So yes, you’ll probably want to take a model trained on math problems to forgoe the step with code. But then, note that writing correct programs is indeed a hi

So yes, you’ll probably want to take a model trained on math problems to forgoe the step with code. But then, note that writing correct programs is indeed a high level solution to a full text math problem, is it not? Going from there to solving it directly should be a matter of some tuning. Finally, who said understanding math? The whole debate is about getting shockingly useful results precisely without symbolic reasoning. If the main contention is that got does not do symbolic reasoning, then we are back at Gary Marcus… yes we know this. It’s not why researchers are so amazed by these models. Its that they output steps (or in this case code since its Codex based) solving university level math with a simple transformer architecture