I think there's a need to separate knowledge from learning algorithm. There's need to be a latent representation of knowledge that models attend to but the way

I think there's a need to separate knowledge from learning algorithm. There's need to be a latent representation of knowledge that models attend to but the way it's done right now (with my limited understanding) doesn't seem to be it. Transformers seems to only attend to previous text in the context but not to the whole knowledge they posses which is obvious limitation IMO. Human brain probably also doesn't attend to whole knowledge but loads something into context so maybe it's fixable without changing architecture. LLMs can work as data extraction already, so one can build some prolog DB and update it as it consumes data. Then translate any logic problems into prolog queries. I want to see this in practice. Similar with usage of logic engines and computation/programs. I also think that RL can come up with better training function for LLMs. In the programming domain for example one could ask LLM to think about all possible test for given code and evaluate them automatically. I was also thinking about using diffusER pattern where programming rules are kinda hardcoded (similar to add/replace/delete but instead algebra on functions/variables). Thats probably not AGI path but could be good for producing programs.