Latest / Elon Musk Thinking / The Vision of Elon Musk About the Future of AI!!!

Transcript

The Vision of Elon Musk About the Future of AI!!! · Elon Musk Thinking

• 0:00
  Speaker 1: Latest interview of elon musk first uh our first speaker
• 0:03
  Speaker 1: is uh is elon musk as the uh ceo of
• 0:07
  Speaker 1: tesla and spacex elon has shown that conquering challenges of
• 0:10
  Speaker 1: physics engineering can be easier than cutting through red tape
• 0:14
  Speaker 1: and i think his uh dedication to doing both whoever
• 0:17
  Speaker 1: continues to inspire people people around the world so Thank you, Elon,
• 0:21
  Speaker 1: for joining us. We were just wrapping up our introductory
• 0:24
  Speaker 1: session where we made clear that under this presidency, we
• 0:28
  Speaker 1: believe that economic growth is one of the most important
• 0:32
  Speaker 1: drivers for all of us, and we believe that technology
• 0:34
  Speaker 1: is a big piece of that. So you've been on
• 0:38
  Speaker 1: the front lines of this for quite a while, and
• 0:40
  Speaker 1: I think one of the key questions, I think, for
• 0:42
  Speaker 1: the group and something that we wanted to talk to
• 0:43
  Speaker 1: you about was, as you have driven innovation across multiple
• 0:48
  Speaker 1: countries around the world, You know, what, in your opinion,
• 0:51
  Speaker 1: separates countries where innovators can successfully turn breakthroughs into deployed
• 0:56
  Speaker 1: technologies from those where progress actually stalls? Because I think
• 1:00
  Speaker 1: there's a lot of lessons learned about what people, governments
• 1:03
  Speaker 1: should be doing and what maybe they shouldn't be.
• 1:08
  Speaker 2: Well, yeah, I think a lot of people have talked
• 1:10
  Speaker 2: about this, and I think some of it, frankly, is
• 1:13
  Speaker 2: pretty straightforward, is that... you have to have an environment
• 1:18
  Speaker 2: that's relatively free of regulation, meaning that new things must
• 1:22
  Speaker 2: be default legal as opposed to default illegal. Um, so, uh, in,
• 1:28
  Speaker 2: in the EU, for example, we find that the regulation
• 1:31
  Speaker 2: level is extraordinarily high and things are generally default illegal. Um,
• 1:35
  Speaker 2: and this inhibits, uh, progress of new technologies. Um, it
• 1:40
  Speaker 2: slows it down. It doesn't, it doesn't ultimately stop it,
• 1:42
  Speaker 2: but it slows it down, uh, quite considerably. Um, Of course,
• 1:47
  Speaker 2: you need to have venture capitalists and an environment that
• 1:54
  Speaker 2: is supportive of new companies. You can think of new
• 1:58
  Speaker 2: companies like small saplings in a forest. So what most
• 2:06
  Speaker 2: countries tend to do is they tend to provide too
• 2:09
  Speaker 2: much support to the large existing trees in the forest
• 2:13
  Speaker 2: and not enough to the small saplings. But the large
• 2:17
  Speaker 2: trees don't need their support. It's the small saplings that do,
• 2:20
  Speaker 2: the startups. And so the system should be generally biased
• 2:24
  Speaker 2: towards supporting the small trees as opposed to.
• 2:29
  Speaker 3: The large ones.
• 2:32
  Speaker 2: But that is rarely the case because the large companies
• 2:37
  Speaker 2: have access to the leadership of the countries and the
• 2:43
  Speaker 2: small startups do not. So you really need to foster
• 2:49
  Speaker 2: the growth of young companies and take active steps in
• 2:52
  Speaker 2: that regard. And like I said, make things default legal,
• 2:56
  Speaker 2: not default illegal.
• 2:59
  Speaker 1: One question that a lot of the countries here face
• 3:02
  Speaker 1: is a question around adoption. I think there's a general
• 3:07
  Speaker 1: recognition that adopting emerging technologies can be very beneficial to
• 3:11
  Speaker 1: economic growth, no matter what sort of shape size of
• 3:15
  Speaker 1: any country is. You know, how do you think about adoption?
• 3:18
  Speaker 1: What should countries be doing to encourage adoption? Does it
• 3:22
  Speaker 1: go back to the same sort of regulatory structures? Or
• 3:25
  Speaker 1: how do you think about adoption as a key driver
• 3:27
  Speaker 1: for growth?
• 3:30
  Speaker 2: Yeah, I think you want to have a lean forward,
• 3:33
  Speaker 2: try new technologies approach to new technologies. Otherwise, as opposed
• 3:38
  Speaker 2: to be somewhat stuck in the past. Naturally, new technologies
• 3:45
  Speaker 2: need some encouragement and I think should be embraced. We
• 3:51
  Speaker 2: are going to see and are seeing, in fact, significant
• 3:54
  Speaker 2: productivity gains from artificial intelligence and we'll see very dramatic
• 4:01
  Speaker 2: gains in productivity from robotics. Things like the Tesla self-driving
• 4:07
  Speaker 2: car is going to be a tremendous boon or is
• 4:10
  Speaker 2: a tremendous boon already to users and I think humanoid
• 4:17
  Speaker 2: robotics will be just an incredible change. I think, just
• 4:21
  Speaker 2: to give you some sense of scale here, I think
• 4:26
  Speaker 2: AI will probably increase the global economy by 20 to
• 4:33
  Speaker 2: 30 percent. That's my rough estimate, meaning on the order
• 4:37
  Speaker 2: of 20 to 30 trillion dollars per year. And And
• 4:41
  Speaker 2: AI will be able to do anything digital, anything that
• 4:44
  Speaker 2: does not require shaping of atoms by hand, probably by
• 4:49
  Speaker 2: the end of next year. So as I'm sure people know,
• 4:53
  Speaker 2: AI is already incredibly good at software. And it's getting
• 5:01
  Speaker 2: to the point where And it won't just be good
• 5:04
  Speaker 2: at software. It'll be what I call Stockfish-level good. So
• 5:07
  Speaker 2: Stockfish is a chess program that can beat the world's
• 5:10
  Speaker 2: best chess players very easily. In fact, at this point,
• 5:15
  Speaker 2: you could run Stockfish on your phone and beat Magnus
• 5:17
  Speaker 2: Carlsen at chess. So at some point next year, is
• 5:23
  Speaker 2: my prediction, software will be so good, AI software will
• 5:27
  Speaker 2: be so good, that it will be Stockfish-level good, meaning
• 5:30
  Speaker 2: that it is impossible for a human to compete in
• 5:32
  Speaker 2: writing software anymore.
• 5:34
  Speaker 3: With AI. AI will just crush all humans as software.
• 5:38
  Speaker 2: And I think it will be extremely good, possibly stock-based level,
• 5:44
  Speaker 2: but certainly extremely good at all forms of engineering and
• 5:48
  Speaker 2: anything digital. Literally in 12 to 18 months.
• 5:55
  Speaker 1: One of the stats.
• 5:57
  Speaker 3: Actually, let me just add one more thing to that.
• 5:58
  Speaker 3: My apologies.
• 5:59
  Speaker 2: So 20-30% increase of total global economy. So this is
• 6:05
  Speaker 2: quite a lot of prosperity we're talking about here. Um, uh, just,
• 6:09
  Speaker 2: just from digital AI, but from robotics, uh, from, um,
• 6:15
  Speaker 2: from humanoid basically think of it like a general purpose, robot,
• 6:17
  Speaker 2: robot AI. I think we'll see many multiples of the
• 6:21
  Speaker 2: global economy, meaning like you can increase the economy by
• 6:26
  Speaker 2: a factor of 10 or more. These are mind boggling numbers.
• 6:31
  Speaker 1: Yeah, I was just about to say that. I mean,
• 6:33
  Speaker 1: one of the stats we gave to the group here
• 6:34
  Speaker 1: was that in the four years since sort of ChatGPT
• 6:38
  Speaker 1: was launched, I think there's over a billion people around
• 6:40
  Speaker 1: the world already using AI. There's been very quick uptake
• 6:43
  Speaker 1: around the world on that. I guess the question to
• 6:47
  Speaker 1: you is I think a lot of people think about
• 6:49
  Speaker 1: sort of the next chapter being in this sort of
• 6:51
  Speaker 1: applied or sort of physical AI space. Where do you
• 6:54
  Speaker 1: see sort of robotics trending? How quickly is it going
• 6:56
  Speaker 1: to be implemented? I think I remember in the first
• 6:58
  Speaker 1: Trump administration, we were talking about sort of automated factories,
• 7:02
  Speaker 1: and now we're a year later, 10 years later. So
• 7:05
  Speaker 1: how quickly do you think these changes are happening in
• 7:08
  Speaker 1: the sort of physical AI world?
• 7:11
  Speaker 2: Yeah, so anything physical always takes longer than anything which
• 7:13
  Speaker 2: is digital. Digital, you know, when you solve something digitally,
• 7:17
  Speaker 2: it's the software that you can easily copy across other computers.
• 7:22
  Speaker 2: When it's physical, you've got to build up an entire
• 7:24
  Speaker 2: massive supply chain. You've got to move a lot of atoms.
• 7:29
  Speaker 2: And supply chains are very much global at this point.
• 7:32
  Speaker 2: So that's why it takes longer. Nonetheless, when you think
• 7:36
  Speaker 2: of humanoid robotics, the way you think about it, I
• 7:41
  Speaker 2: think the right framework is to consider that the usefulness
• 7:43
  Speaker 2: of a humanoid robot, a general-purpose robot, is going to be.
• 7:48
  Speaker 3: Roughly the digital....
• 7:51
  Speaker 2: The AI software, how good is the AI software times
• 7:54
  Speaker 2: how good is the AI chip in the robot times
• 7:57
  Speaker 2: how good is the electromechanical dexterity, especially of the hands? Now,
• 8:02
  Speaker 2: all three of those things are improving exponentially. And the
• 8:06
  Speaker 2: usefulness of the robot is those three things multiplied by
• 8:09
  Speaker 2: each other. Then when you make the robots, the robots
• 8:12
  Speaker 2: will start manufacturing the robots. So you get a recursive effect.
• 8:21
  Speaker 2: So it starts off very slowly, but then it grows
• 8:26
  Speaker 2: at an explosive rate. So if you say like 10
• 8:30
  Speaker 2: years from now, I would say there are well over
• 8:32
  Speaker 2: a billion humanoid robots.
• 8:35
  Speaker 3: Wow.
• 8:37
  Speaker 2: And the productivity per robot will be probably five times
• 8:40
  Speaker 2: that of a human, meaning that the Productivity in 10
• 8:47
  Speaker 2: years of humanoid robots.
• 8:49
  Speaker 3: And I think this is a conservative estimate, by the way.
• 8:52
  Speaker 2: This is one I would be willing to put serious
• 8:56
  Speaker 2: money betting on.
• 8:58
  Speaker 3: That there will be at least a billion robots in
• 9:00
  Speaker 3: 10 years.
• 9:01
  Speaker 2: And that those robots will be at least five times
• 9:04
  Speaker 2: the output of a human. Meaning the billion humanoid robots
• 9:08
  Speaker 2: will be more productive than all humans combined.
• 9:14
  Speaker 3: Wow.
• 9:15
  Speaker 1: Shifting gears just a little bit to a question that's
• 9:19
  Speaker 1: kind of facing the U.S.
• 9:20
  Speaker 3: Today.
• 9:21
  Speaker 1: Data centers have been a big political issue over the
• 9:25
  Speaker 1: last six to eight months here in the United States.
• 9:28
  Speaker 1: And I think there's a general understanding that in order
• 9:32
  Speaker 1: to drive and power the AI revolution that's coming, we
• 9:35
  Speaker 1: need to have the electricity and the data center compute
• 9:40
  Speaker 1: capacity to do the training and the inference of all
• 9:44
  Speaker 1: this AI. You know, how do you think about this
• 9:47
  Speaker 1: particular issue? Where are we on the curve of kind
• 9:51
  Speaker 1: of how much we've built versus how much we need?
• 9:53
  Speaker 1: And as government leaders here think about how to prepare
• 9:56
  Speaker 1: their economies and build the right power and data infrastructure,
• 9:59
  Speaker 1: how should they be thinking? about where sort of compute
• 10:01
  Speaker 1: and power needs are going to be in the next
• 10:03
  Speaker 1: few years?
• 10:06
  Speaker 3: Well, there actually is quite a crisis of power. So
• 10:11
  Speaker 3: this is a.
• 10:13
  Speaker 2: In fact, this is something, if you just sort of
• 10:17
  Speaker 2: follow the AI topic on the X platform, which, by
• 10:21
  Speaker 2: the way, is where almost all of the AI discourse
• 10:24
  Speaker 2: takes place, I think you get a very good sense
• 10:28
  Speaker 2: for where things are headed. So that's how I get
• 10:33
  Speaker 2: my news, and it's incredibly good. Everyone who's anyone in
• 10:36
  Speaker 2: AI posts on X. So that's why I'd recommend just
• 10:41
  Speaker 2: following just go on the AI topic on X and
• 10:44
  Speaker 2: you'll understand all these things and get a day-by-day account
• 10:48
  Speaker 2: of things. And what the consensus is at this point
• 10:54
  Speaker 2: is that there will be a significant power shortfall next year,
• 10:59
  Speaker 2: so not like distant future. There's expected to be, I
• 11:03
  Speaker 2: believe the consensus estimate among analysts that follows the AI
• 11:08
  Speaker 2: space very closely is that there will be at least
• 11:09
  Speaker 2: a 15 gigawatt shortfall of power in 2027 for AI chips.
• 11:18
  Speaker 2: So this is perhaps an obvious thing that one would
• 11:23
  Speaker 2: expect to occur because the rate at which AI chips
• 11:27
  Speaker 2: is being produced has been rising incredibly rapidly. They're sort
• 11:31
  Speaker 2: of rising on the order of 40% to 50% a year.
• 11:34
  Speaker 2: But the power available outside of China has been rising
• 11:38
  Speaker 2: at like 10% to 20% a year. So obviously the
• 11:44
  Speaker 2: faster rising thing will eventually overwhelm the slower rising thing.
• 11:47
  Speaker 2: And so, in fact, even I'd say at this point,
• 11:51
  Speaker 2: there are challenges with power even before next year, which
• 11:55
  Speaker 2: is why Google and Anthropic and many other companies are
• 11:59
  Speaker 2: actually leasing compute from SpaceX because we've been able to
• 12:03
  Speaker 2: turn on AI better than anyone else so far. But
• 12:08
  Speaker 2: this is by constructing our own power plants, right? is
• 12:11
  Speaker 2: the only way we were able to do it. So
• 12:14
  Speaker 2: now China does have a tremendous amount of electricity, but
• 12:18
  Speaker 2: due to GPU export bans, one cannot establish data centers
• 12:25
  Speaker 2: with the latest chips in China. So really the consideration
• 12:29
  Speaker 2: is what sort of electricity growth is there outside of China?
• 12:34
  Speaker 2: And that is currently a significant shortfall relative to AI
• 12:38
  Speaker 2: chip production. So this creates an opportunity, I think, for
• 12:43
  Speaker 2: countries around the world just to say if they're interested
• 12:48
  Speaker 2: in AI data centers, to construct a lot of power.
• 12:54
  Speaker 3: And offer that to AI companies.
• 12:58
  Speaker 2: And in exchange, of course, these data centers would be
• 13:02
  Speaker 2: taxed and have to pay reasonable fees and stuff. But
• 13:08
  Speaker 2: it does create an opportunity for a lot of countries.
• 13:10
  Speaker 3: Absolutely.
• 13:11
  Speaker 1: Well, thank you so much for your time. It meant
• 13:12
  Speaker 1: a lot that you could join us here and really
• 13:15
  Speaker 1: appreciate it. Thank you so much.
• 13:16
  Speaker 3: You're most welcome. Thank you.
• 13:18
  Speaker 1: Thanks for listening.
• 13:19
  Speaker 3: See you in the next episode.