3 ms·
It'd be nice if, generally, people could be more specific when they refer to dangerous systems. 'AI' is a notoriously ambiguous term; we've arrived at and norma
by placesalt 3y ago
It'd be nice if, generally, people could be more specific when they refer to dangerous systems. 'AI' is a notoriously ambiguous term; we've arrived at and normalized countless 'AI' systems so far (chess-playing software, as a trivial example).
Being more specific will allow for more specific responses. If the danger being addressed is the danger from text-prediction software, call it that. If it's something else, then describe it.
- kleer001 3y agoLoss of jobs. It's loss of jobs.
- wsgeorge 3y agoA lot of the recent comments about AI safety are about generative AI models, which looks like a specific enough classification for this subject.
- bbor 3y agoI think this proves the point of who you're responding to: I'd say that the concern is about agential models that employ LLM elements, not LLM models on their own.
- wsgeorge 3y ago> the concern is about agential models that employ LLM elements, not LLM models on their own. Rather? I'd imagine concerns about autonomous agents running amok is further into the future than the digital content manipulation potential of gen AI possible today. Heck, porn deepfakes of celebs were a problem in 2020, and right now Adobe has made gen AI generally available in Photoshop (AFAIK)
- greenie_beans 3y agosame with the word "woke"