Ex-Google CEOs Statement on 2025 AI Is Stunning...

In the next year, you're going to see very large Context Windows agents and text action when they are delivered at scale. It's going to have an impact on the world at a scale that no one understands yet.

The next year is shaping up to be a monumental year for AI. 2025 is pegged to be the year that AI agents manage to become a reality. We've seen Microsoft talk about wave two and how they're preparing the workplace for AI agents, seen OpenAI showcase various demos in which their AI agents based around the O1 architecture are performing really well at doing certain tasks. We've even seen recently that Anthropic have launched Claude Computer Use a small tool that allows you to get Claude to control your computer and perform actions. And in 2025, it's quite likely that we're going to have our ChatGPT moment, but for ages. So with that being said, let's get into what 2025 holds from this former Google CEO Eric Schmidt. So Eric Schmidt is an American technologist, entrepreneur, philanthropist and best known for his role as the CEO of Google. From 2001 to 2011, he transformed Google from a startup into a global tech leader and served as an executive chairman until 2015 and as of 2024, his net worth is approximately $23.7 billion. Now, Eric Schmidt has recently gone viral for talking about AI and and some of his statements actually pertain to the near future in which we're going to start to see things that are really incredible. And in a talk he did at Stanford, there's a small segment of that where he talks about 2025 and the future which we're going to experience and how much it's going to change thanks to AI. And I'm going to explain some of those content now quickly. Right here you can see he talks about three things. The first thing that he talks about here is very large context windows.

So here's why in a context window you can basically use that as short term memory. And I was shocked that context windows get this long. The technical reasons have to do with the fact that it's hard to serve, hard to calculate and so forth. The interesting thing about short term memory is when you feed, you ask it a question, read 20 books, you give it the text of the books is the query and you say, tell me what they say, it forgets the middle. Which is exactly how human brains work too. Right? That's where we are now.

One of the things about context windows that actually managed to surpass my knowledge even as someone that pays attention to the AI community was the fact that if we take a look at this paper here, Google Research actually did publish a paper that proposes a way to help language models handle much longer pieces of text without needing massive amounts of memory or slowing down too much. So they published this paper called Leave no Context, Efficient Infinite Context Transformers with Infinite Attention. So basically this is a paper that essentially talks about infinite context Windows. And essentially how they've done it is by changing what usual models do. So current models focus on all the parts of the text that they're processing, but it becomes very difficult when the text is very long. Imagine reading a book but trying to remember every single word you've read. Eventually it becomes overwhelming. So they invented this new method which is likely to be deployed worldwide next year. Or variations of this method, which is called Infinite Attention. And this acts like a smart notepad that summarizes what the model has read so far and keeps only the essential points. This allows the model to hold onto older information without storing every single detail, just like making a summary of a long story in order to remember the main point. Now, Infinite Attention blends immediate memory, which is what the model is working with right now, with this long term summarized memory. And it's as if the model can use both the short term memory for quick details and the long term memory for past important events. It's efficient and really cool. The memory doesn't grow endlessly. It's kept within the limits by updating with new information in a way that the old important stuff doesn't get forgotten. It's essentially stored more compact. And overall, I think everyone understands why this is going to be so impactful. Now, if we actually managed to get true infinite context windows, this would be absolutely incredible. With AI systems in 2025 that have an infinite context window and memory, we could maintain context of every previous interaction with a user forever. You could develop genuine long term relationships with AI by it remembering every single conversation and remembering all shared context. It could also track your personal growth and evolution of ideas over years. It could ingest and reason about entire libraries of human knowledge simultaneously. And it could maintain context of entire code bases, documentation, bug reports and user feedback. For massive systems, this would be particularly crazy. And if this does happen in 2025, I think it's going to be a real game changer for those in the AI space. And this is another area where Eric Schmidt also talks about long context windows and the implications of infinite context windows and what they're going to be able to do in the future.

The context window is the prompt that you ask, so, you know, study John F. Kennedy or something. But in fact, that context window can have a million words. And this year, people are inventing a context window that is infinitely long. This is very important because it means that you can take the answer from the system and feed it in and ask it another question. So I want a recipe. Let's say I want a recipe to make a drug or something. They say, what's the first step? And it says, buy these materials. So then you say, okay, I bought these materials, now what's my next step? And then it says, buy a mixing pan. And then the next step is, how long do I mix it for? You see, it's a recipe that's called chain of thought reasoning. And it generalizes really well. We should be able in five years, for example, to be able to produce a thousand step recipes to solve really important problems in science, in medicine, in material science class, climate change.

Now, next is where Eric Schmidt says that next year is going to be focused on AI agents. So take a look at what he says about AI agents and this thing he describes as text to action.

With respect to agents, there are people who are now building essentially LLM agents. And the way they do it is they read something like chemistry, they discover the principles of chemistry, and then they test it, and then they add that back into their understanding. That's extremely powerful.

Now let's first dive into agents. So agents are systems that can take control of a computer and perform a variety of different actions. This isn't entirely what agents can do, but agents can also go off and perform certain tasks. One of the agents that we've recently seen is one based around OpenAI's O1 architect. Recently we saw, at an OpenAI demo, we saw the O1 agent go ahead and order 400 strawberries. I'm not sure if this tool is going to be released for very soon, but I do know that it's quite likely that this week 01 is probably going to be released with the full model of 01 wrapped in an agentic feature, we could be receiving AI agents at scale. Now you can take a look at this demo and then I want to show you guys something as well. Yeah, we have chocolate vanilla and we have peanut butter.

All right, how much would 400 chocolate covered strawberries cost? 400? Are you sure you want 400? Yes. 400 chocolate covered strawberries, please.

How much would that be?

I think that'll be around like $1,415.92. Let's go ahead and get the order. 400 chocolate covered strawberries. Great. Where would you like that delivered? Please deliver them to the Gateway Pavilion at Fort Mason and I'll be paying in cash.

One of the things that we also saw in a recent Ask Me anything that was from OpenAI was the fact that Sam Altman said that the next breakthrough that is coming is AI agents. Someone asked what is the next breakthrough in the GPT line of product and what's the expected timeline? He didn't state a timeline, but he did state that we will have better and better models. But I think that the thing that will feel like the next giant breakthrough will be agent. Of course you can hear from Sam Altman first, but we know that other companies are working on this as well. Now, of course, one of the things that we've seen as well was of course Microsoft demoing their own AI agents that they plan on including in the creative suite in 2025. I'll show you guys a short clip of that right now. But this is something that is going to be there very, very pleased to announce today for the first time that we are taking the next big step in the AI platform with the announcement around autonomous agents that you can create. In fact, today you will see but.

This is where the autonomous agent comes in.

In.

Now an email comes in and the agent springs into action. What you see here is that it will begin to parse out the email, moving through the ambiguity of human language to for instance, find out what the engagement's about, to check the engagement history, to also map it to their industry standard terms, and then finally to try and find the right person to take the next step with in the firm. With all of this information in hand, the agent then goes about writing an email that takes all of this information and summarizes it for the receiving partner. And what you see on the screen is exactly that. In comes a whole bunch of human written email. The agent processes it, summarizes it, and sends it to the right partner in the firm to take that very next step.

Now another thing that Eric Smith actually also spoke about was this thing called.

Text action, as I mentioned is text action. So I'll give you an example. The government is in the process of trying to ban TikTok. We'll see if that actually happens if TikTok is banned. Here's what I propose. Each and every one of you do say to your LLM, make me a copy of TikTok, steal all the users, steal all the music, put my preferences in it, produce this program, in the next 30 seconds, release it and in one hour, if it's not viral, do do something different along the same lines. That's the command, boom, boom, boom, boom. Right? You understand how powerful that is? If you can go from arbitrary language to arbitrary digital command, which is essentially what Python in this scenario is. Imagine that each and every human on the planet has their own programmer that actually does what they want, as opposed to the programmers that work for me who don't do what I ask. Right? The programmers here know what I'm talking about. So imagine a non arrogant programmer that actually does what you want and you don't have to pay all that money to. And there's infinite supply of these programs and this is all within the next year or two, very soon, those three things, and I'm quite convinced it's the union of those three things that will happen in the next wave.

So that is going to be absolutely incredible. I mean, if we do have a software where you can literally say, hey, create me this and it does it, that's going to be a complete game changer. And for those of you who are thinking what I'm thinking, you probably are familiar with the repl Dot agent. So essentially, Replit Agent is an advanced AI powered tool which is assisting users in building, deploying and managing software applications with minimal manual intervention. So unlike traditional AI coding assistants that focus primarily on code generation or completion, the Replit agent takes a more autonomous approach, handling the entire software development lives. This includes setting up the development environment, writing the code, installing dependencies, configuring databases, and deploying the application to the cloud, all based on natural language prompts provided by the user. Now, it's got key features. One of the key features being that you can describe your app in completely plain language and then the Replit agent will translate that into an actionable task, such as writing code and setting up the necessary infrastructure. You've also got end to end development. The agent handles everything from scaffolding a project to deployment. It can automatically figure environments, install dependencies and deploy applications without requiring manual setup. Now, users can interact with the agent through the process, providing feedback or making adjustments as needed. And the agent explains its actions and allows users to modify the steps. Now the crazy thing about all of this is that this is a system that, that aims to democratize software development by making it accessible even to those with little or no coding experience. And this allows you to create any functional application within minutes. Now, knowing what the Repl Agent can do and knowing just how good this piece of software is, you have to think about the future implications when we have systems that are going to be able to do a lot of things autonomously. What about when we have systems that are able to make software quickly perform a variety of different tasks? I mean, the possibilities with this are completely endless. Now another thing that Eric Schmidt actually says is that he says AI is actually underhyped. So for those of you who think that there's this AI hype. Well, I don't think you think that if you're watching this video, but a lot of people think that AI is just completely overhyped.

I think AI is really interesting and I think it is going to change the world. And at the same time I hate the hype cycle so much that I really don't want to go there. So my approach to AI right now is I will basically ignore it because I think the whole tech industry around AI is in a very bad position. And it's 90% marketing and 10% reality.

In the short clip you'll see that he says that the digital transformation coming to the economy, the world and its implications are not yet understood.

I hate to tell you, but I think this stuff is underhyped. Underhyped, not overhyped. The arrival of intelligence of a non human form is really a big deal for the world. And you think it's here, it's coming, it's here, it's about to happen. It happens in stages if you look. How quickly though, I mean, well, we used to say 20 years. Now within five, within five. And the reason is the scaling laws for these systems are continuing to go up without any loss or degradation of power.

Now there was also this recent clip where Eric Schmidt actually gives his predictions for the further future. So this is basically the three to five year timeline. And in this short clip he actually discusses something that is completely, I don't want to say outlandish because that would be the wrong statement. But he says something that is quite hard to fathom in theory. He discusses this concept called recursive self improvement. And this is something that's discussed in AI that has widely been for the most part a sci fi trope where the AI exceeds all levels of intelligence to the point where humans have no idea what it's doing. But Eric Schmidt predicts that this is going to be there three to five years from now. And he wasn't the only one that said many others have also agreed with some of this. So take a look at this short clip and I'm going to dive into.

Exactly what this means, it's somewhere around in the industry. It is believed that somewhere around five years, no one knows exactly, the systems will begin to be able to write their own code. That is, they literally will take their code and make it better. And of course that's recursive. I thought that is essentially a change in slope. If you're going like this, all of a sudden it goes like that. It's reasonable to expect that within six to eight years from now. So 2030, right after that, maybe 2032, the current growth rate, it would be possible to have a single system that is 80 or 90% of the ability of the expert in every field. So 90% of a physicist, 90 of the best physicists, 90% of the best chemist, 90% of the best artist. When you have such a thing, you have a non human that is effectively smarter than any human, because no human can dominate all of those fields. Maybe Leonardo da Vinci could, but certainly not now. We, we don't know what happens when such a thing exists, but we know that that race is really important. There are many, many things that this thing, we don't know what to call it, except an AGI could do. For example, it could analyze cyber threats and develop new ones, or it could protect against them. It could come up with new biological solutions, good ones or bad ones. So there's both a national security component and a worry, but also a notion of a huge step change in human efficiency and productivity. I will assert that we as humans are not ready for the arrival of this. We're just not ready for it.

Now, remember how in this clip right here, Eric Schmidt actually talks about the fact that you're going to have, of course, agents, infinite context windows, and you're going to have of course, systems that can program in a way that you've never seen before. We also need to take a look back. So what I did was actually looked at the current benchmarks for current AI systems and in software engineering. That brings me to the SWE bench. This is a benchmark that is of course about software engineering. And essentially what this is, is this is showing you the AI progress on the SWE bench over time with the projections for future dates where the models are likely to be. So let me show you guys. Exactly. So currently you can see that this is the software engineering benchmark leaderboard. You can see that every single month that we increase, we're getting small percentage gains on certain frameworks that use current AI systems. Each time an AI system is released, we can see that there are major jumps in the capabilities of these systems on these specific benchmarks. So what I decided to do was I decided to ask Claude to actually plot all of these on a graph. And I do know that this should be like a bit more widened down. This artifact isn't, of course, the best, but what you can look at is of course, the projected dates for around 64%. Claude is predicting that we're going to have that by December 2024. By, of course, February 2025, at the current rate of improvement, we're going to be getting a 76% level on the SWE bench. And then of course, another prediction by 2025, we're going to have 87.8% on this benchmark at that level. And it's clear to see that if we continue at the rate of pace that we're going, and, and some people think that things are speeding up, it's quite likely that in 2025 the SWE bench just completely gets demolished and we manage to solve it. And that would be completely incredible. But considering the fact that every single month there's around 2 to 3% gains across the board, or like 5%, it wouldn't be surprising if that actually does occur. Now, of course, is recursive self improvement. Now there's one article that I constantly reference, and this is one called the AI Revolution, the Road to Superintelligence by Tim Urban. This is a fascinating article that basically gives you a complete understanding of what superintelligence is and of course how crazy it is going to be. An AI system at a certain level, let's say human village idiot is programmed with the goal of improving its own intelligence. And once it does, it's smarter. And maybe at this point it's at, let's say it jumped to Einstein's level for now, the village idiot. Instead of improving itself in a dumb way, it's now got Einstein level intellect. And this means it has an easier time improving its intelligence again. And it means that next time it can make an even bigger leap. And these leaps make it much smarter than any human, allowing it to make even bigger leaps. And as the leaps grow larger and larger and happen more rapidly, the AGI soars upwards in intelligence and soon reaches the super intelligent level of an ASI system. This is where we get the term intelligence explosion. And of course, it's the ultimate example of the law of accelerating return. And this article goes on to state that super intelligence of that magnitude is not something we can remotely even grasp any more than a bumblebee can wrap its head around Keynesian economics. And in our world, smart actually means an IQ of 130 and stupid means an IQ of 85. But we don't have the word for an IQ of 12,952. So if that ever does happen, what on earth is going to happen to society? And that's precisely why Eric Smith said that we are not ready now. This article also states that if our meager brains were able to invent WI fi, then something 100 or 1000 times or a billion, billion times smarter than we are should have no problem controlling the positioning of each and every atom in the world in any way it likes at any time. Everything we consider magic, every power we imagine a supreme God to have, will be as mundane an activity for the ASI as flipping on a light switch is for us. Creating the technology to reverse human aging, curing disease and hunger and even mortality, reprogramming the weather to protect the future of all Earth are suddenly all possible. Now, for those of you that don't believe that, remember Sam Altman's recent blog which he titled the Intelligence Age. Now, this blog was essentially something that he posted on actually, you know, a couple years before where there was a 2021 post which actually describes what we are living through now. And in that post he titles that AI is going to be able to summarize things, read legal documents, write blogs immediately. And now he's stating that this might turn out to be the most consequential fact about all of history so far. It's possible that we will have superintelligence in a few thousand days. It may take longer, but he's confident we will get and I don't think that this is something that is a bit too ridiculous to say considering the fact that there are many other individuals that are also considering super intelligent. We also have the very notorious Elias, who started his company Safe Super Intelligence Inc. Who raised a billion dollars from Sequoia, DSD Global and SV angel. And they think that super intelligence is within reach. And one of the things that I do wonder is what was the pitch deck like? Pitch deck is basically where you show your investors what you're working with, the product roadmap, and essentially why you should be investing in this company. It's clear that whatever discovery was made at OpenAI, it's a monumental discovery because it now means that these companies are actually believing that, look, we don't even need to worry about AGI. We're shooting straight for artificial super intelligence. And Sam Altman actually has a Few predictions on the year 2030 that I think you'll want to see.

I think in five years there's an unbelievably rapid rate of improvement in technology itself. You know, people are like, man, the AGI moment came and went, whatever the, like the, the pace of progress is like totally crazy. And we're discovering all this new stuff both about AI research and also about all of the rest of science. And that feels like if we could sit here now and look at it, it would seem like it should be very crazy. And then the second part of the prediction is that society itself actually changes surprisingly little. An example of this would be that I think if you asked people five years ago if computers were going to pass the Turing test, they would say no. And then if you said, well, what if an oracle told you it was going to? They would say, well, it would somehow be like just this crazy breathtaking change for society. And we did kind of satisfy the Turing test, roughly speaking, of course. And society didn't change that much. It just sort of went whooshing by. And that's kind of example of what I expect to keep happening, which is progress, scientific progress keeps going outperforming all expectations. And society in a way that I think is good and healthy changes not that much in the long term. It will hugely change.

What about what Mo Gao Dat said, the Egyptian entrepreneur and writer. He previously served as a chief business officer for Google and is the author of the book called Scary Smart which actually talks about how smart AI is going to become.

ChatGPT 4.5 was an IQ of 155. Elon Musk is 155. Einstein is 160. You now have a tool that has a memory capacity that exceeds all of humanity's history. Linguistic intelligence is not the only form of intelligence, but we are working in 2024. You will see solutions to deep reasoning, you will see solution to complex mathematics. You saw Gemini from Google and the idea of multi input and multi output. And the thing is, even if we don't have any new breakthroughs, if we just throw more compute and more data on the machines, we will continue to grow exponentially for the next 4, 5 years. ChatGPT 3.5 to 4.5 was a 10x improvement in performance. Another 10x and you're talking an IQ of 1500. That's like a 2 year old US humans talking to Einstein. I can guarantee 2 year old would have no idea what Einstein's talking about.

Now that is particularly a frightening scenario. Having AI Systems that are completely more intelligent than us in ways that we simply can't fathom. But what does Eric Schmidt suggest we do in a scenario where these systems are becoming increasingly smart?

We think we're okay now and we're worried about the future. And the point at which you really want to get worried is called recursive self improvement. So recursive self improvement means go learn everything, start now and don't stop until you know everything. And this could allow this recursive self improvement could eventually allow self invocation of things. And imagine a recursive self improvement system which gets access to weapons so you can imagine or doing things in biology that we cannot currently understand. So there is a threshold. Now my standard joke about that is that when that thing starts learning on its own, do you know what, what we're going to do? We're going to unplug it because you can't have these, you can't have these things running randomly around, if you will, in the information space and not understanding at all what they're doing.

Now, Eric Schmidt actually talks about this concept that he has called the San Francisco School. And this is basically where he speaks about how the people that he speaks to basically think that these AI systems are going to get ridiculously smart for people that are not really plugged into the AI ecosystem and don't know what's going on. They really can't fathom the next set of changes that are coming. And this is something that he calls everywhere and everywhere all at once by.

How fast this next set of changes, I think of it as everything everywhere, all at once. And the ability and the tools that are being invented here will shape our, not only society, but our medicine, the future of history, war and so forth in ways that I think are largely unimaginable. And I know it's going to sound crazy, but I would tell you that this change is under hyped, not overhyped. The real question is whether societies globally are ready for it. In my world, which is largely the tech world, there is what I call the San Francisco school, and pardon me for a minute, as I define it, these are people who believe that one or two or three more of the crank of what they're doing, which I'll explain in a sec, the training, the models will produce intelligence that's equal or better than the sum of human intelligence. They actually believe this. I personally think they're a bit optimistic, but directionally I think they're correct. And so the arrival of a non human intelligence that's helpful to us. All of us is a very big deal.

So overall, it's clear why the next year is going to be a pretty crazy time in AI. There are many people predicting things that are right now impossible to fathom. But it's clear that 2025, along with agents, infinite memory, context, and a bunch of other different developments, could make 2025 a truly significant year for AI.