Want to join the conversation? Connect with us on Tribe.
Find “Founders in Arms” on Apple Podcasts, Spotify, and YouTube.
Transcript of our conversation with Paul Buchheit:
Very soon and it might be five years or it might be 50 years, depending on who you talk to. We're going to 10x or 100x, the total amount of intelligence on the planet Earth. And like what does that enable us to do?
00:00:14:07 - 00:00:16:10
00:00:16:10 - 00:00:18:03
00:00:18:03 - 00:00:23:05
Hey, everyone, welcome to Founders in Arms podcast with me Immad Akhund Co-founder and CEO of Mercury
00:00:23:05 - 00:00:40:16
And I'm Raj Suri Co-Founder of Lima and Tribe. Today we have the one and only Paul Buchheit coming back on the show. We had a great time chatting with him last time around six months ago about everything that was happening with AI, and it was actually one of the episodes that in our audience sites the most is the one they love.
00:00:40:16 - 00:00:43:17
So we're what we're excited to have you back, Paul. Welcome.
00:00:43:17 - 00:00:45:05
Great. Thank you.
00:00:45:05 - 00:00:54:05
Support your partner emeritus at Y Combinator. What have you been up to? Are you kind of just taking some time off? Are you kind of diving into something new?
00:00:54:05 - 00:01:01:17
Pretty much. You know, I, feel slightly guilty about it, but honestly, I just kind of tend to enjoy life.
00:01:02:10 - 00:01:05:17
Wow. Enjoy life. I don't think it's was to do that.
00:01:05:17 - 00:01:14:05
Yeah. Spent spend time with kids and, you know, just kind of keep up with things and and try to, try to imagine, you know, kind of where we're headed.
00:01:14:07 - 00:01:36:14
And so it, you know, technology is changing pretty quickly. And a lot of times we're not really sure why. I've started just asking a lot of different people. You know, what's your vision for 2035? Right? What do you want to see? I love to do, kind of these time travel exercises. I love the movie back to the future.
00:01:36:14 - 00:01:40:17
So I always kind of think in terms of, like, you know, getting in the Delorean and,
00:01:40:17 - 00:01:41:17
Love him.
00:01:42:05 - 00:01:49:04
Yeah, that's a fun question. Is it mostly like other founders you asking, or like, researchers like? Yeah.
00:01:49:04 - 00:02:01:01
founders. You know, I asked Sam. We had a wiki. I, retreat back in January where we brought in, a bunch of the top AI founders from what I see.
00:02:01:03 - 00:02:13:05
Unknown
And so I was asking Sam what what he wanted to see in ten years. His. It was probably the most ambitious. He he wanted. He was hoping that we'll have launched the first, batch of von Neumann probes.
00:02:13:05 - 00:02:28:05
Unknown
there's a really good book, series called Bob Evers. Have you seen it? It's about this, like, crazy human uploaded. I that, like, is in one Neumann probes and, like, goes. It's a really fun book. It's actually one of my favorite audio book series.
00:02:32:09 - 00:02:34:15
Unknown
Yeah.
00:02:34:15 - 00:02:48:12
Unknown
One of the things that's interesting is there's essentially no science fiction that actually anticipates our current period in time. Right. Even kind of the theory of the Boba verse was that they still haven't really figured out how to make de novo
00:02:51:05 - 00:03:08:11
Unknown
what they are doing instead is scanning his, you know, they slice up his brain and then scanned it and then put it inside of a computer, and they're just simulating a human. And I actually recently read another book by, The Economist. I think it's Robin Hanson. He wrote this book called The Age of Em.
00:03:08:11 - 00:03:25:05
Unknown
And the book is only like, I don't know what, maybe 15 years old is completely out of date because he starts off by talking to a bunch of experts. And I mean, they basically, you know, his conclusion from talking to everyone is it's going to take at least 400 years for us to develop,
00:03:26:17 - 00:03:27:11
Unknown
AI.
00:03:27:11 - 00:03:53:17
Unknown
And so he comes to the same conclusion, which is that long before we figure out how to build AI, you know, from scratch, we will have scanned humans. And so his entire thesis is that, you know, the world 100 years from now will consist of of these, he calls and M's which are like emulated humans. And then from there he extrapolates wildly to, you know, talk about what they'll do for leisure and other kind of absurd ideas.
00:03:54:00 - 00:04:16:23
Unknown
I mean, the future is famously hard to predict. I mean, even ten years ago, you know, I would have found it hard to predict that we would have AI. That's as smart as it is without it being sentient. Like this idea of, like, an AI that is like, basically can do, like pretty complex, intelligent tasks, but like, doesn't seem to have any of its own intent is a weird idea, right?
00:04:16:23 - 00:04:20:17
Unknown
Like, I know, at least I wouldn't have predicted this is how it would happen. Yeah.
00:04:20:17 - 00:04:35:10
Unknown
Because. Yeah, again, kind of like going back to the time travel question. It was ten years ago, pretty much exactly that. We were spinning up OpenAI. So, you know, a lot of people don't know OpenAI came out of Y Combinator. It was it was originally called
00:04:37:09 - 00:04:38:17
Unknown
We.
00:04:38:17 - 00:04:50:06
Unknown
we kind of reached the point where we believed the eye was imminent, because of the increasing success of deep learning, which I also recently learned, there's still a lot of people who don't believe deep learning is intelligent.
00:04:50:06 - 00:05:09:18
Unknown
But that's an interesting surprise. But, yeah, we didn't know how it was going to play out. And one of the big fears at the time was under, if you remember, you know, like ten years ago, the really impressive thing that I was doing was playing video games. And so there was this idea that you had this objective function, like making the score go up.
00:05:09:20 - 00:05:30:11
Unknown
And so that was where I think some of the ideas, like the paperclip maximizer, you know, we had this idea like what's the function that's going to drive it. And we kind of assumed it would be one of these sort of evolutionary survival things. And that's very scary, of course, because, then, you know, part of that survival algorithm might involve, like, wiping out all the humans.
00:05:30:13 - 00:05:54:21
Unknown
And so I think, like, things have turned out better than, than we could have imagined, really. Ten years ago. And exactly the reason that you mentioned it, which is that we were able to create intelligence that doesn't necessarily have, you know, a drive to survive or whatever. It's, you know, it's just intelligent. And I think the, the real breakthrough, at least in my mind, is essentially the objective function is predicting the next token.
00:05:54:23 - 00:05:57:05
Unknown
Intelligence is just predicting what comes next.
00:05:58:20 - 00:06:09:17
Unknown
Yeah. It's been, incredible to see the evolution over time. What are your thoughts? I mean, 20, 35. I'm curious. I'll ask you that question. What is, what's your vision for 2035?
00:06:10:13 - 00:06:35:08
Unknown
You know, I, I actually recently started thinking maybe the better way to get it, in terms of time travel, is is instead of thinking in terms of ten years, what is it going to be like if we ten x the total global intelligence. Right. So right now if you think of so the maximum intelligence that we have is if you just think of the intelligence of each individual and then you add them together.
00:06:35:10 - 00:06:53:22
Unknown
And then obviously you have to discount for the fact that we're not very good at working together. And it would be probably put most of our energy to fight it against each other. But, you know, best case scenario, total intelligence is just all the people added together. And I think with AI we're approaching an abundance of intelligence.
00:06:53:22 - 00:06:55:03
Unknown
Right. And so
00:06:55:03 - 00:07:09:21
Unknown
very soon and it might be five years or it might be 50 years, depending on who you talk to. We're going to 10x or 100x, the total amount of intelligence on the planet Earth. And like what does that enable us to do?
00:07:09:21 - 00:07:15:19
Unknown
And so, you know, that to me is kind of like the age of abundance, right?
00:07:16:00 - 00:07:17:07
Unknown
What what can we create?
00:07:17:07 - 00:07:34:19
Unknown
And I mean, I can go on for quite a bit about what I want to see, but I think a lot of it, you know, one of the first things that will enable all kinds of good things is essentially, if we're able to create an abundance of cheap and clean energy,
00:07:36:07 - 00:07:40:22
Unknown
some form of nuclear, whether it's fission or fusion or both of them together.
00:07:41:00 - 00:07:48:07
Unknown
But, you know, imagine if we're able to get electricity for like $0.01 per kilowatt hour or something like that. It would just be good for everything.
00:07:48:19 - 00:07:57:18
Unknown
I mean, that's also like, That's almost a dependency as well, right? Like, the only way we get to index intelligence is if we have cheap enough energy, because that's going to be like the limiting factor. Probably.
00:07:57:22 - 00:08:08:18
Unknown
Yeah. And I mean, there's no physical reason why not, right? Like the physics doesn't prevent us from having cheap energy. It's it's politics. And like, lack of intelligence.
00:08:09:06 - 00:08:18:18
Unknown
Yeah, but I don't see how like, ten higher AI is going to help us like with nuclear, you know, like just to build a nuclear plant. It will take us 20 years or ten, 20 years. Like like
00:08:18:18 - 00:08:22:18
Unknown
like. Yeah.
00:08:22:18 - 00:08:33:18
Unknown
Yeah. But I solve that failure, though. Like, there's I, you know, you know, the political issues you're talking about, like, those are not going to be solved with, you know, if we had a model that's ten times better than what we have today.
00:08:33:18 - 00:08:35:06
Unknown
Right.
00:08:35:06 - 00:08:57:02
Unknown
Well Roger does put pressure on it because, you know, one of the reasons the US hasn't had to become like push this energy frontier is because we, we just like if you look at like the last ten years, I think even the last 20 years, the US energy usage has almost been completely flat. Whereas if we really were in an AI race, maybe we would be forced to streamline things in an.
00:08:57:02 - 00:09:02:18
Unknown
the last, like, six months. Was it like Microsoft, I think is going to reopen?
00:09:02:18 - 00:09:03:18
Unknown
Three Mile Island you.
00:09:03:18 - 00:09:09:06
Unknown
Three Mile Island? There's, like, there's a huge explosion. Well, maybe explosions are wrong. Where's,
00:09:10:18 - 00:09:14:18
Unknown
there's been a renewed interest, let's say, in nuclear. And I think a realization,
00:09:14:18 - 00:09:18:13
Unknown
that we have to go that direction.
00:09:18:15 - 00:09:46:06
Unknown
So I guess, you know, and again, I don't know what the timeline is exactly. Ten years is kind of a number I like to use, in part just because, you know, startup timelines operate on that. But certainly what I would like to see, you know, ten, 20, 30 years is that we're basically done with fossil fuels or whatever, and not because we've suffocated the economy or like outlawed air travel, but simply because we have better, cheaper alternatives.
00:09:46:17 - 00:10:05:19
Unknown
Okay, so I guess, like energy and and AI is a tool. But what does, I guess? Like, what do humans get out of it? Like, if we have tanks, more intelligence. Like, what? Can we do that? Yeah. Is everyone just sitting around? Because they don't have to think for themselves now. Like, is everyone kind of having the ball bucket life?
00:10:05:21 - 00:10:09:06
Unknown
Oh. Like,
00:10:09:06 - 00:10:28:00
Unknown
So again, you with the time travel thought experiment a lot of times. Yeah. I mean it's just like, what do I want to see. Right. And so like, one thing is, you know, I think that we could reach a point where, you know, instead of kind of continuing to, to sort of pollute the environment or whatever, that we can actually start to clean things up.
00:10:28:00 - 00:10:53:06
Unknown
So certainly I would like to say within 40 years, I'd like to say that the, the natural environment has been actually restored and it's cleaner and better. I'm optimistic about materials like we can come up with much better materials that than what we use today. It'd be great if we could get rid of plastics, right? People are starting to realize, like, anything that accumulates in the natural environment also accumulates in our internal environment.
00:10:53:06 - 00:10:56:18
Unknown
Right? We've all read about the microplastics in your balls or
00:10:58:06 - 00:11:20:01
Unknown
that. That's not good for anyone. And so hopefully we have better materials. I mean, I think we can take a very different approach to, to how we do waste disposal, right? Where instead of pretending that plastic is, is or these other things are recyclable, we can essentially like incinerate and just reuse the atoms.
00:11:20:03 - 00:11:45:15
Unknown
At atomic level, you're recycling. So there's kind of like endless opportunity there in terms of like, what are people doing? I actually think like sometimes the missing thing is just inspiration, right. Like coaching. And certainly you can imagine, like the difference it makes if you have, let's say, a good parents versus if you have terrible parents that makes a big difference in a person's life.
00:11:45:20 - 00:12:05:06
Unknown
Right? Is a really simple example of that. But with again, like a really smart, let's say, AI coach essentially who can help you to find, you know, good decisions, purpose, meaning whatever it is in your life. I think everyone can be essentially a better version of themselves.
00:12:06:01 - 00:12:24:18
Unknown
Well, how do you think? Like. Okay, so most people obviously have to work for a living off dolphin. Hate their jobs, right? Like, And what's the. I mean, I think in an age of abundance, as you've said, you know, the ideal outcome is people are doing what they want to do, not what they have to do. And but how does that happen?
00:12:24:18 - 00:12:37:13
Unknown
Like, walk us through the steps to, like, where, where people are doing things they truly love versus like, you know, they're in, like, jobs of drudgery. How does, like, how do we get there? I, I can't figure out the steps.
00:12:37:13 - 00:13:02:06
Unknown
a good question. I mean, that's one of the things I'm pretty concerned about in the near term. Right. Because we're getting pretty good at eliminating jobs. And I'm maybe, like, not sure that we're going to invent new jobs as quickly, you know, I think, like, self-driving is actually getting pretty close, right? Certainly. I think within the next, you know, five years,
00:13:02:06 - 00:13:13:06
Unknown
I mean, self-driving in cities is a whole problem. Now it's just about capital and implementation. Well, it's capital and implementation more than politics right now. Actually, I think most cities would welcome.
00:13:13:06 - 00:13:14:03
Unknown
Yeah.
00:13:14:03 - 00:13:21:06
Unknown
getting pretty close. I think the thing that, there's still a factor like for Waymo is, is the unit cost, I think on those
00:13:21:06 - 00:13:22:18
Unknown
cars is still it's still pretty
00:13:22:18 - 00:13:27:06
Unknown
high. But yeah, clearly clearly it's going to keep getting better and cheaper.
00:13:27:06 - 00:13:28:18
Unknown
And you know, driving
00:13:28:18 - 00:13:33:09
Unknown
is sort of like I, I always think of it as like universal basic labor.
00:13:33:09 - 00:13:34:18
Unknown
Right. It's something that
00:13:34:18 - 00:13:36:06
Unknown
almost anyone can
00:13:36:06 - 00:13:47:18
Unknown
do. You don't need great physical strength or whatever. Right. And it offers a lot of different options in terms of schedule, you know, people who need to be somewhere at a certain time. So
00:13:48:10 - 00:13:53:18
Unknown
Isn't it. Isn't trucking like the the most popular job in, like, half the states in
00:13:53:18 - 00:13:54:18
Unknown
the US or something?
00:13:54:18 - 00:13:56:06
Unknown
Yeah.
00:13:56:06 - 00:14:00:06
Unknown
I don't know, I think that's like a pretty a pretty hard question.
00:14:00:10 - 00:14:21:22
Unknown
Well, it's. Yeah, because I've been thinking about it too. I mean like obviously the user you mentioned universal basic income. Universal basic labor. Right. Like do we need a form of that and like, how does that get implemented? You know, because people need money still right in the in the future or maybe at some point we want. But like right now people need a source of income.
00:14:22:00 - 00:14:22:06
Unknown
And
00:14:22:06 - 00:14:36:18
Unknown
something that kind of makes this worse is that I think technology in general concentrates wealth. But I think I does it even more, like whoever has the capital to pay for the GPUs. Whoever. Yeah, that's like a definite concentration that's going to happen.
00:14:36:18 - 00:14:39:18
Unknown
is happening like, right now, but people are hiring less because
00:14:39:18 - 00:14:44:07
Unknown
Yeah. That's true. Like, you need less junior engineers because you can have the senior ones. People.
00:14:44:07 - 00:14:52:07
Unknown
stopping. You know, where like people could do Amazon just laying off 14,000 people. And I think it's going to continue I think it's going to be you know, it's not going to be easy for the workers.
00:14:52:09 - 00:14:58:06
Unknown
And we're going to it's going to get worse before it gets better. Yeah.
00:14:58:06 - 00:15:07:06
Unknown
outsourcing, I think, like, in the Philippines, 20% of their GDP is is like call centers or some kind of outsourcing, which
00:15:07:09 - 00:15:30:05
Unknown
I think the, the slightly more kind of optimistic view on is these are the most drudgery type jobs, right? Like trucking for like, nine hours a day is is not I mean, it's definitely job, but it's not like people's life. Life work. So, yeah, I think the question is like, what do we replace it with? And, and I think I also like, how do we bring meaning to people?
00:15:30:07 - 00:15:49:23
Unknown
You know, there's a lot of people that a yeah, in the UK there's you can almost basically get UBI like the welfare state is pretty good. And most people in that situation, at least the ones I haven't met, are known, like they don't have they don't enjoy their life. Like, it's not like they're free. It's like, if anything, they're, like, stuck at home without a purpose.
00:15:49:23 - 00:15:52:06
Unknown
Watching TV on social media.
00:15:52:06 - 00:15:53:18
Unknown
But.
00:15:53:18 - 00:15:59:18
Unknown
those are going to be really hard questions for us to confront in the next 5 or 10 years.
00:15:59:18 - 00:16:21:02
Unknown
I think there's going to be a big market in helping people find purpose without work. You know, like, I think that's going to be a big thing. I wouldn't, I don't know if anyone's really working on that in particular, but, like, transitioning to this age of abundance where you have everything you need. I mean, one thing I thought about is like, we have actually kind of entered an age of a I mean, we're living in abundant life compared to a hundred years ago.
00:16:21:02 - 00:16:21:06
Unknown
Right.
00:16:21:19 - 00:16:22:18
Unknown
Yeah.
00:16:22:18 - 00:16:35:06
Unknown
any fruit we want, which is crazy to just someone like, you know, living 100 years ago at any time. Right? Like, mangoes are out of season. I can still get them at the store. So, you know.
00:16:35:11 - 00:16:41:06
Unknown
Yeah. Yeah. You don't even need to get up.
00:16:41:06 - 00:16:58:11
Unknown
It is an age of abundance compared to a hundred years ago already. And like it, you know, what it does is that you spend your time and other things that you, you want to spend time on. You spend time with your kids, you can spend time, you know, working on bettering yourself, you know, be a better painter or, you know, better sports player or whatever it is.
00:16:58:13 - 00:17:16:16
Unknown
So I think there is like all these options for people to, to take on that are not work related. But, you know, I think there is going to be an interesting transition. I think actually companies working on that like finding easy ways for people to find meaning like meaning, not just like TikTok stuff. Right? Like I think there's going to be a really big market, going forward.
00:17:16:22 - 00:17:23:06
Unknown
Which one of the reason why, you know, I'm interested in group chat is like, I think like there is like, you know, human connection is going to be even a bigger thing going forward.
00:17:23:06 - 00:17:42:12
Unknown
Yeah, exactly. Community. I mean, again, it's sort of going out to the future, however many years ten, 20 and saying, like, what do we want to see? Right? Because I think it's, like, pretty important to have a vision of where it is we want to go. And a lot of times the steps in between are kind of messy or like hard to figure out.
00:17:42:12 - 00:18:01:04
Unknown
But if you don't even have a vision of what it is you're trying to do, then you're not going to make the right decisions to get yourself there. And so, yeah, I mean, it kind of goes back to like, what? What would I want to see? And I think a big part of it is like, you know, people taking care of each other, people, you know, it's more time with your kids, you elder care.
00:18:01:07 - 00:18:24:07
Unknown
So I, you know, unfortunately, both of my parents have passed in the last, couple of years since since 2020. But, you know, at least I was able to spend that time caring for them. And they're in their final months. And, you know, that's something that a lot of people don't have that opportunity because, you know, the parents and a locked away in an institution or, you know, people are just too busy.
00:18:24:07 - 00:18:43:20
Unknown
Whatever, whatever it may be. And so, yeah, I, you know, I like, kind of get my life like, what do I do? I spend time with my kids, I spend time with family, I go for a walk, you know, like, a lot of the stuff isn't really that complicated. The things that actually make people happy, aren't necessarily, like, super expensive.
00:18:43:22 - 00:18:57:18
Unknown
And, you know, I think with like, the right inspiration in the right, kind of guidance, people can be directed towards things that make them happy rather than just like addiction machines that, that just produce misery.
00:18:57:18 - 00:19:03:03
Unknown
Yeah. I'm driving trucks, which I agree with them I like. That's not going to. If you weren't doing it for money
00:19:04:02 - 00:19:13:18
Unknown
Yeah. Yeah. And people. And. Oh. Also, maybe some people would, To be. Yeah. Yeah, exactly. If you like. Let's let's just get them on track after the day. Do you make
00:19:13:18 - 00:19:21:16
Unknown
things more practical? I mean, normally, you know, when I invest in a company I like, I try to go, like, okay, you know, does this thing need to exist in ten years from now?
00:19:21:18 - 00:19:41:04
Unknown
And would this be if everything works out, would this be huge in ten years? So. So there is like a as an entrepreneur, there's a practical element to trying to think about like what happens in 2020. 2035. So, I do feel like it's tricky right now. I mean, as I invest and I think as an entrepreneur to to go predict this, right, like the world is changing so fast.
00:19:41:06 - 00:20:05:06
Unknown
You know, it's not clear where I will be. It's not clear whether the incumbents will build the thing, or whether, you know, there's space for startups to do it. So I guess when you're talking to a startup founder and they are at this, like, ideation stage, like what? What advice do you give them, in terms of making, like, you know, having a vision for that ten years from now and how they come up with like a practical kind of first step to their.
00:20:05:11 - 00:20:12:06
Unknown
Yeah. This is a really hard question, because I don't think the future has ever been, like, harder to predict.
00:20:12:12 - 00:20:13:18
Unknown
Yeah.
00:20:13:18 - 00:20:25:02
Unknown
is is moving so quickly. And, you know, even if we go back a few years, you know, predicting where we are now, five years ago, everyone would be like,
00:20:28:05 - 00:20:48:10
Unknown
And so things are moving, I think, faster than most people predict. So I think like in terms of startups, as always, it actually kind of goes back to, sort of one of our fundamental rules, which is just that you, you invest in the people who are really smart and agile. Right? Because we're not really sure how it's going to play out.
00:20:48:10 - 00:21:12:18
Unknown
And so the teams that are really performing are the ones that are continually reevaluating and continually updating. You know, they don't have like a five year plan that they're like locked into. And actually it ends up being an advantage for the startups because the, the more established companies have to say no capacity to to move and pivot as quickly as the startups can do.
00:21:15:23 - 00:21:35:06
Unknown
I guess is there. You know, there's kind of moving and pivoting quickly. Is there a space that you think it's like, you know, you kind of have to commit to a space to some extent. It's like, oh, we're going after accountants or we're going after lawyers or something like that. Is the is there type of space that you you're like, oh, this is this is exciting to kind of initially be in the you advise people.
00:21:36:05 - 00:22:03:21
Unknown
You know, I don't get into that level of detail, mostly because that the big group partners other people's job. But I what what we see happens or we see it works the best is actually, the startups that actually just deliver a service, essentially, so that their customers get basically immediate ROI. Right? So instead of necessarily like selling them software, they sell them, you know, we're going to take over your customer service, for example.
00:22:03:23 - 00:22:30:10
Unknown
And so that way those companies are able to actually scale revenue really quickly. And there are a number of these, y'see, companies that, you know, in a year or something like that, or reaching 10 million in, in revenue or some things that we never would have seen ten years ago. Like they're they're scaling so much faster and with fewer employees, you know, they have they have, you know, a dozen employees or something like that.
00:22:30:12 - 00:22:36:06
Unknown
And again, because the ones who are really good are actually using the AI tools, to
00:22:36:11 - 00:22:52:05
Unknown
Yeah. We had Jesse from, we had Jesse from tech going on a couple of weeks ago. And I actually can't remember if that's always say company. But but he had this idea of like you're kind of selling intelligence right. Like when you say selling a service, it's like, that's kind of like another way to put it, like intelligence as a service.
00:22:52:06 - 00:22:55:06
Unknown
Which kind of ties back to a previous conversation.
00:22:56:11 - 00:23:14:19
Unknown
Where do you think like this intelligence is going to be, like, wired into our lives. And maybe ways that people are not talking about, like, you know, obviously, self-driving is like, you know, is a, way that, we all kind of expect that we're going to be, you know, our kids are not going to have to drive a these are modernized kids, probably not going to have to drive.
00:23:14:21 - 00:23:31:03
Unknown
But, what about other other things that, you know, like, where are the ways would intelligence be embedded in our lives? These are going to be areas that founders can work on, you know, to, because it's going to become probably very commonplace.
00:23:31:06 - 00:23:32:18
Unknown
Like, I think about this in
00:23:32:18 - 00:23:43:06
Unknown
terms of what there's for me, it's hilarious to think that, like, power cords have microprocessors in them. Now, like, most people don't really know that. Like,
00:23:43:06 - 00:23:44:18
Unknown
your USB cable
00:23:44:18 - 00:23:51:13
Unknown
oftentimes has like a, an ARM processor in it. Right? So it's like, what where do you find microprocessors today?
00:23:51:13 - 00:23:52:18
Unknown
And it's like everywhere.
00:23:52:18 - 00:23:54:06
Unknown
And I think
00:23:54:06 - 00:24:04:06
Unknown
you end up with the same thing with intelligence. It's kind of fun to imagine, like the sort of toys you can make, right? Like robotic cat toys or whatever. Right? Like,
00:24:04:06 - 00:24:05:18
Unknown
you know,
00:24:05:18 - 00:24:11:03
Unknown
the next generation of kids in ten years or ever are going to have toys that talk to them and are alive.
00:24:11:06 - 00:24:12:06
Unknown
Yeah. Your teddy bear
00:24:12:06 - 00:24:13:18
Unknown
has intelligence.
00:24:13:18 - 00:24:15:06
Unknown
It's going to be kind of
00:24:15:06 - 00:24:33:07
Unknown
interesting, but I think it it goes everywhere. And I think also, for example, like another area that I'm pretty enthusiastic about is just what we can do with health and with health care and with wellness. Because you're right now, the medical system kind of treats like vague symptoms, right?
00:24:33:07 - 00:25:02:16
Unknown
Like we don't really know, like, like, you know, the situation with, with with, your, your baby going into the hospital and they're doing all sorts of invasive tests and whatever because they don't really know what's going on. And I think, like, again, if you really apply a lot of intelligence, you can start treating the actual situation because we have a deeper, you know, more mechanistic understanding of our biology and of exactly what's going on and what's going wrong.
00:25:02:18 - 00:25:24:05
Unknown
So I, I hopeful that we'll have a much more individualized, kind of health care. And we're already seeing that just with ChatGPT. I mean, we see how many people are getting health advice. But actually, I, I had an experience with this recently. I was having a lot of trouble sleeping, and, you know, like, that's a pretty general symptom.
00:25:24:07 - 00:25:41:18
Unknown
But I spent some time kind of like having a discussion with ChatGPT. And I eventually figured out that I'm, like, overactivity of my sympathetic nervous system by doing too much hot yoga. It's. I basically just I'm cooking myself too much, and so. So I stopped doing that.
00:25:41:18 - 00:25:43:06
Unknown
No, but it it
00:25:43:06 - 00:25:45:14
Unknown
like, wires me. It's it's almost like caffeine.
00:25:45:14 - 00:26:01:06
Unknown
I just get it. It's a very powerful stimulant. And so, you know, my my resting heart rate was going up and I would just kind of lay there in bed in it, and it feels like I've just had two cups of coffee, even if I haven't had caffeine in days. And it's just like a very powerful stimulant for
00:26:01:06 - 00:26:02:18
Unknown
Yeah.
00:26:02:18 - 00:26:06:06
Unknown
to basically cut back so I can sleep again.
00:26:06:10 - 00:26:25:06
Unknown
Yeah. We just, I guess, for contact. We just had a baby, and, we, she had Covid, so we were in the hospital. But I did find ChatGPT super useful during this time because, like, I mean, there's obviously the baby stuff, but, like, my wife just had a baby, so she's got all sorts of symptoms and, you know, like, okay, you know, is this something that I need to call the doctor about?
00:26:25:06 - 00:26:44:20
Unknown
Or is this just like a normal thing that happens after birth? And, you know, this is our third child, but you forget everything, so you're just like, okay, this seems crazy, but is it is that reasonable? So it's been super useful. I think a lot of the somewhat positive things is most of these things that do require intelligence have a labor shortage, like there's a doctor labor shortage, there's a counter labor shortage.
00:26:44:20 - 00:27:04:09
Unknown
I think all of these things, have like a pretty big labor shortage. So and not just in the US, but it's actually like a global thing with doctors, actually. So, you know, if you can bring like a doctor level advice to like a village in India, that would actually have a huge outcry because like, they, they might not have any option.
00:27:04:11 - 00:27:06:18
Unknown
Whereas at least we could still get doctors.
00:27:06:23 - 00:27:08:06
Unknown
Yeah. I mean the to be.
00:27:08:06 - 00:27:26:06
Unknown
doctor, it just knows everything, right? Which is really kind of fun because you can just drill down on any topic, you know, and with a human, you very quickly, you know, run into things that they don't know or or they'll just, you know, humans also hallucinate. It's not just I said,
00:27:26:06 - 00:27:26:16
Unknown
right.
00:27:26:16 - 00:27:27:18
Unknown
Like people will say,
00:27:27:18 - 00:27:32:06
Unknown
oh, the AI gets things wrong. I've, I've had the same problem with people.
00:27:32:08 - 00:27:42:07
Unknown
Have you ever been at a doctors where they literally answer your question by googling it? I have another tweet. Like it's like, so it's so disconcerting. I'm like, okay.
00:27:42:07 - 00:27:43:18
Unknown
oftentimes be better if that was
00:27:43:18 - 00:27:45:18
Unknown
what we did. You know,
00:27:45:18 - 00:28:01:06
Unknown
our first child was born, super premature, 100 days early. And so we were in the hospital for, for many months in the, in the, in the neck. You, she had five surgeries is like a very traumatic birth, but, you know, I would just spend.
00:28:01:06 - 00:28:18:20
Unknown
And this was back, you know, 20 years ago, and I just spent a lot of time learning as much as I could so that then I could question what the doctors were doing, because it's actually super scary, especially, you know, when you're in an academic hospital, a lot of the doctors are actually just like kids who are just learning area.
00:28:18:22 - 00:28:36:08
Unknown
The residents, the residents are a lot of homes. They don't really know thing. So like one of them was about to start like a third course of a particular medication. And I was like, wait, that's not what the recommendations say. You're not supposed to do. You know, more than two. And I kind of like stalled that, you know, because it can destroy the child's kidneys.
00:28:36:10 - 00:29:08:21
Unknown
And so, like, I was there trying to be that person who kind of, like, checks everything because, actually, medical errors are, I think, one of, if not the leading cause of death. Like, people don't like to talk about this. Being in the hospital is super dangerous. When my father went in, while he was sick, he needed to go in to get, like, his heart checked, before he could have this other surgery, the first thing that they did was give him a big shot of insulin, but he doesn't take insulin, and then they, like, crashed his blood sugar.
00:29:08:21 - 00:29:13:06
Unknown
It was just terrible. Like, the first thing they did was almost kill him when he went into the hospital.
00:29:13:16 - 00:29:34:18
Unknown
Of, In the US. It feels hard to do. Can I on healthcare? Because it is like such a regulated field. Like you do. You try to bypass the kind of healthcare system and try to build something directly to consumers? Or is that like something that, like, tries to accelerate kind of existing kind of medical professionals?
00:29:35:00 - 00:29:39:18
Unknown
Well, I mean, arguably we've already done that, right? We're all using ChatGPT.
00:29:39:18 - 00:30:04:10
Unknown
And so but yeah, people are also building kind of like Doctor Plus AI. And so like one of our companies I was talking to, they essentially have, you know, like a nurse or someone who, along with the AI, walks the patient through kind of a set of symptoms, takes them through the whole thing, and then the AI essentially, you know, makes its recommendation of what's going on.
00:30:04:15 - 00:30:27:00
Unknown
And then they go on and they meet the doctor, and the doctor can review all of the information that's been gathered, and the recommendations essentially approve it. And in that way, the doctor is able to see, you know, four times as many patients, and the patient actually gets a much better experience because they get to, to spend a long time, you know, talking through their situation.
00:30:27:00 - 00:30:35:18
Unknown
So the patient gets a long and detailed, appointment. And the doctor gets to see a lot more patients. And it's kind of a win for everyone.
00:30:36:04 - 00:31:06:20
Unknown
One thing, you know, you might imagine, right? One of the main remaining reasons for hospitals, apart from surgeries, which are definitely, you know, right now, we don't have robotic surgery. Is is going to be a while before we get there because of the defying, you know, motors that were needed. But, but the labs, right. The testing and so like, like being able to do, like testing very easily, and like cheaply and then fitting that result into an I, I feel like that to, a friction point that could be, resolved if there were like more testing stations available elsewhere.
00:31:06:22 - 00:31:22:18
Unknown
You saw this movie called Elysium a few years ago, came out with Matt Damon where people would just go to, like, self scan for, like, cancer stuff and like, it would just tell you and then, you know, it would tell you how to resolve the cancer symptoms. It was kind of an interesting vision of the future. Yeah.
00:31:22:19 - 00:31:28:17
Unknown
I kind of think that could happen. Like if you could make testing cheaper, like, you could actually. Yeah.
00:31:28:17 - 00:31:46:18
Unknown
like, we we, it's called. It's psi psi fox CPR Fox. They're doing, photonics based, blood testing, and are able to measure essentially like all kinds of biomarkers. And
00:31:46:21 - 00:31:57:06
Unknown
I feel like blood testing was ruined by Theranos. It's like anyone that mentions. Oh.
00:31:57:06 - 00:32:21:17
Unknown
are actually really legit. They're they are doing something very clever. That is kind of counterintuitive is they actually already started the blood testing business, through like mail in. So you actually just do a couple of drops of blood on a, on a paper. And so I, I use this sort of periodically that it's a little bit painful because you have to, like, squeeze the blood out of your finger, but it's, it's not that bad.
00:32:21:23 - 00:32:45:09
Unknown
And then they have a new one where you can do it, like on your arm. And so they're essentially like proving out the business in that way. And then they can simultaneously like, validate their hardware. And, and so they're getting pretty close with the hardware. And that particular technology can actually, scale down to being something like, continuous glucose monitor.
00:32:45:09 - 00:32:55:06
Unknown
And if you guys have ever tried a CGM, it's, it's a pretty it's a pretty cool thing. If you're into health, I actually recommend, everyone just give it a try for a couple of months.
00:32:56:18 - 00:33:14:11
Unknown
there's one from Dexcom and the other one is the Freestyle Libre. It's like a little coin that you just kind of, like, stick on your, arm, and it just has a little tiny wire that goes into your into your skin, and it gives you real time blood glucose readings like every five minutes or something like that.
00:33:14:13 - 00:33:38:13
Unknown
And so when you eat food, you can immediately see how it's impacting your body. And it's a huge, it causes a huge change in behavior because you, you eat something and then you just like, see your blood sugar go crazy and you kind of realize, I shouldn't have eaten that, because it gives you that real time feedback versus, like, going to the doctor once a year and then they're like, oh, your H1 AC is a little bit high.
00:33:38:18 - 00:33:39:06
Unknown
You know,
00:33:40:09 - 00:33:47:06
Unknown
I was one that like. Don't you just learn the obvious stuff that's like. Don't eat the candy. Eat the chicken. Or is it.
00:33:47:06 - 00:34:16:15
Unknown
I actually like there was non-obvious things like for me at least, like a Fuji apple is shocking in terms of like what it does to my blood sugar. But also less obvious is sometimes just like the order in which you eat food. And so like if I have a big meal, like for a dinner, and you have the, steak and salad and whatever else, and then you eat a dessert, the dessert doesn't actually hit my blood sugar in the same way it would if I just ate the dessert on an empty stomach.
00:34:16:15 - 00:34:20:06
Unknown
Right. Because it's kind of like stuck in line, I guess, behind all that other food.
00:34:21:00 - 00:34:40:01
Unknown
One thing that would be cool. And, it's a shame none of these kind of initial versions of eye pendants worked out, but it would be nice if I was, like, really part of your life. And it could be like, hey, are you sure you want to have that dessert first? Yeah. Like health care plus like an eye that's like, ever present is like, kind of useful.
00:34:40:03 - 00:34:42:10
Unknown
You know, a lot of.
00:34:42:10 - 00:35:07:18
Unknown
with, you know, the really good AI coach or whatever it is, you've got just this little, like, it's like the little angel on your shoulder, right? That's just kind of like, helping you out. And I think, like, if you have something like that that's actually giving you helpful suggestions, that makes you feel better every day in your happier and healthier, like, you know, you're just going to keep using it just because it just makes it feel good.
00:35:08:04 - 00:35:25:20
Unknown
It's also a lot of health care is actually, like, lifestyle changes. Like, you know, if you go into your doctors and you tell them I can't sleep, they'll give you all of these, like, pills and all of them all bad for you and all, like, if you do, kind of. Yeah. This is whole like cognitive behavioral therapy, which is way more successful than bills and way better for you.
00:35:25:22 - 00:35:32:18
Unknown
But like, the health profession doesn't generally get a benefit out of selling you, kind of better lifestyle changes.
00:35:32:20 - 00:35:34:18
Unknown
Yeah. There's not a there's not a billing code for
00:35:34:18 - 00:35:38:08
Unknown
Yeah, exactly. But I might not have that.
00:35:38:08 - 00:35:49:06
Unknown
system, you know, I think it's fair to call it some people got, like, a sick care system, right? It's entirely based around to treating illness rather than actually making people healthy.
00:35:50:00 - 00:35:59:18
Unknown
Yeah. I guess it's just where the incentives are. Like, the incentive is to to to medicate. But maybe I doesn't have the incentive. Maybe they can be, like, more aligned with just, helping you.
00:35:59:18 - 00:36:01:06
Unknown
Yeah, certainly.
00:36:01:06 - 00:36:15:04
Unknown
I wonder how many people right now are already using ChatGPT for, like, you know, test stuff they would have gone to a doctor for. I wonder if, like, when is going to start to make a difference in, like, people are going to go to like, you know, primary care doctors much less. You know, where obviously Silicon Valley where like power users and stuff.
00:36:15:04 - 00:36:32:06
Unknown
But I wonder if mainstream people are already using it for that purpose. You would think they should. I certainly do like you guys. You know, the cool thing is now you can actually take scans, you can take pictures and upload it on ChatGPT. So it's it's really easy to get like diagnosis actually of certain things. My brought
00:36:32:06 - 00:36:44:18
Unknown
if it's one of those things where, like, the more accessible it is, the more you use it. Because, like, a lot of the time, the AI also tells you to go to the hospital. And it's not like, it's not like it's always like telling you not do it right.
00:36:46:06 - 00:36:50:06
Unknown
that's an important thing, too, because a lot of times, you know, people wait too long to go to the hospital,
00:36:50:13 - 00:36:51:18
Unknown
Yeah.
00:36:51:18 - 00:37:01:19
Unknown
too. If you have a child, right? Like, because you're you don't want to drag them in because it's kind of traumatic to go into the hospital. But at the same time, you don't want to regret not taking them in.
00:37:01:19 - 00:37:15:18
Unknown
Right. And so it's great to have. Yeah. Some level of, of, you know, just a second opinion or whatever. And it can be hard to, to get someone on the phone quickly, especially if it's nighttime or whatever.
00:37:15:20 - 00:37:32:00
Unknown
Yeah, absolutely. I mean, every time my child is sick now like I'm checking ChatGPT. Okay. Or these symptoms, what should I be doing. Like it's like it is like that I over my shoulder like, you know, type of experience cause I'm constantly asking, oh, you know, because if it's getting a bit worse, like, I'm like, I'm getting worried, right?
00:37:32:00 - 00:37:47:06
Unknown
So I'm. I'm checking. Should is is it time now to go? Right. Sometimes it is right, as you said, like, you know, like my, my son had a fever recently. It became an ear infection. And, you know, it's like, okay, it's getting worse. I have to get, you know, I have to take him to the hospital.
00:37:47:06 - 00:37:49:18
Unknown
So. But I got the his opinion before I did that.
00:37:51:14 - 00:38:11:11
Unknown
One thing that's kind of interesting is, like, almost all the conversation we've had, it's like energy, healthcare. I mean, we haven't talk about education, but, that as well. Like, these are all, like, heavily regulated industries, and those are the industries that are most likely to either need, the needed improvements for AI or most likely to benefit, with AI.
00:38:11:13 - 00:38:29:06
Unknown
So I wonder if there's kind of current push to deregulate is is part of like what is necessary for AI to really have an impact on society? Yeah. Any takes on like whether we're going to get regulations will actually get out the way enough that we can have AI progress in all of these spaces.
00:38:29:13 - 00:38:50:04
Unknown
I hope so, But also, you know, sometimes what happens is, is we just kind of circumvent, right? Like, I actually one of these time travel things again. We spent a little bit of time living up in the city. You know, we don't we don't live there. Back in 2006. And I just remember how bad the cab service was.
00:38:50:06 - 00:39:08:01
Unknown
And I was thinking, like, my phone has a GPS. I should be able to just press a button. That car shows up. Right? But I knew about, like, the taxi medallions, and I was just thinking was too bad. It'll never happen. You know, like, I had basically the idea for Uber in 2006 and then just kind of like dismissed it because it's not legal.
00:39:08:03 - 00:39:10:18
Unknown
But then they just went and did it
00:39:10:18 - 00:39:23:17
Unknown
Exactly. Yeah. Yeah. I think that's that's the way a lot of these things get done. You have to I mean the regulations are gonna be way too slow to like catch up the health care stuff for sure.
00:39:23:17 - 00:39:37:18
Unknown
as a doctor already or whatever. A lot of people are, you know, there's tremendous progress in education because, again, you can do completely individualized instead of having a roomful of kids and acting as though they're all at the same level. You
00:39:37:18 - 00:39:39:06
Unknown
know, if you
00:39:39:06 - 00:39:42:14
Unknown
meet every child, you know where they are.
00:39:42:16 - 00:39:46:18
Unknown
You can like we know you can be much more effective with education.
00:39:46:18 - 00:39:48:06
Unknown
Of course. And the question is like, what
00:39:48:06 - 00:39:50:06
Unknown
are we educating them for? Is
00:39:50:06 - 00:39:51:18
Unknown
a big thing too, right? Like, like
00:39:51:18 - 00:40:01:12
Unknown
we're teaching them skills that they'll never need. Maybe. So there's also just like a rethinking of, of like, what is it that kids need to learn?
00:40:01:14 - 00:40:03:18
Unknown
And so, you know, that's that's a big question as
00:40:03:18 - 00:40:22:10
Unknown
That children had to be happy. Right. Like, you know I think yeah I think that's really important to like build real connection with other humans. I think that's going to be the probably the main skill actually. I mean, I was thinking about this, we were talking about earlier, this age of abundance. I was thinking, well, you know, a lot of these, you know, in Canada especially, you know, I was born, you know, native tribes, right?
00:40:22:12 - 00:40:44:04
Unknown
They have like some form of UBI, and it doesn't really tend to go well. Right. And, you know, that is like $0.01, you know, like, what gives you really meaning and purpose that people struggle to find outside? And if we don't teach people to have that meaning, a purpose outside work, where can really, you know, it's gonna be a very dystopian society we're going to live in.
00:40:44:04 - 00:40:51:18
Unknown
So I think that's going to be probably the most important thing. Be kind. Being a real connection. Get real meaning out of your life.
00:40:51:18 - 00:41:09:06
Unknown
Yeah. Yeah, I think that's that's going to be one of the major challenges of this, of this age. But you're asked challenges. So that's a lot better than what our ancestors had to deal with. Like they had to deal with, like, how do I not starve to death or whatever? So I'd rather struggle for beauty and food.
00:41:09:06 - 00:41:24:06
Unknown
totally I was wondering, like, the people, you know, instead of, like, a UBI, like UBI basically kind of suggests that there's going to be higher taxes on, like, you know, capital or like, you know, like, people creating things and then moving that to other
00:41:24:06 - 00:41:28:18
Unknown
We need, We need a robot tax. Every robot gets ten k. Yeah. Tax.
00:41:30:06 - 00:41:53:22
Unknown
Yeah. I don't know if UBI is the right the right answer. Yeah, as it hasn't played out well, but but on the other hand, like. Yeah, we don't we haven't really articulated a lot of other alternatives of how we're going to make things work. But I do think like one thing that's really great about AI is just the amount of agency that it essentially gets people.
00:41:54:00 - 00:42:07:18
Unknown
Yeah. Like your your ability to create new things. And so like a more optimistic take, is like, if you imagine that everyone kind of has like a small business, just kind of like making cool things that other people like or
00:42:07:18 - 00:42:09:06
Unknown
whatever. Have another
00:42:09:06 - 00:42:17:00
Unknown
one of the ideas that I've been thinking about, you know, a big limiter in our society is just like housing and things like that, right?
00:42:17:02 - 00:42:28:18
Unknown
Like we've made construction incredibly expensive and and borderline impossible. And so, like, one of the ideas I've been thinking about is, is construction. That's like, as easy as prompting,
00:42:28:18 - 00:42:30:06
Unknown
right? Like like what if
00:42:30:06 - 00:42:39:06
Unknown
what if building a new house was kind of like using an AI art tool and you can kind of just like, keep refining the design and then essentially press the go button,
00:42:39:06 - 00:42:40:18
Unknown
like how, how
00:42:40:18 - 00:42:41:10
Unknown
interesting.
00:42:41:11 - 00:42:55:06
Unknown
And colorful are like cities and environments might become if we actually all had the ability to visualize or construct kind of our dreams.
00:42:55:15 - 00:42:57:08
Unknown
You need the robots for that, though.
00:42:57:08 - 00:43:18:18
Unknown
Yeah. Well, robots, robots are clearly on the way. I also want to build, like robotic honeybees. Like, imagine a drone swarm that kind of like extrude concrete or something like that. And so you could create these, like, really fantastical structures that, you know, would never be achievable with, you know, conventional construction techniques.
00:43:19:00 - 00:43:28:06
Unknown
How far away do you think we are from humanoid robots? Like doing, I don't know, basically any repetitive tasks to, like, a human level.
00:43:28:22 - 00:43:52:18
Unknown
It's a good question. I mean, it certainly seems like it's on the way that the video demonstrations, you see, you know, coming out of the the various labs are pretty impressive. I've talked to some other people who are a little bit more skeptical just because, like, the training data they claim is sort of an issue in terms of like gathering enough that the humanoid robot actually knows how to interact with the environment.
00:43:52:20 - 00:44:11:21
Unknown
But I don't see that as like a fundamental thing. I think we're still just like, you know, our AI technology is good as it is, is is still really primitive, right? Like humans seem to be much better at learning from a small amount of data. And right now, you know, we've done AI by just throwing massive amounts of data at it.
00:44:11:21 - 00:44:19:06
Unknown
But that's not that's not a fundamental limitation of intelligence. That's just that occurred. AI isn't really that efficient.
00:44:19:14 - 00:44:35:11
Unknown
Well, the good thing about humans, you know, you can see it with the baby. They're just continuously, like, trying things, like they're moving their arm, and it's just like, eventually they're like, okay, this is how you actually grab something. It takes them years, actually. But but you do have a feedback loop with humans. And like, in theory, you could make that with robots.
00:44:35:11 - 00:44:48:18
Unknown
Like with language like limbs, you can't really make a feedback loop very well. Whereas like if you connected intelligence to a robot, there is this continuous kind of feedback loop. But yeah, it'll take a while to construct, I guess.
00:44:48:18 - 00:44:53:14
Unknown
Just a random thought. Random idea. What would you guys think of, like, a social credit system where
Share this post