More Compute Power Doesn’t Produce AGI
nav.al/agi
What Jamie wrote
This is a super important point that many people miss with the radical advances in machine learning.
It’s parroting. It’s brilliant Bayesian reasoning. It’s extrapolating from what it already sees out there generated by humans on the web, but it doesn’t have an underlying model of reality that can explain the seen in terms of the unseen. And I think that’s critical.
To be clear, that doesn't mean the value of these technologies is lessened, but is instead to say that techniques for solving "closed system" problems are very powerful. But "open system" problems are still far from our reach. The challenge is now in figuring out what is a closed and open system. Any game with a clear rule set is closed. Chess, Go, etc. Is driving a car a closed system? My intuition would say no, because it happens in the real world with random things. But maybe not? Driving a car could be managed as a closed system.
So, think harder about how an open system could be modeled as a closed system.