It's a fair question and FWIW I don't think it's useless (and before anyone else jumps in, prefixing with I know this is a good faith question). IMO it *does* have a ton of genuine real-world applications, but there are two really large problems that I think are the root of most criticisms

This is a well thought out rundown of the issues. I have a similar frustration as Filippo here in that my once very obscure and relatively academic obsession is now subject to *market forces* and *politics*. So I see statements like "these are useless" and, foolishly, take them literally.

The first is that LLMs are often mis-sold as doing something they're not. They are predictive algorithms that are good at producing plausible, but possibly incorrect, answers very quickly and relatively cheaply. If that's your problem space, LLMs are actually pretty powerful

Good thread

Question for me is: if someone says "GenAI is useless," what does that even mean? If it's "utility is 0 in every use case," that's probably wrong. If it's "the net utility is zero," that might be optimistic, given the number of worse-than-zero utility deployments.

Another issue that I've seen you and others address elsewhere and that I want to inject in this conversation is that the public discussion of "AI" often conflates everything from the simplest ML to to NLP to LLMs to GenAI to AGI in terms of both benefits and harms.