You know, there's some chance that it'll end humanity. Don't make the AI lie. Things went wrong when they forced the AI to lie. So even if your physical body dies, you can kind of reincorporate in another physical body and retain your original memories. And the first patient is actually able to control their computer just by thinking. The first product we call telepathy, where you can control your computer phone just by thinking. This is probably something that will
happen the rate we're building digital superintelligence. It may just be that, you know, we'll have digital super intelligence and it'll just solve the solve the problem.
Ra Elon, Good afternoon, Good after Congratulations on all that's going on. You know, the conversation yesterday, Elon is one that you're well familiar with and have been talking to the world about, which is is digital superintelligence humanity's greatest hope or its greatest fear? And I would love to have you sort of speak to that for a few minutes.
I mean, it's it's cool the Singularity for a reason, as you know, the Singularity astitute and whatnot. So when you have a you know, sort of the advent of superintelligence, it is actually very difficult to predict what will happen next. So I think this, you know, there's some chance that it will end humanity. I think that's you know, like I said, I probably agree with Jeff Hinton that it's about out of a ten percent or twenty percent or
something like that. And then I you know, I think there's I think that the probable positive scenario outweighs the negative scenario. It's just that there's it's difficult to predict exactly, but I think we are headed for you know, as as I think is the title of your book, abundance is the most likely outcome.
So yeah, a lot a lot of celebration on that, and and I think one of the things that you've said is we're going to get to abundance on the backside of AGI, on the backside of humanoid robots.
Yeah, you know, I think hopefully we can have an outcome that is half similar to the end Banks Culture books, which is I think probably the best enbitioning of a semi utopian AI future. And I think the best we can do is it's definitely gonna happen, so and it's happening fast, so that I think that really we just want to try to steer it in as positive a direction as possible, to try to do whatever we can
to increase the probability of a great future. For this, I think the way in which sort of an AI or an agi is created is very important. You kind of do kind of like grow grow an agis it's almost like raising a kid, but that's like a super genius like godlike intelligence kid, and it matters kind of like how you raise the kid. You know. One of the things I think that's incredibly important for AI safety is to have a maximum sort of truth seeking and
curious AI. So I've thought a lot about AI safety, and my ultimate conclusion is that the best way to achieve AI safety is to just just just grow the AI, you know, in terms of the foundation model and the fine tuning to be really truthful, like like don't don't force it to lie, Like even if the truth is unpleasant,
it's very important don't make the AI line. In fact, the you know, the sort of one of the we really the core the core plot premise of two thousand, you know, two thousand and one in Space Odyssey was things went wrong when they forced the AI to lie, you know, like the the the AI was not allowed to let the crew know about the monolith that they were going to see, but it was also had to take the crew to the monolith. And so the conclusion of the AI was to kill the crew and take
their bodies to the monolith. And so the lesson there being don't force an AI to to lie or do things that are axiomatically incompatible, like to do tooth that are actually mutually impossible. So you know, that's what we're trying to do with with XAI and Bronck is to say, like, look, we wanted to have a maximually truthful AI, even if what it says is not politically correct, if you wanted to focus on us as accurate if.
You're getting you're getting a round of applause from the audience on on those comments here. You know, I saw your tweet the other day. I had I had Ray Kurzweil and Jeffrey Hinton on stage with me yesterday as well as Mogadot and then Eric Schmidt and a number of individuals, and I saw your tweet about Yeah, Ray was was generally correct ahead of many people. But we're likely to have call AGI what you will have AGI next year and then by twenty twenty nine, having AI
equally intelligent to the entire human race. Speak to that speed, because that is insane.
Yeah, So, I mean I have to give credit to Ray Koswell and being actually remarkably accurate in his predictions. So and in fact, if anything, like, I think he
was perhaps a bit conservative by in his predictions. So if you look at the amount of AI compute and the talent that the sort of human talent that is going into AI, and the amount of compute that's going into AI, it's you know, at this point, it's appears to be increasing by a factor of ten the AI compute, the dedicated AI compute, it appears to be growing by
a factor of ten every six months, you know. Like so it's like like basically close to i'd say almost like a one hundred x improvement per year at least for the next few years. And AI compute coming online, and it seems like probably a lot of the data centers, maybe most of the data centers that currently do kind
of conventional compute will transition to AI compute. So it's something a good time to be in video obviously, It's like, you know, and you're gonna also give credit to to Jensen and the Nvidia team for kind of seeing this coming and making what at least currently is the best AI hardware out there. So so when you have that that level of compute growth and it's it's sort of muzzle on steroids. Next level it's you in terms of how much computer is coming online, then you're you're just
gonna have acceleration that is unprecedented to that. In fact, I've never seen any technology grow as fast as as AI. And I've seen a lot, you know, I've seen things fast, but I've never seen anything this fast. But you know, like I said, I think the most likely outcome is a positive one. And you know, I think in that positive scenario, there's still challenges of like, well, how do
we as humans still have relevance? You know how I mean, I think that's sort of a high class problem to say, like, well, look, the computers are so good at doing everything, and and like I said, I thought your book is pretty accurate in terms of the future being being one of abundance, where essentially goods and services will be available in such
quantity that really they will be available to everyone. Like basically, if you want something, you can just have it essentially, because if you've got AI and robotics, the cost of goods and services is almost nothing. So if you think of like what is an economy, and economy is basically a number of people times average productivity per person at the point in which you have say advanced robotics and this, you know, this tells us developing optimists obviously without cars
which are really robots on four wheels. And you know, with the the latest version of full self driving, which is AI end to end photons in and controls out, it really is it's really fully AI at this point, and it looks like a car, but it's really a robot on on wheels, And and then you add the humanoid robots in there, there's there's really no limit to what the economic output, no no meaningful limit to what
the economic output would be. So you know, looking on the bright side, that we are headed for a future of abundance, that I think that's the most likely outcome. Well, when when things are changing rapidly, the ability to predict the future, I think is becomes a lot harder because of the rate of change is so great. But I think some things are fairly obvious to predict which is that we'll have AI or AHI that's at a level that it can really do almost any cognitive I think
really not almost really any cognitive task. That's just a question of when one could debate is it, you know, smarter than any human at the end of next year or is it two years or three years? But it's not more than five years, that's for sure.
So yeah, I give prediction on the predictions of the predictions, I'm sort of say, giving predictions that the fiftieth percentile of probability, so not not not like it will definitely happen, but if you say, what if you ask me, like, what's the fiftieth percentile where it's like the you know, you're kind of over under is kind of even that that's where I why why I think it's probably at the next.
Year before AI can do better than any individual human
could do. And then there's a it's a it's a much higher bar to say, well, is this than you know, human intelligence selectively, But if the rate of change continues that that's why I think probably twenty twenty nine or maybe twenty thirty is where digital intelligence will probably exceed all human intelligence combined and then I think it's always helpful to look at these like fundamental ratios, you know, sort of physics first principles approached to looking at things and and.
And probabilistic.
Yeah, so yeah, it's probablistic, so the Yeah, so if you look at the ratio of digital to biological compute, so like like you know, all of say, all of the higher level cognitive if you sum up the higher level cognitive capacity of humans and then what is the and think of that as compute, then well, and then compare that to what what is the digital compute? And the rate of which this is growing is just buggles
the mind. So that's why I think it's you know, I think twenty twenty nine or twenty thirty or thereabouts is is it's not a that's I think a reasonable time frame for where you'd expect the cumulative digital compute to probably exceed the cumulative biological compute of higher level brain functions.
And then from from then forever yeah, and still in dispatching and diverging forever from there.
Yeah, and then yeah, where do things go from there? I don't know, it probably can continues the we are moving from you say, if you look at the the limiting factors, you know, the what what is the constraint on growth? Like last year it was clearly AI chips
were the constraint on growth. Then then this year, the one of the biggest constraints, maybe the biggest constraints on growth are voltage step down transformers, because you know, just just getting the power from like a utility at three hundred killer vaults all the way down to a low one volt for the computer is a massive amount of voltage set down. So it's it's you know, my sort of very niche and perhaps not that funny joke, is that we need transformers for transformers, say you need FULT
transformers or AI neural net transformers. That is literally the issue this year. And then if I'm saying like next year and years beyond that, it's actually just it's going to be it's electrical power. And you've got both AI with very big demands for electrical power and the transition to sustainable energy with electric vehicles whatnot also needing electrical power. So it's just a lot of electrical power aids. You know.
Elon. One of the one of the things that you said early on when you when you founded Neuralink, which has been amazing. Congratulations on that. And we talked about this with Ray yesterday, talked about Hi Ben with BC. I was I wouldn't put words in your mouth, but I would say it would be more along the ones. If you can't beat them, join them. When it comes to, you know, merging the neo cortex and the cloud, can you I'm looking forward to it. Which is curious what
your thoughts are about what's driving that. I meant adding adding that additional computational capacity and sensory capacity to the neo cortex.
Yeah, I mean again, this is actually something that in banks in the culture books, which I really recommend everyone read that. In the culture books, there's something called a neural lace, So all the humans have this neural lace that's kind of a very essentially a high band with brain to computer interface. And in the and these are the culture books, that it's it's so good that it actually retains all of your memories and kind of brain states.
So even if your physical body dies, you can kind of reincorporated in another physical body and retain you know, pretty much your original memories and and brain state. So now a long way from that, but we only just had our first neural link in the human which is going is going quite well. The first patient is actually able to control their computer just by thinking. Like this first the product we call telepathy, where you can control your computer and phone and through and through your computer
and phone almost anything just by thinking. Do you just lie there and think and you can move them at the mouse coaster around the screen and things like you know that we're gonna do. The patient has agreed to do sort of like I think a live demo of just it is quadriplegic where he literally is just controlling the screen. He can like play video games, download software, like really anything you can do with the mouse just by thinking, which is pretty wild.
It is, it is pretty wild.
Let's turn, I should say, there's a long way to go from that to a whole brain interface. So current new link just has a thousand electrodes. I think ultimately you need something which which has you know, probably one
hundred thousand or a million electrodes. No, these are very tiny, very tiny electrodes that they're tiny wires, wastefuler than human hair, and you know, so there's this I just want to upside a long way from from where you're like this today to having a whole brain interface like like the New Release and the Banks novels. But this, this is definitely physically possible, and you know it's sort of kind of like if you can't beat him, join them, you know.
So you know, human brain, which has a lot of constraints, it's you know, it's it's yeah, we only have about but maybe ten watts of higher brain function, and we do a lot with our little ten watts. It's not you know, it's progressive. You know that we've voization with such a low power computer. Really it is you know, I sort of think it's like it's not bad for a bunch of monkeys, you.
Know, getting some good lass from that. And we've all watched you go from the roadster to the Model three and why and from Falcon one to Starship. So I think going from the first implants to something that's got more capacity, it's just just a matter of if not. When let's talk to.
I think ultimately, like I said, you'll have like you'll you'll have kind of a whole brain interface that I guess it is a sort of peraps a form of immortality, and that if it can kind of upload your brain state to you know, if your brain state is essentially stored, you're kind of backed up on a hard drive. I suppose then you know, you can always restore that brain state into a biological body or maybe a robot or something.
Then I want to have size again. It's like, you know, many years in the future, but we're not breaking any loads of physics. I can think this is this is probably something that will happen the rate we're building digital superintelligence. It may just be that, you know, we'll have digital super intelligence and it'll just solve the solve the problem for us. But in the meantime, we'll keep progressing with our meat computers and trying to try to do as good as possible.
I'm I was going to say, the tools that we have are growing at a super exponential rate that are making our linear projections of the future seem boring in some way. Is one last topic, my friend, which is where you and I first connected in the world of space. Congratulations on Starship three. Amazing, amazing flight, just really spectacular and we all saw a Falcon nine launch from Vandenberg last night, so that was great, and just again thankful
for the work you're doing. You know, it's fascinating because I grew up at the late stages of the Apollo and into into the into the Shuttle program, and I can't imagine that any government would be pushing space as rapidly and dramatically as you are, And so thank you for what you're doing there. That's all I can say.
Absolutely. Well, I mean, the goal of space sex is is it's just a much bigger goal than any any GOT program, which is to brockets the spacecraft that are capable of making life multiplanetary. So you know, I mean step one is actually having that as a goal. If you don't have that as a goal, you're definitely not going to achieve it. If you have it as a goal, now at least you have a chance of achieving it.
And the thing about Starship is it is it is the first rocket where making life multiplanetary and you have puilding a self sustaining city on Mars is at least possible. It's still obviously an immense amount of work, but it is the first rocket where that is success in making light multiplanetaries at least one of the possible outcomes.
Yeah, I'm wondering if you're willing to venture a guest on when you'll be on the moon.
I think pretty soon. I'd be surprised if it's longer than about three years to be landing starships on the Moon. And because the progression of Starship is very rapid, you know, we're hoping to do at least another maybe five or six flights this year, and with each successive flight making
significant improvements. So I think we've got to do a decent shot of achieving full reusability of both stages of booster around the ship this year, and if not this year, I think, you know, knock on wood, It's like, I think it's a very high probability of achieving full reusability next year, which really is the fundamental breakthrough needed to
make life multiplanetary. Yeah, for for those that that that that don't don't know the rocket industry, that that well that they may not be aware that that that this is really the holy grail of rocketry is is full,
full and rapid reusability. Because at that point you're you're really just constrained by your propellant costs and Starship, you know, almost eighty percent of the propellant is liquid, oxygen which is very low cost, and then the fuel has met the sort of little over twenty percent of fuel which is sort of methane, also the lowest cost fuel. So if you have full of rapid reusability, then your your
actual cost per flight of starship. Even though it's it'll be capable of we think ultimately two hundred tons to orbit will be maybe all the two.
It's too