Pretraining + RL itself is the scaling limit. If you feed it the entire dataset before 1905, LLMs aren't going to come up with general relativity. It has no concept of physics, or time even.
AGI happens when you DON'T need to scale pertaining + RL.
AGI happens when you DON'T need to scale pertaining + RL.
Link?
There is a mountain of data pre-1905. Certainly enough to train a decent 30B parameter model.
Now, digitizing & OCRing all of that data... THAT is a challenge.