I think it’s important to still give a critical eye towards the use of AI, but at this point I think it’s clear that not only is the use of AI going to stop (even once the bubble bursts), but also that the top-end models are just becoming more and more capable every month.

A couple years ago I was giving GPT-3 complex prompts and laughing at how bad and error-prone the output was, but last week I was using GPT-5 to give me information in a field I have little knowledge of, and it’s giving me perfect answers in seconds that takes me 20+ minutes to verify as correct, and that’s tens of times faster than actually learning the field myself. Even if I were to take a year to learn it all myself, I’d then need to not only retain all of that information, but also keep up-to-date on advancements in that field, which an AI will just do over time. This way I can concentrate on the fields of work I already know and follow, but can dabble in other fields without expensive retraining or bugging others in those fields with basic questions.