OpenAI's Sam Altman trusts AI will treat most diseases by 2035

We have said many times that in a well functioning society, you know, OpenAI would have been a government effort. I could say that about a lot of other industries too.

For longevity specifically, I'd say that the broader category is an interest in healthcare. And I'm interested in longevity for a bunch of reasons, but I remember once someone saying that most diseases are diseases of age. You know, 20-year-olds don't get nearly as sick as 80-year-olds. And so if we can figure out the general purpose thing here, maybe we can improve healthcare outcomes across the board. I had a passing interest in longevity for a long time and then when the partial reprogramming discoveries began to happen...

The idea that we could sort of rewind a cellular age a little bit, not all the way back to a stem cell, which, you know, you don't want to be turned back into goo, but we could like make your cells a little bit younger, right? So that seemed like a kind of crazy exciting discovery and I ended up making a large investment in a company called Retro to pursue that.

I think kind of all of the above. I was, as I mentioned, really interested in healthcare and is there like a... I think my whole kind of approach to startups and technology is can we identify like the crux issue for something. And then for healthcare, if we could truly keep you, you know, health-wise younger for much longer, that would address a lot of the health issues. And so rather than like kind of try to like fight disease after disease and problem after problem, maybe this was like, you know, the Gordian knot we could cut through all.

No, I don't like Steve. Yeah. By the way, I don't think a technology like Retro will make you live for... No, that's correct. I think you'll be healthier for longer and then a very steep... So, you just have healthier for longer, not live forever. I definitely don't want to get, you know, really sick for a long time.

At the time, there were not as many. This broad area of partial reprogramming seemed to me like the kind of thing that looked incredibly promising in the lab. Was one of the few approaches that I thought could really work like across a whole person. And seemed worth pursuing. I have become increasingly negative on the ability for academia to make things happen. I think academia is still really great at some things, early discovery, but it's like all right, if we really want to like take this idea and maybe make it into something that can really help people, maybe a startup is a good idea, right?

The goal is add 10 years of health span. So, you know, maybe it doesn't keep you alive any longer at all, but you're healthier for longer. And the approach that I am personally most excited about is give you some proteins, some RNA or DNA that will express those proteins, whatever it is, but somehow do something to your cells that will partially reprogram them back to a younger state.

Healthcare is one of the biggest uses of ChatGPT and with GPT-5 we made a big push to improve it for healthcare. People use it to understand their health all the time. They upload their medical records. They ask you questions about symptoms. You can like read a lot of stories on the internet about people using this in super important ways and obviously we think that's great. So with GPT-5 we did a big push to make it better at health related queries.

I've heard a lot of doctors and healthcare workers say that, you know, they hope that in 10 years everybody gets, and they think this, they literally mean this, everybody gets better healthcare than the best healthcare anyone can get today. You know, they imagine a world where you can start with an AI medical advisor service and you can say here's what's going on for me or hey, I'm just, this is my normal life. Here's my sleep data, what I'm eating, my blood tests, my symptoms. And you know, a service could early on say, 'Hey, you're at risk of this. You need to go test for this. You need to make this change. You need to go to this specialist doctor. This is a little weird. You need to go here.'

Mental health in general is something that I'm increasingly worried about with AI. Again, I think it can be done for tremendous good. But this is happening very fast, you know, like I think ChatGPT is like a little bit more than or coming up on three years old, right? And for the first year it was used as a sort of curiosity, maybe a Google replacement a little bit, right? Maybe like, you know, write my essay for school or whatever. Only with more recent evolutions did people start to use it as a life coach, sort of a therapist and you know close advisor, things like that. We have learned that a lot of people need someone to talk to and to give them advice. And I am worried that given how quickly this is moving that societal scale behavior could change in a way that affects societal scale mental health. And we are working super hard to get a handle on that and understand what's happening. But it is something I'm worried about.

No, it tends towards that way. I am pretty sure though that people even if they know the human doctor is going to perform worse than the AI doctor will significantly prefer the human doctor. I certainly would. I want my doctor to use all these tools. I want to use these tools myself. But there is like something very deep of... I have not had that many serious health issues but when I did have one I really would not have been happy like leaving my fate in the hands of an... I really cared about that human doctor and their humanness.

GPT-5 is the first time where there have been real small but real examples of this.

There was like a... I saw on Twitter a couple of days ago there was like a math proof and then someone said maybe we can do a little bit better and ask GPT-5 Pro to work on it and came up with a novel and better solution.

Yeah. Treat, mitigate, something like that. Right. Right. But I think we are, I mean again this research project with Retro is one example where it's like wow this is like this just pushed us years ahead in our research and I have to mostly go with what the biologists and the pharma companies say here. This is way out of my field of expertise.

Both speeding up the rate of discovery of new insights and then making the insights themselves. Yeah.

We are not that smart and biology is enormously complex. We can imagine kind of anything. We can have someone explain enormously complex things to us and then we can get it. But to discover these things requires a lot of brain power and in many cases maybe more than we have. If an intelligence, you know, a thousand times or a million times smarter than us can figure out really how all of this works and then explain it to us, I believe we'll be able to understand it even if it was too complex for us to figure out in the first place.

Yes. And then we'll go, 'Oh, right.' You and I can understand very complex physics, but it would have taken lifetimes of smarter people than us to figure it out.

Is in there somewhere. Yeah. What I think we can do for sure is really empower individual people to be able to advocate for better outcomes for themselves.

We have not seen the kind of rejection from the medical community that people told us we were going to get. In fact, every time I go visit like a clinic or a hospital or whatever, they're and I'm kind of asked, I'm like, you know, how's it going with us, whatever, they're like, don't tell anyone, but every doctor here uses this all the time and we just do it at home and, you know, please make a HIPAA compliant version so we can do it at work, but we all like we all love it.

Yeah, I think these are like somewhat different. I mean, I think that what we've been talking about mostly is giving consumers access to like healthcare good information is one thing about themselves in particular. Like a theoretical thing. This is like here people do it. A lot of people do it every day. And then there's this other category of can we use AI to do better longevity research with any of the companies or approaches that are working on it and I think those are pretty different. I mean the same intelligence matters for both of them. So AI will work for both of them but they're quite different approaches.

So, I find that obviously really distasteful. It speaks to how much people want something, you know? It's like everybody wants a magic pill. You tell people like sleep eight hours a night, get exercise, eat reasonably healthy, and don't drink so much. They're like, 'Nah, [__] you. I'm not going to do any of that. I want a pill.' And so in the search for this cheat code, people take all kinds of crazy stuff. Limited to no medical evidence.

I mean, it's all around. When the GLP-1s first came on the scene, long time ago, like many, many years ago, right, there was this compounding pharmacy that'll make it for you. I had no idea what a compounding pharmacy was. So they make it for me. They screw up the dosing and I do I don't remember the exact number, but a mega overdose. Yeah. And I had never taken before, so I had like no tolerance or whatever, right? And I like injected like a, you know, it's like a Sunday or something. Let's call it 11:00 a.m. 2 p.m. 3 p.m. start feeling really bad. Convinced I'm like really sick. I have my friends over for dinner that night. My now husband looks at me. He's like, 'You're dying.' I don't know. You're like, 'You got some crazy you're dying.' Yeah. And I hadn't put it together that it was this yet. Oh, wow. And I go to bed feeling horrible. Puking all night. Whatever. I now have kind of put together that maybe it has to do with this. Maybe and I you know my pulse is like 150 something like that like blood sugar unbelievably like just like fully crashing in the morning with the ER. Yeah. And then like over the next day, they gave me like a glucose strip, whatever. Over the next day, we kind of figure out, but there's nothing you can do. You just wait, right? And an interesting and unknown thing about this is the you or not that well known is it makes you not have a desire to eat. Take enough of it makes you have a desire, not have a desire for anything else. So, I spend and I'm usually like a very kind of like I need a lot of like I need like new stuff to think about. Yes, I know. I'm the same way. And I have these few days laying in this hospital bed staring up at a white ceiling, right? Thinking nothing, not looking at my phone, not wanting anything, right? Feeling terrible, right? But just no desire for anything, right?

One thing that is better about longevity is everybody gets old. And so there's like a mass mass market there. You don't have to worry about that part. I do think these things where there's only like, you know, 20,000 people in the world that have a disease, it's not so obvious about how to like make capitalism work for this. I think anything that we could come up with that would add 10 years of health span, we'll have the benefits of capitalism to help us make it cheap. And I think that's what I would do.

I mean, the version of this I think about this much more for AI than longevity, but for AI, yes, it absolutely has to be fixed. I think we're actually doing a really good job right now of distributing the benefits of AI quite broadly. But you can imagine a world where it doesn't go that way. And if you have a lot of capital and you put it all into your AI brain, that's like a huge advantage over people who don't.

I actually don't know the size of the chat health team. It's not tiny, but it's and it's growing fast. And then we have the scientific research teams.

I mean, again, it's like I would bet it's a top three use case of ChatGPT at this point. Like it's a significant thing. It is fair that it's only one of our goals. Like we also want to help people write better code and we want to have like a better kind of search engine and we want to help people be more productive. We also have it just entertain people whatever. So it's like we're not a health company. It's one of many things but it is a significant part of the human experience.

I think in all of those cases, people are confused about whether they want to think about it as like a tool or a creature. I think it's like this incredible Rorschach test and the way people talk about it says much more about them than it has anything to say about it. No one knows obviously and people can do it but you know if you're like I think it's kind of always really revealing to people hear people. How do you look at it? Um, I think of it as this tool. Like I think of it as it's going to be it's going to do like amazing things, but it is not going to be this like self-aware thing from the sci-fi movies. Let's say the next CEO of OpenAI is an AI. You know, I like it. I get to like write the prompt for like what I would like that AI to go off and do. And it may then go do crazy complex things. It may look to the world like it's paving these highways and doing whatever. But it was still like some person had to ask the system to go like accomplish some goal.

Yeah. Now again, it could be so complex over such a long time horizon. So many like side quests along the way that it doesn't feel like that.

Ethics already runs behind. I was gonna say I have some hope that AI will help us figure out better like win-win systems. You know, I think we need new ideas in the world. And clearly we're not coming up with them on our own. So that'd be my positive case there.

Well more generally than just healthcare although this is it fits really well into this. I think the prompt that most people want some version of is help me flourish in accordance with my goals, whatever those are. Maybe help me figure them out, but once I figure them out, help me flourish in accordance with my goals over a very long time horizon. And that is flourish mentally, physically, everything.

The ice bath people are so annoying to me. They're so religious, smug about it. Yeah. Like they take such pleasure in like telling you, 'Oh, my ice bath is, you know, 36 degrees. I sit in for 45. Mine's 35. I sit in for 50.' And they're like so convinced it works. Maybe it does, but there was this study recently that it doesn't. Yeah. And I just sent it to like all these friends that brag about how miserable they are. Yeah. And they don't like it. Yeah. Not the fasting people. They drove me crazy. There was a whole period in Silicon Valley where I had to go to dinner parties with fasting people and then I just sit there and eat the steak or whatever it happened to be.

I am much I very skeptical. I'm much more focused on the downside. I can see ways where that's horrible. Yeah. I want to be like mistreated in a simulation for all of eternity. Right. For all of eternity. Right. Yeah.

No, I like my body. I like the world. Like I...

There's like gradations, right? There's like some version of that where I would augment. Suddenly you play piano, something like that. But oh man, would I be skeptical.

Quickly and painlessly.

I don't know. Like, you know, a stroke or heart attack in my sleep when I'm really old and I've had a great something like that.

I turned 40 recently. I was like, 'All right, this is kind of traditionally about the halfway point.' And I like felt okay about that. I assume I'll live much longer than 80 if things go well, but I was just like had like a very rich life. I'm like I wouldn't make change. No. But I do really want to stay healthy. Like I really value that.