Comment by torginus

Haven't read the article, but multiple times during my life I have had to research how to do things in a domain, and have come up against academia (both literature and people), who seemed to have formed an opinion on how to approach certain problems, and what sort of properties a desirable 'correct' solution should have. When I have dug into the reasoning of why things have to be this way, the arguments often devolved into a matter of taste, or the strength of justifications have weakened considerably.

Nevertheless, if one were to turn their backs on existing research, not only would they draw the ire of academics (idc that much), but also would have to do a lot of stuff from scratch themselves, which usually is quite the undertaking.

These flamewars are not uncommon in CS, and usually involve forces and intellects much bigger than mine, and we often end up looking back at viable things left behind in computing history, just because the mainstream decided to go a different way.

If we are to outsource thinking to AI models, we will outsource these consensus driven 'rational' decisions to them, and then AI companies will have a pretty large say on what 'correct' opinions will look like.