AGI is a science problem. We've lucked into these amazing abilities by just scaling. But we don't really understand how they work. And they are obviously mis

AGI is a science problem. We've lucked into these amazing abilities by just scaling. But we don't really understand how they work. And they are obviously missing a piece, some self-reflection, or continuous-loop operation perhaps, which we again don't understand. Perhaps we'll do all this engineering and luck the solution again, but I think probably not.