One doubt that I have about LLMs is what would happen if I could retrain the model on each interaction. Like, from scratch, without costs and instantly, with th

One doubt that I have about LLMs is what would happen if I could retrain the model on each interaction. Like, from scratch, without costs and instantly, with the whole dataset + today's interactions. Would this feel closer to (a dumb) AGI? Or is that not how an LLM's internal memory representation works? Why?