If we had implemented everything I suggested in my 2020 arXiv paper 'The Next Decade in AI', we might have actually achieved AGI by now. Over the past three years, after a digression motivated by the false promise of simple

→ View original post on X — @garymarcus