Clarity check: this model has not been trained yet at this time, correct? I understood you to be saying the model had already been trained initially, due to the word is in “is a foundation model”, as well as “can improve” (“could improve”?), etc. Just wanted to double check and point out the ambiguity.
Seems like a potentially interesting idea though, an interesting sketch of things to try to combine. Doesn’t seem like this would help scale to superintelligence, but it could be interesting at least at a more useful near human capability approach. It also seems like it would be quite a research project.
Clarity check: this model has not been trained yet at this time, correct? I understood you to be saying the model had already been trained initially, due to the word is in “is a foundation model”, as well as “can improve” (“could improve”?), etc. Just wanted to double check and point out the ambiguity.
Seems like a potentially interesting idea though, an interesting sketch of things to try to combine. Doesn’t seem like this would help scale to superintelligence, but it could be interesting at least at a more useful near human capability approach. It also seems like it would be quite a research project.
Yes, I’ve changed the title of the post and added a footnote on “is a foundation model”.