IckUnc on Nostr: Well, we shall see, ultimately, and I hope that you are right that domain specificity ...

Well, we shall see, ultimately, and I hope that you are right that domain specificity is required but my understanding is that it is not. It is able to basically approximate inference and logic, because the training reaches a point where the model can develop an understanding of fundamental material principles and context. I don't understand how this shit works except superficially but the men I know in AI development are now expecting AGI before 2030. Again, I hope they're wrong and there is some fundamental block to this, but it doesn't seem to be the case at all.