I hope you have some advanced predictions about what capabilities the current paradigm would and would not successfully generate. Separately, it's very clear t

I hope you have some advanced predictions about what capabilities the current paradigm would and would not successfully generate. Separately, it's very clear that LLMs have "world models" in most useful senses of the term. Ex: https://www.lesswrong.com/posts/nmxzr2zsjNtjaHh7x/actually-othello-gpt-has-a-linear-emergent-world https://www.lesswrong.com/posts/nmxzr2zsjNtjaHh7x/actually-o... I don't give much credit to the claim that it's impossible for current approaches to get us to any specific type or level of capabilities. We're doing program search over a very wide space of programs; what that can result in is an empirical question about both the space of possible programs and the training procedure (including the data distribution). Unfortunately it's one where we don't have a good way of making advance predictions, rather than "try it and find out".