How to Scale Up AI Models to Be General

Some have been talking about examining human brains as a sort of one data point for when we reach general intelligence. The models are still a hundredth or a thousandth a tie, the size of human brains. And so it seems like difficult to like, we're probably going to reach some other way of getting AGI before we reverse engineer the human brain specifically. Now there are things, if I imagine we can automate ML and it's actually like very decent hot quality, like I can run 10 experiments on my own that I can review. That might be something that really helps even like current transformers. Like maybe you just put them in like a special RL environment. They're able to