- cross-posted to:
- [email protected]
- [email protected]
- cross-posted to:
- [email protected]
- [email protected]
Reddit’s blockchain-based “Community Points” rewards crash after sunsetting::Tokens based on subreddit reputation saw dips over 85% after the announcement.
It’s what happens when the only use case for a tech product is its ability to interest venture capital.
We’re seeing the same thing now with “AI”
Eh, no, I’m going to disagree with you there.
Yes, AI is trendy, hyped, and a buzzword at the moment, but at the core of it there really is a very useful and practical technology (and it’s also much more varied than just LLMs). Take for the fact that generative AI is actually being used in practice, to do more or less what it was advertised to.
You’re right to say that generative AI has at least some practical applications. That’s a valid distinction as compared to crypto, and it’s exactly why the venture capital world pivoted to it so hard.
However if you compare how these companies are selling their technology to what it actually does, the gulf is astonishing. Almost every story of a company actually trying to use this tech has been a disaster. And what we hear from advocates is the constant refrain of “It’s early days”, but that’s exactly what we heard about crypto and distributed blockchain for over a decade. Meanwhile a lot of the experts are increasingly of the opinion that most of the proposed applications of AI will never reach the point where the tech actually works as advertised. Sooner or later, it’s gonna recommend a food bank as a tourist destination. And human validation isn’t going to solve this because inattentional blindness is a thing.
I do think that there are probably, eventually, real uses for this tech, but we’re clearly no where near them yet. Right now it’s still all just magic beans.
You need to be more specific than just say “AI” here. Experts do not say that AI will never reach a point of true general intelligence, but they may say that LLMs cannot do that.
AI is a big field, and it has seen massive improvements over time. Sure, don’t oversell its current capabilities, but we don’t really know where the current path leads. Current AI is already plenty impressive.
Yes, I’m aware. I thought that could have been easily inferred from context, given the topic of conversation. That’s, y’know, how conversation works. We’re not writing essays here.
I see a lot of people conflating concepts in AI, so I cannot be too sure just from context.