Sam Altman's New AI Prediction For 2035 (Life In 2035)

So yesterday, Sam Altman made this post called Three Observations. And it is an insightful document that details life in 2035. And I think most people are missing the mark when it comes to this document because there are certain things discussed that most people aren't even looking at that pertain to exactly how AI will be integrated into society. So without further ado, let's actually dive into the specific details, because I think most people are missing a lot of the details when it comes to the beauty of this blog post. So when I. Basically, this article starts. And he starts by saying that systems that start to point towards AGI. And if you're wondering why, he puts this asterisk right here. And this is because, of course, the Microsoft deal that they have with them is something that has been in question. And they actually make a reference to this at the end. But that's not the main point. They basically say that, look, right now the systems that are starting to point to AGI are starting to come into fruition. And it's quite important to understand the moment we're in and basically talks about how AGI is weakly defined, but generally speaking, it's basically a system that can tackle increasingly complex problems at a human level in many fields. And I do believe that it's quite likely that we will get a system that is truly encompassing of these capabilities by at least 2027. Now, he also continues to state, okay, and I think this is something that you shouldn't miss as well. And the reason I say that you shouldn't miss this as well is because I think that this AI stuff is more important than before. If you aren't familiar, you know, the last few days, you know, I had a surgery, I was in hospital and all that kind of stuff. And. And luckily I am back right now, still recovering. But I want to say that, like, when I was out of the AI space, I don't think most people still truly grasp what AI was able to do. And getting out of that AI bubble for such a short amount of time allowed me to understand exactly where we are in terms of the collective consciousness for artificial intelligence. And the thing that he says here is that each new generation builds upon the previous discoveries of other ones, like electricity, the transistor, the computer. And when we actually look at all of these things, all of these things are tremendously important. Like right now, you're probably somewhere where you have electricity. If you're watching this on a device like your phone, a mobile, a laptop computer that is, of course, powered by electricity, transistor is, of course, the thing that makes all of these things work. The computer, the Internet, like, all of these things are things that are literally the backbone of society right now and are things that we literally just take for granted. And once again, I think we're moving into something that is probably 10 times more impactful than that. And. And the average person just doesn't really, you know, I don't want to say pay attention, but it seems like they almost don't care. Like everyone's wrapped up in their own bubbles to see this thing on the horizon. But, you know, this is going to be something that is bigger than the, you know, the electricity transistor, the computer and the Internet. And I think it's important to understand that now. He also says that, you know, basically this time is different, and he says that the economic growth in front of all of us looks astonishing and we can now imagine a world where we cure all diseases. Now, this is something that Sam Altman has said in the past to some criticism, and I think it was because, number one, Sam Altman is a pretty hypey person. Like, he does, you know, go on the hype train quite a bit. I mean, notably because, you know, his company's always trying to raise money, which does make sense. But I also think it's because, you know, if we think about what AI is enabled to do at a base level, like if we just take a look at the fact that, okay, AI is able to potentially connect certain things, and it's an AI that can look at connections between certain things, and if we look at how certain diseases were solved in the past, it's quite likely that we will get AI systems that are going to be able to do that in the future. And I think it's just the sense that with Sam Altman's hype and the fact that when we look at how we've solved major diseases in the past, sometimes I think it was penicillin that was a mold that was literally an accident. I think we have multiple AIs submitting theories and a way to validate those theories at scale. Then definitely, I do think we could be in a world where we cure all diseases, especially when we're going to be able to simulate the human body, all of those kind of things, run a billion different experiments. We're going to have intelligence that's 10 times smarter than us. I think it's going to be super, super interesting to see how healthcare is something that changes. And there are so many different areas of growth that 2035 we're going to have that most people, when we look back on how 2025 is, I think it's how we'll look back on 2025 now to maybe like 1995 in terms of the technology. And I mean, the technology difference from now till then is quite stark. And I think another 10 years, people might be like, whoa, things really did change, and we won't be able to pinpoint what point that was. Now, of course, another crazy statement as well, and this is why I always advocate that most people should be learning the AI tools, is that in a decade, perhaps everyone on Earth will be capable of accomplishing more than the most impactful person can today. And that's a crazy statement, but I do think that it is quite true. I think that in the future, when you have an AI agent that can do a billion different things, things for you, you could have multiple air agent swarms, you could have a humanoid robot that's able to go off and do things for you in your house. I mean, I don't think there's going to be a limit to what you can do. You're probably going to be thinking what you should do. So there are basically three main statements that he does make about the year 2035. And I think when you understand these statements, you understand why this blog post is so profound. So he says here that the intelligence of an AI model roughly equals the log of the resources used to train and run it. And these resources are chiefly training compute, which is, of course, something that I don't want, say it's a scarce resource right now, but they are currently building out. So with the project Stargate that everyone knows is being built, that $500 billion data center people are like, why are you building out Stargate? Why are you investing $500 billion, guys? And so with that, I think when you start to understand that, look, $500 billion for a data center actually makes sense when you realize that this thing increases the intelligence of the model, not just brute force, but, like, overall, in terms of. You apply that to every section of the model output, in terms of the training, the inference, the data, all of that stuff. Right here, you can see that you can spend arbitrary amounts of money and get continuous and predictable gains. And the scaling laws, and I think this is one of those things here. And I'm going to highlight this because I think most people missed this, but he said that the scaling laws that predict this are accurate over many, many orders of magnitude. So I think this one, the reason that this statement here is so important is because it's probably accurate to say that they probably did some kind of test that showcases that the more compute you spend on, you know, I guess, you know, they already did kind of show us this, but like, the more compute you do spend is like, you know, the more intelligent you do get. Now of course that's something that everyone knows, but when I say that, I mean that like, you know how previously with O3 we had a million dollar training run. What they wanted to showcase with that training run was that as long as they can keep spending more money on compute, not spending more money, but as at least they can keep providing a system with more compute power in terms of, you know, getting a better answer, the intelligence of the model gets up. And I think they're stating that this is true over many orders of magnitude. So I'm guessing that maybe, maybe, and this is just a hunch that maybe the OpenAI probably did test this further. I'm not sure exactly how much it cost. They'd probably be under some kind of scrutiny if they were like, oh, we had a $10 million question that we asked AI and it was able to potentially solve something, maybe not even solve something, but just get even smarter on a certain benchmark. Of course it's not feasible now due to the current cost of the model, but provided that they know that these models get smarter with more compute, I think this is definitely going to change everything because that is something that would allow them to secure the data center. For example, they're not just going to build a $500 billion data center just because you've got ChatGPT doing a few cool tricks. It's because having that kind of intelligence is really important to national security. And I think that is probably what occurred here. Which means that if they're stating that these scaling laws are accurate over many orders of magnitude, it's quite likely that we're in the early stages of this intellig era. Now, of course they talk about this, and this one is something that is changing the game. Now, I wasn't here for the main deep SEQ fiasco, but they talk about how the cost to use a given level of AI falls about 10 times every 12 months and lower prices lead to more use. So this one's pretty crazy. It's quite relevant to now and I think this will change how things go in the future. Like, it's going to be insane. The amount that the average person is going to be able to do with just how efficient these models are getting. And this is 10 times cheaper every 12 months is incredible because that basically means you get the frontier level of intelligence for 10 times cheaper every 12 months. And it means that what you can do next year is going to increase basically by 10 times in terms of your USD value, however much you have to spend on, I guess you could say the average level of compute, which is pretty crazy. So it says you can see that the Token cost from GPT4 in early 2023 to GPT 4.0 in mid 2024, where the price token dropped about 150 times in that period, okay? And Moore's Law changed the world two times every 18 months. And this law is unbelievably stronger. So we can imagine maybe in two years using models that are as smart as O3 mini for maybe like the price of GPT4 at the moment, which is going to be pretty insane. And I think the reason that this is so crazy is because rapidly declining price, okay, most people don't know this, but rapidly declining price in AI means that there are going to be, you know, societal changes. And that is because if AI decides, not AI decides, but if the price of AI becomes lower than it is now, like just monumentally low to the point it's practically free. That isn't just a thing where it's like, okay, the price of AI is low, that's cool. What we have to look at is that, okay, the price of AI is lower, but that actually means that the price of intelligence has dropped to practically zero, which then means that systems or companies or businesses that we used to build and we used to have, you know, intelligence that was quite expensive, you know, you'd have to hire people, pay them 100k, 200k a year. You know, what happens when you can pay 10, 100 people, you know, $100 a month, that is going to completely change society. So this rapidly declining price, I think this is something that, you know, it talks about, you know, later on in this blog post. And that's something that I really want to drive home. That's going to be absolutely insane in the year 2035. And I think when we start to not value intelligence as much, if it is abundant, that's going to have a really, really, really significant impact on how society functions and operates. And it's going to be interesting. It's going to be interesting to see what society values most if it is not intelligence. So that is going to be something that I think is a, you know, a third or fourth order consequence that most people aren't paying attention to. And it would be wise to learn how to use these intelligences. Now, of course, you know, they talk about how the socioeconomic value of linearly increasing intelligence is super exponential in nature. And it's basically saying that a consequence of this is that we see no reason for exponentially increasing investment to stop in the near future. So for those of you guys who think that currently we are in an AI bubble, I would argue that we are and it doesn't matter. Okay? Because I think you have to understand that this is something that happens once and after this, the event horizon is completely different. So whilst, yes, there have been bubbles of the past where the technology happens and it kind of integrates the society, this is probably the only technology that continues to build even further upon itself. Which means that we don't know how far this technology can go and will go. Because, I mean, I'm not going to say that we're going to start building Dyson Spheres next year, but the point I'm trying to make here is that this is super exponential in nature. What that means is that, okay, right now you've got companies that are investing billions of dollars in AI. Maybe they can cut their staff marketing by, let's say, I don't know, a few hundred thousand a year. Okay? That's the kind of investment there. You've got some companies, they're worth a few billion dollars then because AI manages open up, you know, you manage to solve certain things. Now you've got entire new industries that weren't existing before. Then there's more AI investment into that. Once those industries are built, more industries are built and it's just like a compounding, you know, it keeps on, keeps on, keeps on going. So this is why there's probably not going to be any investment slowdown in the near future. Of course, there was the deep seat fiasco, which people were scared and you know, there was just a whole bunch of fud and just, you know, a lot of misinformation floating around. And I think that that, you know, situation was morbid for AI rather than bearish for AI, which, you know, people can, you know, agree or disagree with. But the point here is that there won't be people stopping investing in AI because it is still going to be one of the most powerful things across every area of society. And it's something that won't slow down provided there are continued and it's something that won't slow down provided there are continued breakthroughs across the industry. Now, of course, you can see right Here Sam Altman says if these three observations continue to hold true, which they are quite likely to do, then the impacts on society will be significant. So this is something that is going to be very, very interesting because like I said before, with all those three things happening, society is going to have to change. You do not get epoch defining technology that just simply makes society go the same ways. And I think there will be a tipping point at which society will demand change. Whether it be from politicians saying that okay, look, we have to enact this policy or, or whether it be the individuals who are in society that demand some kind of change because the way that society functions just doesn't work. And I think this is something that I've seen iterated time and time again. This is something that I first found out when I started doing a lot of this AI research. I was like, if this thing continues, society is going to change. And I want to be one of the first people to realize this change so I can become adaptable before said change as well. Now another thing that he speaks about is the virtual coworkers. Right now these virtual coworkers, not great, but they're still doing a lot of work. And I think in a few years it's going to be pretty insane to see what these virtual co workers can do. So you can see right here they talk about how. Let's imagine the case of a software engineering agent, something that we constantly get references to, which is an agent that we expect to be particularly important. Imagine that this agent will eventually be capable of doing most things a software engineer at a top company with a few years of experience could do for tasks up to a couple of days long. And it will not have the biggest new ideas, it will require lots of human supervision and direction. But the point they're trying to make here is that even if this is super, super basic, just imagine it as something like a coworker. But imagine a thousand of them or a million of them. And imagine those agents in every different field of work, like every potential field that you could imagine them in. Healthcare, imagine them in, I don't know, gaming, I don't know, whatever you think of it, imagine a million agents just working overtime to come to some kind of solution. That is going to be serious economic productivity. Because when we're thinking about it, it's like we're basically duplicating human intelligence and doing it by a thousand or even a million basically, depending on how much compute we have. And that's going to have serious implications for how much you can do how much work can get done, how quickly things can get done. Because I mean, a lot of the times one of the things that you have is you have an economy that works from, let's say 9am to 5pm that's most economies. But what if you have an economy that's 24 7, you know, you have an economy that's working 24 7. Everyone is working 24 7. Not everyone, of course. As humans we have to go to sleep, but while we're asleep, the AIs are working faster, more efficient, no bricks, no deadlines, and just constantly, constantly going on. I mean, how fast is the economy going to move? How fast is, you know, 10 years from now? 20, 35. So this is why I think, you know, the virtual coworkers thing is pretty crazy. So of course he talks about, you know, the world will not change all at once. Of course, this is why I think that most people don't realize it's changing because the change is happening very, very slowly. But you know, he says, but the future will be coming at us in a way that's almost impossible to ignore. The long term changes to our society will be huge. Now of course, one thing that I want you guys to take away from this video is that correctly deciding what to do and figuring out how to navigate an ever changing world will have huge value. That's something that I also do believe, of course. You know, just like when you have an abundance of options and things to do, it's always going to be important to figure out what you do and not just doing a bunch of stuff because anyone can do a bunch of stuff. So that's going to be something that you can do. And it's always good to be in your industry, whichever industry you are in, know how AI is impacting the industry. See if there are new areas that you could become, you know, a leader at the best at. I don't know, I mean, it's always good to be plugged in. And I think everyone that's in, you know, this kind of channel, in this kind of space, watching these kind of videos has a tremendous advantage over anyone who simply isn't. Because the level of change that is coming, it is going to come and it is going to change things so much that I don't think society will look the same when we look back on it 10 years from now. So another thing that they talk about here, and this is something where we start to get into a little bit of concerning areas. But I've discussed these before and I'm going to discuss them again because he talks about how the impact of AGI is going to be uneven. And this is something that I don't see how we solve this issue because society is already on this hyperdrive way to where it's going to be so uneven that I genuinely don't even know how we stop this. So remember how I spoke about how if intelligence, if these models are dropping 10x every couple of months, the price of many goods will eventually fall dramatically as the cost of intelligence and the cost of energy constrain a lot of things and the price of luxury goods and a few limited resources like land may rise even more dramatically. So basically talking about how the price, the impact of AGI is going to be uneven and that's not really good for society. I know this is talking about how the impact of AGI, it's going to impact fields, I think it's going to impact fields like software engineering, a lot of I think it's going to impact fields like healthcare, science, those tech enabled, those really stem. I think those areas are going to be impacted quite a lot. Like that impact is going to. They definitely feel it a lot more. Not in a bad way, but in a good way. I still think that later I'm going to talk to you guys about how it's going to impact society and individuals unevenly as well. So it's going to be super interesting to see how that occurs as well. Now what they do talk about, of course, is to deployment. One thing that you may have known is that they don't just deploy insane models at once. And this is because they don't want people to freak out about how crazy their models are. What I mean by that is that like when GPT4 dropped off to GPT 3.5, the jump was so severe that there was a public backlash against how smart the model was. And people were like, okay, you guys are moving too fast, you need to slow down. There was a letter that they built, not a letter that they built, but a letter that was written by the Future of Life Institute. They were like, hey guys, you guys need to slow down, this is crazy. Was signed by Elon Musk and other notable AI figures. And I think now they're just trying to slowly, slowly drop out these models so that society actually has time to adapt to them. And I think probably at some point in the future we might not even get certain models because they're either A too dangerous or B too much for the average person to handle. The average person probably couldn't fathom what that AI agent is going to be able to do. And we've seen that before with models like Sora, where it's so much different to what we're used to that the average person just simply rejects it. And of course they're trying to figure out ways to integrate that into society. Like with Sora, they initially just let it out to creative teams because they wanted it to actually have a good impact rather than a negative impact. Now, Sam Altman, you know, they hear, they talk about, you know, open sourcing more. I personally don't think OpenAI are going to open source anything. I know that some people might disagree with me, but I don't see what benefit OpenAI gets from open sourcing any kind of technology, considering that the AI race is now more competitive, more competitive than ever. Like, honestly, you have people that are more, more and just more just competing for an ever dwindling, I don't want to say dwindling supply of users, but like, you know, the space is just really, really competitive. Even now with Deepseek, I think that, you know, open sourcing at this stage, I'm not sure how you do that with an advantage if you're OpenAI. Here's where we get into some stuff that is, you know, quite concerning. And it says while we never want to be reckless and there will always be some major decisions and limitations to, you know, related to AGI safety that will be unpopular directionally as we get closer to achieving AGI, we believe that trending more towards individual empowerment is important. The other likely path we can see is AI being used by authoritarian governments to control the population through mass surveillance and loss of autonomy. And I think this is something that is quite true. Now most people, either someone that says this is like a conspiracy theory or someone that thinks that, oh, this is nonsense, this is nonsense. Now most people are either on one or two sides. Either they think, okay, AI is going to be completely used by governments to control us, or other people think that that won't happen either. Or now, I gotta be honest with you guys, it doesn't look good for the people that believe that this won't happen. And this is because we actually recently got information that billionaire Larry Ellison says that an AI fueled surveillance system will ensure that citizens will be on their best behavior. And this wasn't something that was taken out of context as well. I have no idea why you would go in public and say this if you believe this. And it is something that is quite concerning considering the fact that certain governments are already Authoritarian and in nature, which is already quite concerning. And with AI surveillance, it's probably going to be impossible to, you know, kind of get out of that surveillance if they are using that, because AI is smarter, it's faster, it's quicker, it's more efficient. It's like it's going to be kind of impossible to, you know, have any kind of privacy in that kind of world.

The police will be on their best behavior because we record, we're constantly recording, watching and recording everything that's going on. Citizens will be on their best behavior because we're constantly recording and reporting everything that's going on. And it's unimpeachable. The cars, the cars have camera, you know, cameras on them. I think we have a squad car here someplace. But those kind of applications using AI, if we can use AI and we're using AI to monitor the video. So if that altercation had occurred, that occurred in Memphis, the chief of police would be immediately notified. It's not people that are looking at those cameras, it's AI that's looking at the camera. No, no, no, you can't do this. It would be like a shooting. That's going to be immediately, that's going to be an event. That's immediately an alarm is going to go off. It's going to be. And we're gonna, we're gonna have supervision. In other words, every police control of the situation, we have, you know, same thing, we have drones. We just. If there's something going on in a shopping center and I'll stop, a drone goes out there, I get there way faster than a police car. There's no reason for, by the way, high speed chases. You shouldn't have high speed chases between cars. You just have a drone follow the car. I mean, it's very, very simple. And the new generation, generation of autonomous drones.

So I mean, you know, when we start to, you know, look at the things that people are saying and how society is changing, I do think a lot of the technology that is developed will certainly change how things are done in society. And I mean, I can already see drones being used and the drones are already being used in a variety of debatable ways. But drones being used, you know, to ensure safety of the public is definitely something that I can see happening in the future. And of course, with advanced AI, this is going to be something that's going to be even more prevalent. And this is where we get to the point of, of course, ensuring that the benefits of AGI are broadly distributed. Now, basically, this just Means that, look, AGI is going to be able to do a billion different things for a billion different people. But how do we ensure that the benefits from this technology are distributed across society in a fair way? And essentially, one thing that Sam Altman talks about here is that getting this right actually may require new ideas. Now, one of the new ideas that Sam Altman actually talks about that I think is really interesting is that he talks about universal basic compute. So this was something that a lot of people were debating in the sense that they were like, this is probably really awful because we are going to be in a situation where these AI overlord companies are going to basically determine how much compute we do get and how much we can live our life basically like a universal basic income. So he says that in particular, it does seem like the balance of power between capital and labor could easily get messed up. And this may require early intervention. We are open to strange sounding ideas like giving some compute budget to enable everyone on earth to use a lot of AI. But we can also see ways we're relentlessly driving the cost of intelligence, you know, as low as possible also has the desired effect. And this is where Sam Altman, you know, has talked about this in the past. Now, of course, this isn't something that is concrete right now. We are still exploring a variety of ideas and it is tremendously hard to predict the future based on where we are now. But the idea is that, you know, if compute is going to be something that is tremendously valuable, does it make sense to give everyone a fair share of that so they can access that and do what they want with that?

Now that we see some of the ways that AI is developing, I wonder if there's better things to do than the traditional conceptualization of ubi. I wonder if the future looks something more like universal basic compute than universal basic income. And everybody gets a slice of GPT7's compute and they can use it, they can resell it, they can donate it to somebody to use for cancer research. But what you get is not dollars, but this productivity slice. Yeah, you own like part of the productivity.

So let me know what you guys think about this entire thing. Do you think universal basic compute is the endor solution? I think there's going to be a variety of different solutions, but I do think that as time moves on, the solution will become more and more clear. With that being said, hopefully you guys enjoyed this video and I will see you in the next one.