Hacker News Reader

Diffusion models were the first gen AI tech to plausibly be able to replace humans, no? I feel like image generation was ahead of text generation 2015-2022, and prior to 2015 the best "gen AI" was markov chains which I don't think is within the scope of what we're talking about.

View on HN

I remember being blown away by BERT and all its variations (BART etc) after the Transformers paper. For topic modeling and other semantic understanding of text, that was bee's knees. But certainly not replacing any humans.

Yeah, BERT was cool, but IMO things like StyleGAN were equally cool. I think DALL-E was the first time that there was practical, nonphilosophical, discussion about AI replacing human industries. Maybe there was discussion within the data labeling industry about text classification models replacing humans prior to that? But also, data labeling is one of the industries that probably got a lot bigger due to AI, so...

We had some really neat powerful classification around then.

Reply on news.ycombinator.com