I think generative AI has definitely been the most important area of progress in the last year, but it’s by no means confined to that. I think we’re starting to see self-driving cars become a commercially viable technology, and Google’s DeepMind had a big protein folding breakthrough a couple of years ago that could drive a lot of progress in biotechnology.

But I think that one reason LLMs have so many people (including me) excited is that they seem like they might be a fairly general-purpose technology. There are lots of jobs out there—lawyers, programmers, accountants, management consultants, marketing, executive assistants, etc.—that are mainly about reading, analyzing, and producing documents. So if LLMs are on a path toward truly understanding documents in a general way, that seems like it could have big social and economic implications.