The fact that it’s losing hype is proof that… it’s losing hype. (A tautology always proves true.)
I do think that generative models will remain relevant in the near and medium future, but the article is failing basic logic, and switching cause and effect. To be able to fit the earlier history of the public attitude into the “hype cycle” says little to nothing about their future potential.
LLMs in specific seem to be shorter-lived for me. Still useful, mind you, but I don’t think that usefulness will get too above what we’ve already seen. I think that neuro-symbolic approaches will eventually supersede them, at least when it comes to AGI research.
Archive link.
The fact that it’s losing hype is proof that… it’s losing hype. (A tautology always proves true.)
I do think that generative models will remain relevant in the near and medium future, but the article is failing basic logic, and switching cause and effect. To be able to fit the earlier history of the public attitude into the “hype cycle” says little to nothing about their future potential.
LLMs in specific seem to be shorter-lived for me. Still useful, mind you, but I don’t think that usefulness will get too above what we’ve already seen. I think that neuro-symbolic approaches will eventually supersede them, at least when it comes to AGI research.