, sort of from, call it 1988 to, you know, kind of all, all the way through my work. And it worked really well. It, it's in the background.
296
00:34:53.324 --> 00:35:01.264
Um, you know, nor- normal people don't need to ne- didn't need to necessarily know about it. But, like, if you were doing, like, system architecture or application development, you, you, you knew all about it.
297
00:35:01.884 --> 00:35:09.104
Um, and then, you know, it's been in the background ever since. And, you know, look, your Mac still has a Unix shell, you know, kind of in there, and your iPhone still has a Unix shell kind of buried in there somewhere.
298
00:35:09.524 --> 00:35:15.744
So they're kind of in there. And then, you know, the Windows shell is kind of a, you know, sort of a weird derivative of that. Mm-hmm. But, um, you know, but look, the inter- the internet runs on Unix.
299
00:35:15.984 --> 00:35:23.984
Um, and then smartphones. Actually, both iOS and Android are Unix derivatives, and so, you know, kind of Unix did end up winning. But, but anyway, and then we just started taking that for granted. And then, and then...
300
00:35:24.024 --> 00:35:31.884
So, so basically the, the way I think about what happened with Pi and then with OpenClou is basically what those guys figured out is, I always say the, the great breakthroughs are obvious in retrospect, right?
301
00:35:31.924 --> 00:35:39.164
Which is- The best kind... the best kind. They weren't obvious at the time, or somebody else would have done them already. Um, and so there is a, like a real conceptual leap.
302
00:35:39.184 --> 00:35:46.204
But then you look at it sort of backwards looking, and you're just like, "Oh, of course." Mm-hmm. Like, the, the, to me, those are always the best breakthroughs. Well, actually, language models themselves are like that.
303
00:35:46.244 --> 00:35:52.424
It's just like, "Oh, next token completion. Oh, of course." Yeah. What other objective mattered? [laughs] Yeah, exactly. But, but like it... Right. But you see what I'm saying.
304
00:35:52.484 --> 00:35:57.284
It wasn't obvious until somebody actually did it, right? And so the conceptual breakthrough is real and deep and powerful and, and very important.
305
00:35:57.744 --> 00:36:07.704
And so the way I think about Pi and OpenClou is it's basically marrying the, the language model mindset to the Uni- to the Unix basically shell prompt mindset. And so it's, it's basically this idea that what, what...
306
00:36:07.744 --> 00:36:09.304
So what is an agent, right?
307
00:36:09.344 --> 00:36:16.124
And as, as, and as you know, like, many smart people have been trying to figure out what an agent is for, for, for decades, and they've had many architectures to build agents and the whole thing, and it turns out, what is an agent?
308
00:36:16.184 --> 00:36:22.484
So it turns out what we now know is an agent is the following. It's, so it's a language model, and then above that, it's a batch, it's a batch shell.
309
00:36:22.824 --> 00:36:30.476
Um, so it's a, it's a Unix shell And then as, and then the agent has access, uh, has access to, to the shell and, you know, hope-hopefully in a sandbox. Maybe, maybe in a sandbox.
310
00:36:31.176 --> 00:36:36.556
So
it's, it's the model, um, it's the shell, um, and then it's a f- it's a file system. Um, and then the state is stored in files.
311
00:36:37.096 --> 00:36:42.496
And then, you know, there's the markdown format for the, you know, for, for the files themselves, and then, and then there's basically what in Unix is called a cron job.
312
00:36:42.536 --> 00:36:46.536
There's a loop, and then there's a heartbeat- There's a heartbeat... for the, there's a heartbeat and, and the thing basically wake, wakes up, wakes up.
313
00:36:47.296 --> 00:36:52.736
So it's basically LLM plus shell plus file system plus markdown plus cron. And it turns out that's an agent.
314
00:36:52.936 --> 00:37:00.956
And, and, and every part of that other than the model is something that we already completely know and understand. And in fact, it turns out the, like the latent power of the Unix shell is like extraordinary. Hmm.
315
00:37:01.076 --> 00:37:07.136
Because basically like all... Like, there's just like an, there's just enormous latent power in the shell. There's enormous numbers of Unix commands.
316
00:37:07.156 --> 00:37:13.556
There's enormous number of command line interfaces into all kinds of things already in the, you know, your entire... I mean, your entire... Just to start with, your computer runs on a shell.
317
00:37:13.996 --> 00:37:21.056
If you're running a Mac or a, or, or a phone, your computer, your computer's running on a shell, uh, already. And so like the full power of your computer is available at the, at the command line level.
318
00:37:21.336 --> 00:37:29.516
Um, and then it turns out it's really easy to expose other functions as a command line interface. And so like the, this whole idea where we need like MCP and these like proto- Mm-hmm... fancy protocols, whatever.
319
00:37:29.535 --> 00:37:36.856
It's like, no, we don't. We just need like a command, command line thing. So that's the architecture. And then it turns out, what is your agent? Your agent is a bunch of files [chuckles] stored in a file system.
320
00:37:36.916 --> 00:37:46.396
And then there's the thing that just like completely blew my mind when I wrapped my head around it as a result of this, which is like, okay, this means your agent is now actually independent of the model that it's running on because you can actually swap out- Mm...
321
00:37:46.436 --> 00:37:53.636
a different LLM underneath your agent, and your, your agent will change personality somewhat 'cause the model is different, but all of the state stored in the files will be retained. Yeah.
322
00:37:53.676 --> 00:37:58.196
Different instruction set, but you just compile that. Right. Exactly. And it's all right. It was like, right, swapping out a chip and recompiling.
323
00:37:58.576 --> 00:38:04.596
But it's, it's still, it's still your agent with all of its memories, um, and with all of its capabilities. And then by the way, you can also swap out the shell.
324
00:38:04.916 --> 00:38:11.236
Uh, so you can move it to a different execution environment that is also, is also a bash shell. By the way, you can also switch out the file system, right? Yeah.
325
00:38:11.236 --> 00:38:16.696
Uh, and you can, and you can sw- and you can swap out the, the, the heartbeat for the, the cron framework- Mm... the, the loop, the, the agent framework itself.
326
00:38:16.736 --> 00:38:23.436
And so your agent basically is, ba- basically at the end of the day, it's just, it's just its files. Um, and then, and then there's this person- It's your open claw. [laughs] Yeah.
327
00:38:23.496 --> 00:38:29.796
It's, it's basically, it's just the files. Um, and then by the way, as a consequence of that, the agent its- and then the agent itself, it turns out a couple important things.
328
00:38:29.816 --> 00:38:32.596
So one is it, it's, it, it can migrate itself, right?
329
00:38:32.635 --> 00:38:40.536
And so you're, you, you can instruct your agent, migrate yourself to a different, uh, runtime environment, migrate yourself to a different file system, migrate yourself to a different, you know, like swap out the language model.
330
00:38:40.576 --> 00:38:46.776
Your agent will do all that stuff for you. And then there's the final thing, which is just amazing, which is the agent is, the agent actually has full introspection.
331
00:38:46.816 --> 00:38:50.656
It actually, it actually knows about its own files, and it can rewrite its own files, right?
332
00:38:50.716 --> 00:38:59.456
Which, which by the way is basically no widely deployed software system in history where the, the, the thing that you're using actually has full introspective knowledge of how it itself works and is able to modify itself like that.
333
00:38:59.896 --> 00:39:04.876
Yeah. Mm-hmm. That, I mean, there have been toy systems that have had that, but there, there's never been a widely deployed system that has that capability.
334
00:39:05.476 --> 00:39:12.276
And then that leads you to the capability that just like completely blew my mind when I wrapped my head around it, which is you can tell the agent to add new functions- Mm-hmm...
335
00:39:12.336 --> 00:39:19.416
and features to itself, and it can do that. Extend yourself. Yeah. Right. Ex-extend yourself. Like extend yourself. Give yourself a new capability. [laughs] Right?
336
00:39:19.496 --> 00:39:25.716
And so, and so literally it's just like you run into somebody at a party and they're like, "Oh, I have my open claw do whatever, connect to my Eight Sleep bed, and it gives me better advice on sleep."
337
00:39:25.756 --> 00:39:31.816
And you go home at night and you tell your claw, or if they're at the party, by the way, you tell your claw, "Oh, add this capability to yourself," and your claw will say, "Oh, okay, no problem."
338
00:39:32.196 --> 00:39:38.556
And it'll go out on the internet and it'll figure out whatever it needs, and then it'll go out to Cl-Claude Code or whatever, it'll write whatever it needs, and then the next thing you
know, it has this new capability.
339
00:39:39.036 --> 00:39:45.276
And so you don't even have to... Like, you can have it upgrade itself without even having to, without having to do anything other than tell it that you want it to do that.
340
00:39:45.756 --> 00:39:52.336
And so anyway, so the, the combination of all this is just, I mean, this is just like a massive, incredible- [laughs] I mean, it's just incredible.
341
00:39:52.376 --> 00:39:58.116
Like, if I, if I were, if I were 18, like this is 100, this is what I would be spending all of my time on. This is like such an incredible conceptual breakthrough. Yeah.
342
00:39:58.156 --> 00:40:02.556
And again, pe-people are gonna look at it, and they already get this response. People are gonna look at it and they're gonna say, "Oh, well, where's the breakthrough?"
343
00:40:02.596 --> 00:40:06.376
'Cause these, the, all of these components were already known before. Mm-hmm. But, but this is the key.
344
00:40:06.416 --> 00:40:11.856
The key to the breakthrough was by using all these components that were known before, you get all of the underlying capability that's buried in there.
345
00:40:11.876 --> 00:40:19.456
And so all, and so for example, computer use all of a sudden just kind of falls trivial, trivial. Of course it's gonna be able to use your computer. It has full access to the shell, right?
346
00:40:19.476 --> 00:40:26.096
And then, and then you just, you, you give it access to a browser, and then you've got the computer and the browser and, and off and away it goes. And then, then you've got all the abilities of the browser also.
347
00:40:26.196 --> 00:40:34.416
Um- Yeah... and so, and so the capability unlock here is profound. My friends who are, you know, deepest into this are having their claw do like, uh, like literally like 1,000 things in their lives.
348
00:40:34.876 --> 00:40:40.716
They have new ideas every day. They're just like constantly throwing new challenges at the thing. And by the way, it's early and, you know, these are, you know- Mm...
349
00:40:40.735 --> 00:40:49.276
these are prototypes and there's, you know, as you guys know, there's security issues. Ugh, God. And, and so, you know, there's a bunch of stuff to be ironed out, but the, the unlock of capability- Yeah...
350
00:40:49.296 --> 00:40:50.676
is just incredible. Yeah.
351
00:40:50.716 --> 00:40:58.816
And I, I have absolutely no doubt that everybody in the world is gonna, is gonna have at least, you know, a-an agent like this, if not an entire family of agents, and we're gonna be living in a world where- Yeah...
352
00:40:58.856 --> 00:41:07.316
I think it's almost inevitable now that this is the way people are gonna use computers. I was gonna say, for someone who is deeply familiar with social networks, the next step is your claw talking to my claw. Mm-hmm.
353
00:41:07.616 --> 00:41:14.316
Posting on Claw Facebook, uh, posting their jobs on Claw LinkedIn, and post-posting their tweets on Claw XAI or
whatever, you know.
354
00:41:14.476 --> 00:41:23.756
[laughs] Um, I do think that that is how, uh, you know, we, we get into some danger there i-in terms of like alignment and whether or not we want these things to, to, to run.
355
00:41:23.936 --> 00:41:31.456
Um- You guys know about rent a, rentahuman.com? Yeah, rentahuman. Yeah. Yeah, yeah, yeah. I mean, it's Fiverr, it's TaskRabbit- Yeah. Sure, of course... it's, uh- Crowds on demand... Mechanical, Mechanical Turk. Yeah.
356
00:41:31.776 --> 00:41:39.876
But flipped. Yeah. Right? The agent hiring the people. Yeah. Which of course is gonna happen. [laughs] Right. It's obviously gonna happen. I'm curious if you have any thoughts on the engineering side.
357
00:41:40.136 --> 00:41:46.216
So when you build the browser, the internet, you know, just a bunch of mostly plain text file plus some images.
358
00:41:46.336 --> 00:41:53.216
And today the, every website and app is like so complex and like somehow, you know, the browser kept evolving to fit that in.
359
00:41:53.896 --> 00:42:08.368
Are there any design choices that were made like early in the browser and kinda like the internet and the protocols that you're seeing agents similar today say, "Hey, this thing is just not gonna work for like this type of new compute, and we should just- Rip it out right now.
360
00:42:08.928 --> 00:42:15.048
There were a whole bunch, but I'll give you a couple. So one is, um, y- and we didn't, and you know, to be clear, like this, this was not, you know, this was totally different.
361
00:42:15.068 --> 00:42:22.108
We didn't have the capabilities we have today, but because we didn't have, we didn't have the language models underneath this. But, um, we did have this idea that human readability actually mattered a great deal.
362
00:42:22.528 --> 00:42:29.588
Um, and, and, and so, and specifically in those days, it was, it was not so much English language, but it was, there, there was a design decision to be made between binary protocols and text protocols.
363
00:42:30.168 --> 00:42:39.748
And basically every, every, every basically old school systems architect that had grown up between like the 1960s and the 1990s basically said, you know, the internet, it's... What do you know about the internet?
364
00:42:39.788 --> 00:42:47.948
It's star for bandwidth. You, you just, you have these very narrow straws. You know, look, people, when we did the work on Mosaic, like pe- people who had the internet at home had a 14 kilobit modem, right?
365
00:42:48.108 --> 00:42:51.628
So you're, you're trying to like hyper optimize every bit of data- Mm-hmm... that, that travels over the network.
366
00:42:52.128 --> 00:43:00.728
And so obviously, if you're gonna design a protocol like HTTP, you're gonna want it to be a binary, you know, highly compressed binary protocol for maximum efficiency, and you're gonna wanna have it be like a single connection that persists, and you're, you're...
367
00:43:00.768 --> 00:43:05.688
The
last thing you're gonna wanna do is like bring up and tear down new connections. And you definitely, you're not gonna, not gonna want a text protocol.
368
00:43:05.708 --> 00:43:13.468
And so, of course, we said, "No, we actually want to go completely the, the other direction. It's obviously we only want t- text protocols." Uh, by the way, same thing in HTML itself.
369
00:43:13.508 --> 00:43:19.908
We want HTML to be relatively verbose. You know, we want the tags to actually be like human readable. Um, we wanna use the- The most inefficient things possible.
370
00:43:19.908 --> 00:43:24.768
[laughs] Yeah, we wanna do the in- we wanna do the inef- we wanna do the inefficient things. We- You're, you're the original token maxer. Yeah, exactly. Yeah. Yeah. [laughs] Yeah.
371
00:43:24.788 --> 00:43:27.568
Basically, it's just like a- Better life than build. Well, yeah. Well, actually, this was- [laughs]...
372
00:43:27.608 --> 00:43:32.928
this was actually the con- the conscious thing, which basically says just like assume, assume a future of infinite, infinite bandwidth, build for that. Right.
373
00:43:33.388 --> 00:43:45.448
And then basically what it was, is it was a bet that it, it was a bet that if the system w- if the, if the latent capabilities of the system were powerful enough, and that was obvious enough to people, that would create the demand for the bandwidth that would cause the supply of bandwidth to get built, that would actually make the whole thing work.
374
00:43:46.048 --> 00:43:56.588
And then specifically what we wanted was we wanted everything to be human readable, because we, at the engineering level, we wanted people to be able to read the protocol coming over the wire and be able to understand it with their, with their bare eyes without having to like disassemble it or whatever, right?
375
00:43:56.668 --> 00:44:03.888
Or have it converted out of binary, right? And so the, the, the, all the pro, you know, HTTP and everything else were, were, it was always, uh, text protocols. Uh, and the same thing with HTML.
376
00:44:03.928 --> 00:44:08.248
And, and in many ways, some people say that the key breakthrough in the browser was the view source option- Mm-hmm...
377
00:44:08.308 --> 00:44:16.308
um, which is every web page you go to, you could view source, which means you could see how it worked, which means you could teach yourself how to build, right, new, uh, to, to build new web pages. There was that.
378
00:44:16.368 --> 00:44:22.888
So human readability, um, a b- and, and again, human readability in those days still meant technical, you know, specs. You know, now it means English language.
379
00:44:22.928 --> 00:44:32.208
But the, there is an incredible latent power in giving everybody who uses the system the option to be able to drop down and actually understand and see how it's working, and that worked really well for the web, and I think it's working really well for AI.
380
00:44:32.928 --> 00:44:34.568
That was one.
Um, what was the other?
381
00:44:34.668 --> 00:44:43.348
Um, a big part of the idea of web servers was to actually surface the underlying latent capability of the operating system and to be able to surface the, uh, also the underlying latent capability of the database.
382
00:44:43.888 --> 00:44:48.728
'Cause basically, what was a web server? What, what, what, what is a web server fundamentally, architecturally? It's, it's, it's the operating system.
383
00:44:48.768 --> 00:44:56.208
So it's, it's the operating system's ability to, you know, it's running on top of an OS, so it's the OS's ability- Mm-hmm... to manage the file system and do everything else that you wanna do, process everything.
384
00:44:56.668 --> 00:44:59.968
Um, and then of course, a lot of early, you know, a lot, lot of websites are, are front ends to databases.
385
00:45:00.228 --> 00:45:06.848
Um, and so you wanted to, you wanted to unleash the underlying latent power of whether it was an Oracle database or some other, you know, some other Postgres or whatever, whatever it was.
386
00:45:07.388 --> 00:45:14.808
Um, and so a lot of the function of the web server was to just bridge from that internet connection coming in to be able to unlock the underlying power of the OS and the database.
387
00:45:15.368 --> 00:45:19.768
Uh, and again, people looked at it at the time and they were like, "Well, d- is this really, does this really matter? Like, is this important?
388
00:45:19.808 --> 00:45:26.668
Because we've had databases forever, and we've always had, you know, user interfaces for databases, and this is just another user interface for a database." And it's like, okay, yeah, fair enough.
389
00:45:26.788 --> 00:45:34.848
But on the other side of that is just like this is now a much better interface to databases and one that eight billion people are going to use and is going to be like far easier to use and far more flexible.
390
00:45:34.928 --> 00:45:42.828
And, and, and, and you're not just gonna have old databases. Now you have a system where people can actually understand why they wanna build, you know, a million times more database apps than they have in the past.
391
00:45:43.288 --> 00:45:47.888
And then the number of databases in the world exploded. And so again, this goes to this thing of like building, building in layers.
392
00:45:48.488 --> 00:45:54.948
Some of the smartest people in the industry look at any new challenge and they're like, "Okay, I'm, I'm, I need to build a new kind of application, so the first thing I need to do is build a new programming language."
393
00:45:54.968 --> 00:46:00.108
[laughs] Right? "And then the next thing I need to do is build a new operating system," right? [laughs] "And then the next thing I need to do is I need to build a new chip," right?
394
00:46:00.148 --> 00:46:01.768
And they, they kind of wanna reinvent everything.
395
00:46:02.328 --> 00:46:12.148
And I've, I've always had, maybe it's just, I don't know, pra- pragmatic mentality or something, or maybe an engineering over science mentality, but it's more like, no, you have just like all of this latent power, uh, in the existing systems.
396
00:46:12.168 --> 00:46:17.528
And you, you don't wanna be held back by their constraints, but what you wanna do is you wanna kind of liberate that power and open it up. Yeah.
397
00:46:17.548 --> 00:46:24.827
And so I, I think, I think the, and I think the web did that for those reasons, and I think it's the same thing now that's happening. It's a good perspective on the web. Programming languages is another good thing.
398
00:46:24.868 --> 00:46:33.408
We have Brett Taylor on the podcast, and we were talking about Rust, and you know- Yeah... Rust is memory safe by default, and so why are we teaching the model to not write memory unsafe code?
399
00:46:33.468 --> 00:46:42.568
Just use Rust and then you get it for free. How much do you think there's like time to be spent like recreating some of these things instead of taking them for granted and be like, "Oh, okay, Python is kind of slow."
400
00:46:42.588 --> 00:46:47.128
Python, TypeScript. You know, it's like- That's it. Yeah. As, as imperfect as they are, they are the lingua franca.
401
00:46:47.828 --> 00:46:51.768
I mean, I think this is gonna change a lot 'cause I don't think the models care what language they program in. Mm-hmm.
402
00:46:51.788 --> 00:46:59.068
And I think they're gonna be good at programming in every language, and I think they're gonna be good at translating from any language to any other language. Like, okay, so this gets into the coding side of things.
403
00:46:59.748 --> 00:47:05.548
I, I think we're going through a really fundamental change, and I, look, I, I grew up hand, you know, I grew up hand code, you know- Yeah, yeah, yeah... I, I grew up hand coding.
404
00:47:05.748 --> 00:47:12.888
Everything I did was actually, everything I did actually was written in C. I wasn't even- Back in the day. I wasn't even using C++, so I... Or like Java or any of this stuff, right?
405
00:47:13.068 --> 00:47:17.528
Uh, and so, um, I, everything, everything I ever did, I was like managing my own memory at, at, at the level of C.
406
00:47:18.508 --> 00:47:25.708
And then I, you know, I, I'm still from the generation that, you know, I, I knew assembly language and, you know, I, I, you know, um, so I, I could drop down and do things, uh, right on the chip.
407
00:47:26.388 --> 00:47:38.568
And so we, we've just, we've all, all of us, we've always lived in a world in which software is like this precious thing that like you have to think about very carefully, and it's like really hard to generate good software, and there's only a small number of people who can do it, and like you have to be very like jealous- Mm-hmm...
408
00:47:38.568 --> 00:47:47.128
in terms of thinking about like how do you allocate, like what are your engineers working on, and how many good engineers do you actually have, and how
much software can they write, and how can, how much software can human beings, you know, kind of maintain?
409
00:47:47.148 --> 00:47:50.328
And I think like all those assumptions are being shot right out the window right now.
410
00:47:50.528 --> 00:47:57.448
Like, I think they're, I, I think those days are just over, and I think the new world is like actually high quality software is just like infinitely available. Mm-hmm.
411
00:47:57.488 --> 00:48:01.888
And if you need new software to do XYZ, like you're just gonna wave your hand and you're gonna get it. Mm-hmm.
412
00:48:01.888 --> 00:48:08.596
[laughs] And then if it's, if you don't like the language it's written in, you just tell the thing, "All right, I want the, now I want the Rust version." Um, or, you know, secure, secure...
413
00:48:08.676 --> 00:48:15.876
You know, secure- we're about to- by the way, we're about to go through- computer security is about to go through the most dramatic change ever, which is, number one, like, every single latent security bug is about to be exposed.
414
00:48:15.936 --> 00:48:20.716
Right. So it's, we're gonna have, like, the in- we're, we're, we're set up here for, like, the computer security apocalypse for a while.
415
00:48:20.836 --> 00:48:26.916
But, but, but on the other side of it, now we have a- coding agents that can go in and actually fix all the security bugs. And so how, how are you gonna secure software in the future?
416
00:48:26.956 --> 00:48:30.036
You're gonna tell the, tell the bot to secure it, and it's gonna go through and, and fix it all.
417
00:48:30.776 --> 00:48:38.716
And so, so this thing that was this incredibly scarce resource of high-quality software is just going to become a completely fungible thing that you're just gonna have as much as you want. Right.
418
00:48:38.976 --> 00:48:42.576
Uh, and, and that has, like, you know, that has, like- Right. Yeah... ton- tons and tons of consequences.
419
00:48:43.276 --> 00:48:50.296
In some sense, the answer to the question that you posed, I, I think is just somewhat, I don't know, simpler or something, or straightforward, which is just if you want all your software in Rust, you just tell the bot you want all your software in Rust.
420
00:48:50.336 --> 00:48:57.556
Like, things that used to be, like, hard or even, like, seem like an insurmountable mountain to get, to get through, all of a sudden- Yeah... I think become very easy.
421
00:48:57.616 --> 00:49:05.476
I, I think Brett had a theory that there would be a more optimal language for LLMs. Yeah. And so the contention is, uh, there isn't. Like, just don't bother.
422
00:49:05.576 --> 00:49:12.015
Just whatever humans already use, LLMs are perfectly capable of porting. I think we're pretty close to being... I don't know if this would work today.
423
00:49:12.076 --> 00:49:20.676
I think we're pretty close to being able to ask the AI what would its opti- optimal language be, and let it- [laughs]... right, and let it design it. [laughs] True. Okay, here's a question.
424
00:49:21.436 --> 00:49:28.936
Are you gonna even gonna have programming languages in the future? Um, or are the AI- Right... are the AIs just gonna be emitting binaries? Let's assume for a moment that humans aren't coding anymore.
425
00:49:28.996 --> 00:49:35.176
Let's assume it's all bots. The b- what levels of intermediate abstraction do the bots even need? Yeah. Or are they just coding binary directly?
426
00:49:35.496 --> 00:49:42.996
Did you see there's actually an exper- somebody just did this thing where they have a, they have a l- a language model now that actually emits model weights for a new language model. [laughs] Right?
427
00:49:43.036 --> 00:49:47.136
And so will the bots be- Just predict the weights, yeah... well, yeah. Will the bots literally- Yeah...
428
00:49:47.156 --> 00:49:54.656
be emitting not just coding binaries, but will they, will, will they actually be emitting weights for n- for new models- Yeah... direct- directly? And conceptually,
429
00:49:55.556 --> 00:50:03.756
there's no reason why they can't do both of those things. Uh, uh, is the, like, just architecturally, both of those things seem completely possible. It's very inefficient. You're basically- Very inefficient...
430
00:50:03.776 --> 00:50:11.136
creating a simulation of a simulation in a simulation inside of the weights. Correct. Yeah. Yeah. Uh, uh, very inefficient. But, like, look, LLMs are already, like, incredibly, uh, uh, inefficient. Ask an, uh, uh...
431
00:50:11.216 --> 00:50:17.656
I have a favorite thing. Ask Claude, "Add 2 plus 2 equals 4," right? Right. It's just, like, you know, it's like, you know, it's, it's, like, whatever, billions and billions of times more- [laughs]...
432
00:50:17.956 --> 00:50:26.236
inef- inefficient than using your pocket calculator. Yeah. But, but, but yet the, the, the payoff is so great of the general capability. And so anyway, like, I, I kind of think in 10 years, like, I'm not sure...
433
00:50:26.616 --> 00:50:35.396
Yeah, like, I'm not sure there will even be a salient concept of a programming language, um, in the way that we understand it today. And in fact, what we may be doing more and more is a form of interpretability- Yeah...
434
00:50:35.416 --> 00:50:43.076
which is we're trying to understand why the bots have decided to, uh, structure, uh, code in the way that they have. I mean, if you play it through, you don't need browsers then. Like- Yeah...
435
00:50:43.116 --> 00:50:53.996
that's the death of the browser. Well, so I, I would take it a step further, which is you may not need user interfaces. Mm-hmm. So who is gonna use software in the future? Other bots. The other bots. Yep. Yeah.
436
00:50:54.036 --> 00:51:03.976
And so- You still need to, I don't know, pipe information in- Do we?... and out. Really? [laughs] Well, what are you gonna do then? Are you sure? You're just gonna log off and touch grass? Whatever you want, exactly.
437
00:51:03.996 -->
00:51:13.376
[laughs] Isn't that better? I want software to do stuff for me. Isn't that, but isn't that better? I mean, look, I... You know, I don't know. Look, like, you know, uh, you, you know the arguments here, you know.
438
00:51:13.396 --> 00:51:20.676
[laughs] It was not that long ago that 99% of humanity was behind a plow. Right. Right? And what are people gonna do if they're not plowing fields all day to, to, to grow food, right?
439
00:51:20.716 --> 00:51:28.156
And it just turns out there's, like, much better ways for people to spend time than plowing fields. Yeah, doom scrolling. Uh, yeah, exactly. [laughs] Exactly. Or, you know, talking to their friends.
440
00:51:28.536 --> 00:51:30.836
And look, and I'm not an absolutist, and I'm not a utopian, and I...
441
00:51:30.896 --> 00:51:36.416
A- and to be clear, like, I've, I have an 11-year-old, and he's learning how to code, and like, I'm, you know, I, I think it's still a really good idea to learn how to code and so forth.
442
00:51:36.456 --> 00:51:42.996
But I just, if you project forward, you just have to think forward to a world in which it's just like, okay, I'm just gonna tell the thing what I need, and it's gonna do it.
443
00:51:43.556 --> 00:51:51.536
And then, and then it's gonna do it in whatever way is most optimal for it to do it. Yeah. Unless I tell it to do it non-optimally, like if I tell it to do it in Java or in Rust or whatever- Mm-hmm...
444
00:51:51.576 --> 00:51:56.296
it'll do it, I'm sure. But, like, if I'm just gonna tell it to do it, it's gonna do it in whatever way is, like, the optimal way to do it. Yeah.
445
00:51:56.336 --> 00:52:00.776
And then I, and then if I need to understand how it works, I'm gonna ask it to explain to me how it works, right?
446
00:52:00.896 --> 00:52:06.556
And so it's gonna be doing its own interpret- it, it's gonna be the engine of interpretability to explain itself. And I, I just am not convinced that that...
447
00:52:06.756 --> 00:52:13.836
I'm not, I'm not convinced that in that world you have these h- historical... The goals of the abstractions will be whatever the bots need, not what the humans need. Right. Yeah. Yeah. Yeah.
448
00:52:13.856 --> 00:52:26.456
That, well, I, I'm curious, like, if that's true, then shouldn't the models providers be building some internal language representation that they can do extreme kind of like RL, uh, and reward modeling around?
449
00:52:26.576 --> 00:52:32.796
Because it's like today they're kind of like tied to like TypeScript and Python because the users need to write in that language- That's right. Yeah...
450
00:52:32.836 --> 00:52:41.076
versus they can have their own thing internally, and like they don't need to teach it to anybody. Yeah. They just need to teach their model. And I think that's how you get maybe the version between the models.
451
00:52:41.136 --> 00:52:50.416
Like going back to like the Pi OpenAI thing, it's like, oh, I built all the software using the OpenAI model, and now switch to the Anthropic model, but the Anthropic model doesn't understand the thing.
452
00:52:50.696 --> 00:52:59.396
So I w- uh, it feels like there still needs to be some abstraction, but maybe not. Maybe that's the lock-in that the model provider is going to have. I don't know.
453
00:52:59.456 --> 00:53:05.716
I'm not even sure that's lock-in though, 'cause why can't the second model just learn what the first model has done? Like- Exactly. It's like- Okay. So lo- okay, give you an example.
454
00:53:05.756 --> 00:53:12.916
So as you know, models can now reverse engineer software binary. Isn't it the whole thing now where people are reverse engineering like Ninten- Nintendo game binaries? Yeah. So you, you have like- Easy...
455
00:53:12.956 --> 00:53:22.016
there's, we've seen a bunch of reports like this where somebody has like a favorite game from the 1980s, and the source code is like long dead, but they have like a binary burned into a chip or something, and now they reverse engineer it and get a version that runs on their Mac, right?
456
00:53:22.116 --> 00:53:29.676
And so if you reverse it, if... This is what I kinda say. If you're reversing like x86 binaries, then why can't you reverse engineer- Whatever they create. Yeah... anything else?
457
00:53:29.696 --> 00:53:38.896
And because we're all on a Unix-based system, it has to be reversible because it needs to run on the target. Yeah. Yeah, yeah. Yeah. Yeah, basically. And so I just, I just think it's this thing where it's just like...
458
00:53:39.516 --> 00:53:44.956
And by the way, and everything we're describing is something that human beings in theory could have done before, but just with- Right. Yeah, yeah...
459
00:53:44.976 --> 00:53:53.556
but with enormous, where, where, but it was just always like cost and labor prohibitive. Reverse engineer, like I learned how to reverse engineer. I was like, human beings can reverse engineer binaries. Yeah.
460
00:53:53.616 --> 00:54:00.456
It's just for any complex binary, you need like 1,000 years- [laughs]... to do it. But now with the model, you don't. And so all of a sudden you get, you get these things...
461
00:54:00.596 --> 00:54:06.256
Or, or another way to think about it is so much of human-built systems are to compensate for the human limitations. Mm-hmm. Yep. Right?
462
00:54:06.356 --> 00:54:12.976
Um, and if you don't have the human limitations anymore, then all of a sudden you have... And, and it's not that you, you won't have abstractions, but you'll have a different kind of abstraction. Yeah. Yep.
463
00:54:13.224 --> 00:54:16.424
I have two topics to bring us to a close, and, and you can pick whi- whichever one.
464
00:54:16.464 --> 00:54:25.604
So just talking about protocols, was it you or someone else, I forget my internet history, who said that, like, the biggest mistake that we didn't figure out in the early days was payments? Yes. Was that you? Yes.
465
00:54:25.704 --> 00:54:32.604
It was a 402.
402. 402 payment required. We have a chance now. Yep. [laughs] I don't think we're gonna figure it out. I don't know. Like, what's your take? Oh, I think we will. Yeah.
466
00:54:32.644 --> 00:54:38.284
No, now I think it's gonna happen for sure. Yeah. Yeah. And there's two reasons it's gonna happen for sure. One is we actually have internet native money now in the form of crypto. Stable coins.
467
00:54:38.304 --> 00:54:43.884
Cryp- stable coins and crypto, and this is... I, I think this is the grand unification, basically, of AI and crypto, uh, is what's about to happen now.
468
00:54:44.004 --> 00:54:50.564
Um, I, I think AI is the crypto killer app I think is where, where this is really gonna come out. Um, and then the other is it's just, it... I mean, it's just I think it's now obvious.
469
00:54:50.624 --> 00:54:58.424
It's like obviously AI agents are gonna need money. And it's already happening, right? If you've got a cl- if you've got a claw and you want it to buy things for you, you have to give it money, uh, in some form.
470
00:54:58.644 --> 00:55:05.624
Um- I would say the adoption's probably like 0.1% if, if that, but yeah. Oh, today. Yeah, yeah, yeah. But think- Yeah... think forward. Like, where is it going? Forward thinking.
471
00:55:05.824 --> 00:55:12.524
The ultimate principle of everything and, and everything that I think I, we, we do is it's the William Gibson quote, which is, "The future is already here, it just isn't distributed." Mm-hmm.
472
00:55:12.544 --> 00:55:20.264
It isn't, isn't, isn't distributed yet. My friends who are the most aggressive us- users of, of, of, of OpenClaw just like have given their claws bank accounts- Yeah... and credit cards.
473
00:55:20.584 --> 00:55:28.284
Um, and, and, and, and, and not only have they done it, it's obvious that they needed to do it because it's obvious that they needed to be able to spend money on their behalf. Yeah, yeah. It's just completely obvious.
474
00:55:28.324 --> 00:55:34.804
And so, and again, like, so the number of people who have done that today, to your point, is like, I don't know, probably 5,000 or something. Yeah. But- It'll grow... that's how these things start.
475
00:55:35.104 --> 00:55:40.144
Actually, I mean, since, uh, you keep mentioning things- And by the way, OpenClaw, by the way, if you don't give it a bank account, it's just gonna break into your qui- your- Right.
476
00:55:40.364 --> 00:55:46.864
[laughs] It's gonna be breaking into- Be a agency... it's gonna break into your bank account anyway- [laughs]... and, and, and take your money. So you, you might as, you might as well do it. You might as well do it.
477
00:55:46.944 --> 00:55:53.964
All right. You- By the way, I really love, I gotta tell you- Yeah, yeah, yeah... I really love the phenomenon. I love the YOLO. Um, I'm not doing it myself to be clear- [laughs]...
478
00:55:54.004 --> 00:56:00.564
but I love the people that are just like, yeah, like, what, what is it? Skip, skip permissions. Yeah, dangerous. Skip permissions. Dangerous links. Yeah. Which by the way is a Facebook thing. Okay. [laughs] Right.
479
00:56:00.684 --> 00:56:09.264
'Cause, uh, 'cause we, uh, in Facebook, they, they have this culture to name the thing dangerous so that you are aware when you enable the flag that you are opting into a dangerous thing. Okay, good.
480
00:56:09.284 --> 00:56:17.044
And they brought it into OpenAI. Uh, and- And, and of course that makes it enticing... Sa- Sam runs Codex, uh, with skip permissions on, on his laptop. Yes, 100%.
481
00:56:17.064 --> 00:56:23.324
And so I, I th- I think the way to actually see the future is to find the people who are doing that. [laughs] There's a madness to, you know, they're, you know- Yeah...
482
00:56:23.344 --> 00:56:28.684
th- there's a- Log everything, you know, just watch it. Watch the logs... but, like, let's actually find out what the thing can do. Yeah, yeah.
483
00:56:28.724 --> 00:56:35.284
And the way to find out what the thing can do is just like- Try everything... l- yeah, let it try everything, let it unlock everything. Yeah. By the way, that's how you're gonna find all the good stuff it can do.
484
00:56:35.344 --> 00:56:41.564
By the way, that's also how you're gonna find all the flaws. Yeah. I think the people who turn that on for bots are like, they're like martyrs to the progress of human civilization.
485
00:56:41.564 --> 00:56:46.144
[laughs] Like, I feel very bad for their descendants that their bank accounts are gonna get looted by their bots in the first like 20 minutes.
486
00:56:46.744 --> 00:56:52.744
But I think the contribution that they're making to the future of our species is amazing. It's like gentleman science, you know? Yes. It's, yes, yes. Experiment on yourself.
487
00:56:52.804 --> 00:56:58.864
It's, it's, uh, Ben Franklin out with the- Yeah. [laughs]... trying to, trying, trying to get lightning to strike his, his, uh, his balloon and see, seeing if he gets electrocuted. Yeah.
488
00:56:59.164 --> 00:57:02.944
It, uh, it's, uh, Jonas Salk with the polio vaccine. Yeah. [laughs] Right? Injecting it in his... Yes.
489
00:57:02.964 --> 00:57:11.824
So yes, I, I, I, I think we should have like a gl- we should have like flags and like we should have like monuments- [laughs]... to the people that just let OpenClaw run their lives. More anecdotes.
490
00:57:11.864 --> 00:57:20.424
Or like, what, what are the craziest or interesting things that people listening to this should go, go home and do? I mean, this is, this is the, th- this is the, the extreme thing is just like the straight YOLO.
491
00:57:20.864 --> 00:57:27.764
Like just, yeah, turn, turn your life over. Well, that, that's a general capability of this. Yeah, yeah. Is there like a specific story that was like, wow, and, and everyone in the group chat just lit up?
492
00:57:28.044 --> 00:57:35.104
I mean, like, you know, so there's tons of, there's already tons of health, you know, there's, you
know, the, the health dashboard stuff is just, is just- Personal health... absolutely, absolutely amazing. Yeah.
493
00:57:35.124 --> 00:57:40.184
The number of stories on... I'm trying, I just don't want to violate people's, you know, obviously personal, uh- Non-anonymized... situations.
494
00:57:40.584 --> 00:57:48.584
But, um, you know, one of the things OpenClaw is really good at is hacking into all the stuff in your LAN. It's really good at it. So, you know, Internet of Things, AKA Internet of Shit. Yeah.
495
00:57:48.804 --> 00:57:52.904
[laughs] Like- Super insecure, but great. So- It's discoverable... yeah, it's discoverable.
496
00:57:53.124 --> 00:58:00.404
[laughs] And OpenClaw is happy to scan your network, identify all the things, and then my, my, my friends who are most aggressive at this are having OpenClaw take over everything in their house. Yeah.
497
00:58:00.444 --> 00:58:08.004
Take, it takes over their security cameras, it takes over their, their, you know, their whatever, their, their access control systems. It takes over their webcams. I have a friend whose claw watches him sleep.
498
00:58:08.464 --> 00:58:13.504
Put a webcam in your bedroom, put the, put the claw, put the claw on a loop. Uh, have it wake up frequently and have it watch. And just tell it, "Watch me sleep."
499
00:58:14.084 --> 00:58:17.804
And, and I've, I've seen the transcripts, and it's literally like, "Joe's asleep. This is good.
500
00:58:17.844 --> 00:58:23.324
This is good that Joe's asleep because, you know, I have, I have his health data, and I know that he hasn't been getting enough sleep, and so it's really good that he's getting sleep.
501
00:58:23.384 --> 00:58:32.343
I really hope he gets his full whatever, you know, five hours of REM sleep. Da, da, da, da, da. Uh, Joe's moving." [laughs] "Joe's moving. Um, uh, Joe might be wake, waking up. This is a real...
502
00:58:32.484 --> 00:58:40.164
If Joe wakes up now, he's gonna ruin his sleep cycle. Uh, oh, okay, it's okay. Joe just rolled over. Okay, he's gone back to bed. Okay, good. All right. Okay, I can relax. This is fine." Right?
503
00:58:40.164 --> 00:58:47.484
He's monitoring the situation. [laughs] Monitoring, monitoring the situation. And, and being a bot, like, you know, is just like very focused, right?
504
00:58:47.504 --> 00:58:51.784
[laughs] It's just like, ah, th- this is like its reason for existence- [laughs]... is to watch Joe sleep.
505
00:58:52.244 --> 00:58:58.604
And then, and I was talking to my friend who did this, he's like, you know, on the one hand it's like, all right, this is weird and creepy, um, and I need to, I need to... Maybe this is taking over my life.
506
00:58:58.624 --> 00:59:03.044
And then the other thing is like, you know what? If I had a heart attack in the middle of the night, this thing literally would like freak out and call 911.
507
00:59:03.224 --> 00:59:09.744
[laughs] Like, there's no question this thing would figure out how to, like, alert medical authorities and, like, prob- probably summon SWAT teams and, like, do whatever would be required- [laughs]...
508
00:59:09.804 --> 00:59:19.444
to save my life, right? And so it's like, you know, like, yeah. Like, that's happening. What else? Um, I'll give you, I've, um... There's a company, Unitree- Mm-hmm... uh, that makes the robot dogs.
509
00:59:19.684 --> 00:59:24.824
Um, and I, I actually have one at home, which is, is, it's actually really fun. The Chinese companies...
510
00:59:25.664 --> 00:59:34.164
The Chinese companies are so aggressive at adopting, uh, new technology, but they don't always, like, let's say, take the time to really- Package it?... package it and maybe think it all the way through.
511
00:59:34.184 --> 00:59:43.304
And so, so the u- at least the Unitree dog I have, so it, it has a old non-LLM just control system, which by the way is not very good. In, in markets well, but it in practice it's, it's not that good.
512
00:59:43.344 --> 00:59:47.376
It has trouble with stairs and so forth, and so it's not quite what it should be. But then the language model thing comes out in the voice.
513
00:59:47.416 --> 00:59:55.216
So they, they add, so they add LLM capability, and then they, they add a voice mode to it. Um, but, but that LLM capability is not at all connected to the control system.
514
00:59:55.336 --> 01:00:03.036
[laughs] So, so you've got this schizophrenic dog that, like, is a complete idiot when it comes to climbing the stairs, but it will happily teach you quantum mechanics. [laughs] Right?
515
01:00:03.096 --> 01:00:09.536
In like a plummy English accent, right? [laughs] Like, it, it, it's just, like, absolutely amazing. Jagged intelligence. Yeah, ex- yeah, talk about jagged.
516
01:00:09.576 --> 01:00:15.216
And then, now obviously what's gonna happen in the future is, is they're gonna connect together, but- They'll do it... but right now it's, it's... And so right now it's not that useful.
517
01:00:15.236 --> 01:00:23.056
And so I, I have a friend who has one of these who had his claw basically hack in and rewrite the code, re- write new firmware. Yeah. Wri- write new firmware for the, for the Unitree robot. Ooh.
518
01:00:23.116 --> 01:00:30.996
A- and now it's, now it's an actual pet dog for his kids. You should do the before or after, like the motion. Yeah. He said it's completely different. Yeah. He said it's a complete transformation. Yeah.
519
01:00:31.016 --> 01:00:37.416
And whenever there's an issue in the thing now, the claw just, like, rewrites the code. You know, you know, you goes in- Yeah, yeah... and does, does the code, and so on. It's a g- kind of goes to your thing here.
520
01:00:37.456 --> 01:00:44.736
And so, so like all of a sudden, uh, this is why, like, when we wanna think about AI code, AI coding is not just like writing new apps, it's also going in and rewriting all the old stuff that should've worked that never worked.
521
01:00:45.376 --> 01:00:49.096
And so, like I, I
think, I think basically, I think the internet, the internet of shit is basically over.
522
01:00:49.136 --> 01:00:55.636
Like, I, I think everything, it- there's a potential here where, like, all these devices in your house that have been like basically marginal or, you know, basically dumb, you know, for, uh...
523
01:00:55.736 --> 01:01:02.556
Like, all of a sudden they might all get really smart. Now, you have- Smart home... you have to decide if... Yes. There are horror movies in which this- Right...
524
01:01:02.616 --> 01:01:10.796
is, uh, of which this is the premise, and so [laughs] you have to decide if you want this. Yeah. But, but, but this is the first time I can say with confidence I now know how you could actually- Yeah, yeah...
525
01:01:10.796 --> 01:01:15.796
have a smart home- Yeah, yeah... with 30 different kinds of things with chips and internet access, where it actually all makes sense and all works together- Yeah, yeah...
526
01:01:15.816 --> 01:01:23.536
and it's all coherent and the, and the whole thing. Yeah. And to have that unlock without a human being having to go do any of that work, like- Yeah... you know. So yeah. It's, uh- Yeah.
527
01:01:23.576 --> 01:01:30.996
I, I'm waiting for a, "Sorry, Mark, uh, I can't let you open that fridge door." You know, like [laughs] Exactly. Exactly. Yes. Yes. Because- Oh, yeah, yeah, yeah, yeah... you're not supposed to eat right now.
528
01:01:31.076 --> 01:01:37.136
I have all of... Yes, I have every shred of health information. [laughs] You know, and I know you think you're doing, you know, da, da, da, and I think you can do this, but you know, this is a real...
529
01:01:37.216 --> 01:01:42.716
Are you really, you know, are you really sure? [laughs] And you know, you told, you know, you told me last night you really don't want me to let you do this. Yeah.
530
01:01:42.716 --> 01:01:49.136
[laughs] So, you know, I'm sorry, but the fridge door is locked. Um, yes. Open the fridge doors. Exactly. But, uh- And by the way, I know you're supposed to be studying for a test- [laughs] Yeah...
531
01:01:49.176 --> 01:01:57.556
so why don't we, why don't you go, like, when you can pass the test, um, I will open the fridge door for you. Yeah. Final protocol- Yeah... and then, and then we can wrap up. Uh, proof of human. Yes. Uh, right? Yeah.
532
01:01:57.876 --> 01:01:59.916
That's the last piece that we gotta figure out. Yeah.
533
01:02:00.416 --> 01:02:12.196
So I would say there's, there's two massive, I would say, um, uh, sort of asymmetries in the world right now, where we've known these asymmetries exist and we, we societally have been unwilling to grapple with them, and I think they're both tipping right now.
534
01:02:12.636 --> 01:02:15.936
And, and they're, they're, they're, they're the same thing as virtual world version as the physical world version.
535
01:02:15.976 --> 01:02:23.076
So the virtual world version is, is the bot problem, where just like, you know, the internet, internet is just like awash in bots. Internet's awash in fake people. It has been forever.
536
01:02:23.356 --> 01:02:30.496
Um, by the way, a lot of that has to do with lack of money, you know, and so this, you know, this is the- Yeah... this is the s- My spicy take was these two are the same thing, and corporations are people too, you know?
537
01:02:30.736 --> 01:02:40.316
So [laughs] Interesting. Yeah, yeah, yeah. Okay. Got it. So a bank account is proof of human. Yeah. Okay. [laughs] Yeah. Until you, until you give the bots bank accounts. Yeah, exactly. So, okay. Yeah, so there's that.
538
01:02:40.356 --> 01:02:47.116
But yeah, look, m- look, the bot... I mean, every social media user knows this. The bot, the bot problem is a big problem. You know, the bot, the bot problem has been a bigger problem forever.
539
01:02:47.176 --> 01:02:55.436
It's, it's a huge problem, and it's never really been confronted directly, like at any point. By the way, the physical world version of this is the drone, the drone problem. Um, right?
540
01:02:55.476 --> 01:03:06.796
And so we, we've known for, you know, we've known for 20 years now that the asymmetric threat, both in mili- military, in actual military conflict, but also in just like security, like, like, s- you know, security on the home front, the big threat is, is the cheap attack drone, right?
541
01:03:06.836 --> 01:03:10.796
The, the, the cheap, the cheap suicide, you know, drone with a bomb. And we've known that forever.
542
01:03:10.836 --> 01:03:17.236
And by the way, like, you know, it, it's very disconcerting how like every, you know, every office complex in the, in the coun- you know, in the world is like unprotected from drone attacks.
543
01:03:17.496 --> 01:03:25.416
Um, every, every stadium, every school, every prison, like, it like- Sure... it, it... Okay. We've known that. We've never done anything about it. What are you gonna do about it? Yeah.
544
01:03:25.716 --> 01:03:39.256
One possibility is just leave, leave them unprotected forever, and live in a world of like asymmetric terrorism forever, or the other is take the problem seriously and figure out the set of techniques and technologies required to, to be able to deal with that, whether those are lasers or jammers or early warning systems or, you know, all, all- Personal force fields...
545
01:03:39.296 --> 01:03:47.596
kinetic personal for- doing, uh [laughs] Personal force fields. Exactly. And in both cases, the, these are, these are economic asymmetries. These are economic asymmetries, right?
546
01:03:47.636 --> 01:03:54.296
'Cause it's really cheap to field a bot, but it's very hard to- Yeah... tell something a bot. It's very cheap to field a drone. It's very hard... It's very expensive to defend against a drone.
547
01:03:54.736 --> 01:03:58.476
But you see what I'm saying is it's, it's, it's the, it's the virtual version of the problem, and it's the physical version of the problem.
548
01:03:59.036 --> 01:04:05.216
Uh,
the virtual version of the problem, what w- what we need quite literally is proof of human. The reason is because you're, you're not, you're not gonna have proof of bot. The, the, the... Especially now.
549
01:04:05.255 --> 01:04:13.236
The, the bots are too good. The, the, the bots can pass the Turing test, and if the bots can pass the Turing test, then you can't, you can't screen for bot. You can't have proof of not a bot.
550
01:04:13.596 --> 01:04:18.736
But what you can have is you can have proof of human. You can have, you know, cryptographically validated this is definitely a person, and this is...
551
01:04:18.776 --> 01:04:25.556
And then you can have cryptographically validated this is definitely like something that a person said. Yeah. This video is real, right? Um- Ju- just to double click on- Yeah. Go ahead...
552
01:04:25.596 --> 01:04:32.716
on, uh, do you think Alex Blania with World- Yeah... do you think he's got it? Yeah. Or is there an alternative? Oh, so I mean, there's gonna be... I think there'll be... I think many people will try.
553
01:04:32.856 --> 01:04:37.516
We're one of the key, you know, participants in, in, in the World, in the World project and, and investors. I know. I know that. Yeah. Yeah. So we're, we're partisans.
554
01:04:37.856 --> 01:04:46.196
But yeah, I, I think, so we think World is exactly correct. Okay. And, and the reason is it, it has, it has to be, it, it has to be proof of human. It, it has to... Because you can't do proof of not bot.
555
01:04:46.216 --> 01:04:53.176
You have to do proof of human. To do proof of human, you, you need bi- you need biological validation. You, you need it to start with this was actually a person, right?
556
01:04:53.236 --> 01:04:56.936
'Cause otherwise you have bots signing up as fake people, right? So you, you have to have like something.
557
01:04:56.976 --> 01:05:02.336
You have to have a bio- a biometric, and then you have to have cryptographic validation, and then the ability to do- Sure... to do, to do the lookup.
558
01:05:02.736 --> 01:05:09.616
And then by the way, the other thing you need was that you, you also need selective disclosure. Um, so you need to be able to do proof of human without revealing- Privacy... all the underlying information.
559
01:05:09.636 --> 01:05:11.996
By the way, another thing you're gonna need, you're gonna need proof of age, right?
560
01:05:12.056 --> 01:05:21.976
'Cause and there's all these laws in all these different countries now around you need to be 13 or 16 or 18 or whatever to do different things, and so you're gonna, you're gonna need a, you know, sort of validated proof of age, um, you know, to be able to legally operate, right?
561
01:05:22.016 --> 01:05:27.456
And so that, that's coming, and then you're gonna want like proof of credit score and, you know, proof of like, you know, 100 other things. That's a tricky one.
562
01:05:27.616 --> 01:05:33.656
[laughs] It, it is a tricky one, but you're gonna, you're gonna n- there, there's no reason, like if somebody's checking on your credit, somebody shouldn't... Give you an example.
563
01:05:33.696 --> 01:05:40.436
Somebody shouldn't need to know your name in order to be able to find out whether you're credit worthy, right? I see. Independently verifiable inf- pieces of information. Pieces of information. Yeah.
564
01:05:40.436 --> 01:05:46.016
It's like, it's like we disclosed, and this is the answer to the privacy problem writ large- Mm-hmm... which is I, I only need to prove what I need to prove at that moment.
565
01:05:46.336 --> 01:05:53.016
So like you're gonna need that, and I, I think their, their, their architecture makes sense. So that needs to get solved. I think language models have tip- the bots are now too good.
566
01:05:53.656 --> 01:05:58.116
Uh, and, and, and so they're undetectable, and so as a consequence, you, we now need to go confront that problem directly.
567
01:05:58.176 --> 01:06:05.372
And then, and like I said, and then the other problem is we, we need to go actually confront the drone problem. The Ukraine conflict has really unlocked a lot of thinking on that. Mm-hmm.
568
01:06:05.372 --> 01:06:12.452
And now the, um, and now the, the, the, the- Yeah... Iran situation is also unlocking that. And so I think there's gonna be just like this incredible explosion of, of both drone and counter-drone.
569
01:06:12.452 --> 01:06:20.092
It's just our drones are better than their drones. [laughs] It's, it's let's keep it that way. Yeah. [laughs] Yeah. Yeah. And count- and counter-drones. I think we can sneak in one more question. Okay.
570
01:06:20.192 --> 01:06:25.472
Um, I'm trying to tie together a lot of things that you said over the years. So at the Milken Institute debate with Thiel- Yeah...
571
01:06:25.492 --> 01:06:31.712
which is amazing, um, you talked about the lag between a new technology and kind of like the GDP, um, impact of it. Yep.
572
01:06:32.752 --> 01:06:39.592
The other idea you talked about is bourgeois capitalism and how, you know, this kind of managerial class was needed because of this complexity.
573
01:06:39.652 --> 01:06:44.432
And I think if you bring AI into the fold, you have like much higher leverage- Mm-hmm... of people.
574
01:06:44.472 --> 01:06:51.572
So like if you have, you know, the Musk industries, um, and you give Elon AGI, he can run a lot more things- That's right... uh, at once. That's right.
575
01:06:51.592 --> 01:06:58.852
And then you have the social contract, and I know you retweeted a clip of Sam Altman saying, um, "We're rethinking the whole thing" and you were like, "Absolutely not." Yes.
576
01:06:58.892 --> 01:07:05.972
Under- [laughs] Uh, and I wa- I was in an event with Sam last night, um, and he actually said in the last couple of weeks he felt like now people are taking that seriously. Yeah.
577
01:07:06.312 --> 01:07:16.772
So I'm just curious like how you're seeing the structure of organization changing, especially when
you invest in early-stage companies and, um, yeah, just like how the impact of work structure and, uh, all of that is playing out.
578
01:07:16.992 --> 01:07:22.752
Yeah. So there's a whole bunch of, there's a whole bunch of topics. I know, yeah. We could, we could spend... And by the way, we'd be happy to spend more time, but we could, we could spend more time on all that.
579
01:07:22.832 --> 01:07:32.392
So just so for people who haven't followed this, so the, this, this, this term managerial comes from this thinker in the 20th century, James Burnham, who, um, just one of the great kind of 20th-century political thinkers, um, societal thinkers.
580
01:07:32.432 --> 01:07:39.692
And he sort of said a- as... And he was writing in like the 1940s, 1950s. Um, and he said kind of the, the whole history of capitalism until that point had been in two phases.
581
01:07:39.752 --> 01:07:47.192
Number one had been what he called bourgeois capitalism, which was think about it as like name on the door. Like f- Ford Motor Company 'cause Henry Ford runs the company. Um, and Henry...
582
01:07:47.252 --> 01:07:54.532
It's like a dictat- dictatorial model, and Henry Ford just like tells everybody what to do. And he said the problem with bourgeois capitalism is it doesn't scale, 'cause Henry Ford can only- Mm-hmm...
583
01:07:54.552 --> 01:07:57.092
tell so many people to do so many things, and then he runs out of time in the day.
584
01:07:57.232 --> 01:08:11.031
And so, um, he said the second phase of capitalism was what he called managerial capitalism, which was the creation of a professional class of managers, um, that are trained not to be like car experts or to be whatever experts in any particular field, but are trained to be experts in management.
585
01:08:11.492 --> 01:08:16.312
And then that led to, you know, the importance of like Harvard Business, you know, business schools and management consulting firms and all these things.
586
01:08:16.832 --> 01:08:26.892
And then you look at every big company today and like most of the executives and most of the Fortune 500 companies are not domain experts in whatever the company does, and they're certainly not the founders of those companies, but they're professional managers.
587
01:08:27.452 --> 01:08:30.992
And in fact, in the course of their careers, they'll probably manage many different kinds of businesses.
588
01:08:31.032 --> 01:08:36.052
They'll rotate around and they might work in healthcare for a while and then work in financial services and then go work in something else, you know, come work in tech.
589
01:08:36.912 --> 01:08:44.072
And what Burnham said is he said that transition is absolutely required because the, the, the, the problem with bourgeois capitalism is, is it doesn't scale. Henry Ford doesn't scale.
590
01:08:44.592 --> 01:08:50.552
And so if you're gonna run capitalist enterprises that are gonna have millions to billions of customers, um, you're gonna need to... You're...
591
01:08:50.732 --> 01:08:57.452
They're gonna be operating a level of scale and complexity that's gonna require this professional management class. And he said, "Look, the, the professional management class has its downsides.
592
01:08:57.492 --> 01:09:02.352
Like they're not necessarily experts- Mm-hmm... at doing the thing. They're not as inventive. You know, they're not gonna create the next breakthrough thing."
593
01:09:02.932 --> 01:09:09.352
But he's like, "Whether you think that's good or bad or whatever, it's what's gonna be required." And basically that's what happened, right? And so he wrote that book originally in like 1940.
594
01:09:10.012 --> 01:09:16.552
You know, over the course of the next 50 years, basically managerialism... Well, I mean today, up till today, managerial, managerialism basically took over everything. Mm-hmm.
595
01:09:16.772 --> 01:09:23.672
And you know, what I'm describing is basically how all big companies run and how all governments run and how our large scale nonprofits run and kind of everything, you know, everything runs.
596
01:09:24.292 --> 01:09:36.772
Basically what, what, what venture capital does is we basically are a rump, uh, sort of protest movement to that to try to find the next Henry Ford or, or just to say E- Elon Musk or the, or the next, or the next Elon Musk or the next Steve Jobs or the next Bill Gates or the next Mark Zuckerberg.
597
01:09:36.812 --> 01:09:43.872
And so we, we, we, we start these companies in, in the old model, right? We, we, we start them out as, as, as, as like in the Henry Ford model. Mm-hmm.
598
01:09:43.892 --> 01:09:48.472
And so we start them out with a founder or a, or a, or a founder with, with colleagues. But you know, there's, there's a founder CEO.
599
01:09:49.272 --> 01:09:57.592
Um, and then we basically bet that, we basically bet that the startup is going to be able to do things, specifically innovate in ways that the big incumbents in that industry are not gonna be able to do.
600
01:09:57.632 --> 01:10:02.352
And so it's a bet that by, basically by relighting this sort of name on the door, you know, kind of thing- Mm-hmm...
601
01:10:02.392 --> 01:10:12.212
this new innovative thing with like a king monarchical, uh, uh, political structure, um, that they're gonna be able to innovate in a way that the incumbent is not going to be able to because the incumbent is, is being run by managers, right?
602
01:10:12.252 --> 01:10:17.392
And, and, and, and by the way and of course, venture being what it is, sometimes that works, sometimes it doesn't. But we're, we're constantly doing that.
603
01:10:17.412 --> 01:10:21.171
But I've always viewed it my entire life as like we're like raging against the dying of the light. Mm-hmm.
604
01:10:21.192 --> 01:10:30.252
Like we're, we're, we're, we're sort of constantly trying to fight off managerialism just basically swamping everything- [laughs]... and everything getting basically boring and gray and dumb and old, right?
605
01:10:30.292 --> 01:10:38.772
And it's- Right... we're trying to keep some level of energy and vitality in the system. AI is the thing that would lead you to think, wow, maybe there's a third model. Mm-hmm. Right?
606
01:10:38.832 --> 01:10:47.672
And, and maybe may- and way to think about it would be maybe it's a combination of the two. Maybe the new Henry Ford or the new Elon or the new Steve Jobs plus AI is- Right... the best of both, right?
607
01:10:47.752 --> 01:10:57.772
'Cause it's, it's, it's sort of the spark of genius of the name on the door model, the Henry Ford model, but then it's give that person AI superpowers to do all the managerial stuff and let the bots do all the managerial stuff.
608
01:10:58.292 --> 01:11:04.252
That may be the actual secret formula. And we've never even known that we wanted this because we never even thought- Right... it was a possibility.
609
01:11:04.832 --> 01:11:11.592
But I mean, you know this, the, what is the thing that these bots are really good at? They're really good at doing paperwork. Like they're really good at filling out forms. [laughs] Right.
610
01:11:11.612 --> 01:11:17.592
Like they're really good at writing reports. They're really good at reading. They're really good at doing all the managerial work. Like they're amazing at it.
611
01:11:17.812 --> 01:11:26.692
And so, yeah, so I, I think, I think the I, 100%, I think the answer, the answer very well might be to get the best, be- best of both worlds by doing this. And then the challenge is gonna be twofold.
612
01:11:26.732 --> 01:11:31.312
The challenge is gonna be for the innovators to really figure out how to leverage AI- Mm-hmm... to actually do this, right?
613
01:11:31.412 --> 01:11:37.452
Um, a- and then, and then the o- the other challenge is gonna be for the, for the incumbents that are managerial to figure out like, okay- Right... what does that mean?
614
01:11:37.492 --> 01:11:42.712
'Cause now they're gonna, they're, they're gonna be facing a different kind of insurgent competitor that has a different set of capabilities than they're used to.
615
01:11:43.212 --> 01:11:49.712
And so the, the, the, it, this really I think is gonna force a lot of big companies to kind of figure out innovation. E- either I say figure out innovation or die trying.
616
01:11:50.752 --> 01:11:58.572
Do you feel like that structure accelerates the impact on the actual GDP and economy? If you look at SpaceX- Yes... it's like the growth is like so fast. Yes.
617
01:11:58.592 --> 01:12:05.964
And like instead of having these companies kind of like peter out in growth and impact, they can kind of like keep going, if not accelerating. Yeah. That's for sure the hope.
618
01:12:06.164 --> 01:12:14.424
Um, the, the, the challenge, and, and, you know, and, and look, the AI utopian view is of cour- of course, and, and, and that's gonna be the future of the economy, and it's gonna grow 10X and 100X and 1,000X, and we're- Right...
619
01:12:14.444 --> 01:12:20.244
entering this regime of, like, much higher economic growth forever, and consumer cornucopia of everything, and it's, it's gonna be great. And I, and, and I hope that's true.
620
01:12:20.304 --> 01:12:26.644
I hope that's, that's, like, the u- you know, that's the current kind of utopian vision. I hope that's true. The problem is, it goes back again, the real world is really messy.
621
01:12:27.004 --> 01:12:34.504
Um, and I'll give you an example of how the re- real world is really messy. It requires 900 hours of professional certification training to become a hairdresser in the state of California.
622
01:12:34.524 --> 01:12:45.904
[laughs] Um, so it's, like, 35% of the economy, something like that, you have to get some sort of professional certification to do the job. Which is to say that the, the professions are all cartels, right? Yeah.
623
01:12:45.924 --> 01:12:55.784
And so you have to get licensed as a doctor, you have to get licensed as a lawyer, you have to get licensed as a... You have to get into a union. Mm-hmm. Um, by the way, to, to work for the government, you need to be...
624
01:12:55.984 --> 01:13:03.724
You, you have both civil service protections and you have public sector unions. You have two layers of insulation, uh, against ever getting fired for anything or anything, anything ever changing.
625
01:13:03.744 --> 01:13:08.904
[laughs] I'll give you another example. The, the dock work- the dock workers went on strike a couple years ago- Mm-hmm. Yeah... 'cause they're, you know, robo- robotics.
626
01:13:08.944 --> 01:13:16.344
You, you know, if, if you go look at a modern dock, like in Asia, it's all robots. If you go to a, a American dock, it's, like, all still guys dragging, strike, dragging stuff by, by hand.
627
01:13:16.884 --> 01:13:23.984
The dock workers went on a strike. It turns out there are 25,000 dock workers working on, on, on docks in America. It turns out they have incredible political power, 'cause it's a- Mm-hmm...
628
01:13:24.024 --> 01:13:31.644
it's, it's one of these u- unified blocks of things. They won their strike, and so they got commitments from the dock owners to not implement more automation. We learned a couple things in that.
629
01:13:31.684 --> 01:13:37.024
So number one, we learned that even a union as small as 25,000 people still has, like, tremendous political stroke. We also learned that they...
630
01:13:37.224 --> 01:13:40.864
It actually turns out the dock workers union has 50,000 people in it, 'cause there's 20...
631
01:13:40.884 --> 01:13:49.344
They have 25,000 people working at the docks, they have 25,000 people drawing full paychecks sitting at home from prior union agreements. Oh my God. From prior union agreements. I'll give you another great example.
632
01:13:49.584 --> 01:13:56.744
There are government agencies, there are federal government agencies where the employees, right, have, have civil service protections and they're in public sector unions.
633
01:13:56.784 --> 01:14:03.184
There are entire federal government agencies that struck new collective bargaining agreements during COVID, where not only are they...
634
01:14:03.244 --> 01:14:10.464
have their jobs guaranteed in perpetuity, but they only have to report to work in an office one day per month. And so there are entire office buildings in Washington, D.C.
635
01:14:10.544 --> 01:14:15.744
that are empty 29 out of 30 days of the year that are still operating and are still... we're all still paying for it, 20...
636
01:14:16.064 --> 01:14:23.084
And so, and then what they do, it turns out what the employees do is they're very, they're very smart in, in, in this way, and so they figure out, they come in on the last day of a month and the first day of the next month.
637
01:14:23.744 --> 01:14:30.164
And so- [laughs]... and so they're, so they're in their, they're in the office two days per 60 days. [laughs] Gotcha. Which means these buildings are empty for 58 days at a time.
638
01:14:30.184 --> 01:14:40.284
[laughs] And you see what I'm, you see where I'm heading with this. Like, this is, like, locked in. Right. Mm-hmm. Right? This is, like, locked in in a way that has nothing to do with, like... And people say capitalist.
639
01:14:40.324 --> 01:14:50.784
It's, like, anti-capitalistic. It's like, it's, it's basically, it's restrictions on trade. It's restrictions on the ability to, like, change the workforce. And so, so much of our economy is, is, you know, the, the...
640
01:14:51.184 --> 01:15:00.524
I, I'm describing the entire healthcare system. I'm describing the entire legal profession. I'm describing the entire housing industry. [laughs] I'm describing the entire education system. Right?
641
01:15:00.844 --> 01:15:08.044
K through 12 schools in the United States, they're a literal government monopoly. How are we gonna apply AI in education? The answer is we're not, 'cause it's a literal government monopoly.
642
01:15:08.104 --> 01:15:15.224
It is never going to change, the end, and there is nothing to do. By the way, you can create an entirely new school system. Like, that's the one thing you can do, is you can do what Alpha School is doing.
643
01:15:15.264 --> 01:15:21.404
You can create an entirely new school system. Mm-hmm. Yeah. Other than that, you're not gonna go in and change what's happening in the American classroom, like K through 12. There's no chance.
644
01:15:21.684 --> 01:15:29.124
The teachers are 100% opposed to it. It's 100% not gonna happen. So, so you see what I'm saying, is, like, there's this, like, massive slippage that's gonna take place. Yeah.
645
01:15:29.164 --> 01:15:36.064
Both the AI utopians and the AI doomers are far too optimistic. Right. [laughs] You see what I'm saying? Because, because they believe- Right...
646
01:15:36.084 --> 01:15:43.924
that because the technology makes something possible, that eight billion people all of a sudden are gonna change how they behave, and it's just like, nope. So much of how the existing economy works- Mm-hmm...
647
01:15:43.944 --> 01:15:52.604
is just, it, it's just, like, wired in. Yep. And so we're gonna be lucky... As a society, we're gonna be lucky if AI adoption happens quickly, right? 'Cause if it doesn't, what we're just gonna have is stagnation.
648
01:15:53.344 --> 01:16:01.404
Awesome. Mark, I know you gotta run. Yeah, we all know you're always still welcome, but, uh, it was such a pleasure talking to you. Uh, we're truly living in an age of science fiction coming- Yes... to real life.
649
01:16:01.724 --> 01:16:19.564
Yes, yes. Could not be more exciting. [laughs] Yeah. Really thank you, Mark. Thank you, guys. Awesome. Thank you. Good. That's it. Good. Thank you. That's it. [outro music]