Hi Joanna, thanks a lot for your comment. Actually I came to exactly the opposite conclusion!

LLMs today are incredibly good at both inference and creativity. If you ask them to reframe a problem based on a different framework, or to brainstorm on different frameworks that could be relevant for a problem, they will do a great job, better in fact than most humans.

LLMs are not based on if-this-then-that like traditional programming, but are better visualized as some kind of "compressed" version of the content internet that is able to reproduce human activity as seen through text. And although of course the machine does not have understanding, what is does is by no means mere regurgitation - no more than human activity is. After all, our eccentricity and creativity also come from having absorbed countless ideas and concepts from the world out there - and then looking at them through an unexpected framework, which LLMs can easily reproduce.

I wrote extensively about the topic here, and I will publish more soon. https://medium.com/@jonathan-kahan/generative-ai-and-the-future-of-problem-solving-71b6ffc9a686