Gary Marcus

Noah Smith (@noahopinion) has become deeply intellectually dishonest. If he actually followed the field he would a. understand the difference between pure LLMs and LLMs with neurosymbolic harnesses, code intepreters, verifiiers b. realize that a central point of my entire career, in work spanning 30 years, was advocating for neurosymbolic hybrids, (generally against great hostility from the field). c. realize that the great leap forward in math came from adding neurosymbolic components (harnesses and especially Lean) to neural networks. d. realize that that great leap forward in math actually *vindicated* my central claim. Instead he completely ignored all that to take a cheap shot, out of context. (Though I acknowledge I was off the timing of math.) —- *As Terence Tao has noted there are still lots of things that (some) world-class mathematicians can do, like invent new principles, and write clear proofs, that as far as I know AI has not done, though of the course the results are very impressive.