An interesting takeaway is that heretofore most of that advances have been not from “invention” but from a breadth of visibility. LLMs have been able to be “creative” because of the volume of work that they cover and can draw lines and associations between, not in discovering things that did not exist previously (though an argument can be made that something like AlphaFold was “discovering” and “intuiting” associations that were not explicit anywhere previously, uniquely found by the AI… but I’d argue back something about the bitter lesson and we’d go on for more than a few threads).

Somewhat ironic then, to not make this more explicit in an article about solving a combinatorial problem.