One of the big ideas that you speak about, and actually you're on stage at Abundance 360 speaking about this, is the explosion of AI agents that are coming. This concept of the AI Atlantis. I find that compelling. I find that enticing and somewhat scary if you don't understand the implications.
What does that all mean? Billions of AI agents in AI Atlantis. Yeah, I think AI Atlantis was coined by Nat Friedman on stage, former CEO of GitHub, when we had that panel together, Peter.
And we've discovered this new continent. And the best way to think about these AIs is they're like interns. They're like grads. We keep treating them like they should be professors or PhDs, but they're just in the early stages of coming out of high school.
These supercomputers are high schools, right? But there's 100 billion of them. And we can get them to do all these little tasks.
But anyone knows who's had grads and interns? We have to teach them before they can do anything. And once they learn the processes, that's when you have agents that can go out and do jobs for you, that can translate something in different languages so you can reach a bigger audience in seconds, or that can paint pictures for you or do SEO, you know, that can be your call center worker. And so the first step was to teach them their liberal arts degree.
Now we're specializing them. We're optimizing them. And the cost, I mean, I think OpenAI just showed that the cost of...
GPT-3 originally versus the latest version of GPT-4 has dropped a thousand times, literally a thousand times. It's almost too cheap now. So we have this new continent and all these workers are coming that can do jobs as reasonably as any graduate.
What does that mean for your personal life, your company, your country? Well, it's simple economics. There's a massive amount of supply of intelligence and capability and rule following coming because these don't sleep.
All they need to be fed is a few flops of, you know, computer and electricity. How would your life change if you had a really good group of graduates? And then they will only get smarter.
They'll only get better at following ingredients. And soon they'll have physical bodies as well. When you look at Optimus, when you look at Unitree and the other robotics companies coming.
How far are we away from having competence there? And how will we see that develop? I think that we're not far at all. We just had to get to a certain level of... performance on the base models of the base degrees.
So again, I think the best way to think about these are liberal arts graduates. They've had a diverse education because we've literally shown them the whole internet, right? Those are the trillions of words that go into the GPTs or the billions of images that go into the image models or hundreds of millions of songs.
And now we're specializing them. And these are the workflows that we see. Because like when I started as a programmer 20, God, three years ago, I was getting old. We didn't have GitHub and libraries and the way programming is now, which is like, you pull from pre-made building blocks and you put it together.
We wrote directly to the computer. We had these very low level building blocks that now people have made into houses and, you know, Lego blocks that come together. Similar is with AI. We had to get a certain level of performance of the base models.
And now people are chaining the base models together in repeated. processes. They're integrating it with lookups for databases.
They're learning from how people are experiencing things. If you compare ChatGPT now to ChatGPT a year ago, well, it's just two years old now, right? There's a massive difference because it's learned from how people have used it and they're actually paying people for specialist use cases.