The cost of running AI models has dropped roughly 10x per year for the past several years, which is faster than Moore's Law ever moved. If that rate continues, the things that are economically impossible to automate with AI today become possible very quickly.
I think we are at an early transition point where the world is going to change more in the next 10 or 15 years than it has in really any comparable period in human history. And I want people to really take that seriously, not just as a throwaway line, but really think about what that means — that the things that are going to be true about the world 15 years from now will be things that seem unimaginable or at least unbelievable to most people today.
The thing I try to remind myself is that if you think about what's happened in the last 100 years — where people live, the nature of work, what life expectancy is — all of the things that we would have said to be true in 1925 would be wrong. And I think we will see more change in the next 10 years than we saw in the previous 100.
The other thing I'd say about AI is it has this strange property where the more you use it, the more you realize how transformative it's going to be. Most technologies, the more you use them, the more you understand their limitations. AI is almost the opposite — the more deeply you engage with it, the more you realize the ceiling is higher than you thought.
If you're Jeff Bezos, what is what's he worth? $300 million, $300 billion, right? Even if he makes 5% a year, right? Which he's going to make way more than that. Even if he makes 5% a year, he's making $15 billion a year, right? And it's just going to grow, right? It's going to grow unbelievably quickly.
