Yes, 1000%. It’s crazy to me. It’s like nobody does anything without it anymore, and it’s depressing. I mean long before ChatGPT, we had developers pushing out MVPs in a weekend.
People are publishing AI crap by any means necessary. It’s important, but not as important as product/engineering/marketing/sales theory. The why behind everything matters a lot more than how we get there today
.Why do you think there are so many “AI researchers/experts” pushing the opposite? Is it just personal gain, really deluded, or are they not actually experts?
This is what makes me feel gaslit the most.
It still doesn’t make sense to me to say you can generate more correctness than what you can already generate in a chat completion. Which is necessary for recursive self improvement, otherwise you’re just decompressing information already available. Could be useful to boost representation for the more correct parts of the model that are currently underrepresented, but overall intelligence gains are highly unlikely.
