3 ms·
NYT assumes that LLMs = AI, which is far from truth. This is just a recent hype which relies on getting insane amounts of data to train, but we had and will ha
by elAhmo 2y ago
NYT assumes that LLMs = AI, which is far from truth.
This is just a recent hype which relies on getting insane amounts of data to train, but we had and will have AI models that do not rely on training using data without consent.
- kredd 2y agoTechnicalities make things harder to understand for general masses. “The Data That Powers LLMs” means little to nothing to an average user. My 70 year old parents use ChatGPT, and they have no idea what LLM is, but call it AI.
- wkat4242 2y agoIt's not just LLMs but all generative models that rely on extreme amounts of training data. Text but also images, video, speech, music. And it makes sense that LLMs were the ones to trigger the hype. AI in general has been making steady progress but most usecases are really hard to explain to layman. Give them a chatbot and they naturally understand what's happening. The only thing I'm seeing is that people are using them wrong. They're using them like all-seeing oracles, which is exacerbated by the confidence with which LLMs provide their answers, and the innate human idea that "it's a computer so it must be right". But knowledge isn't really where LLMs shine, at least not without search engine integrations. It's rather generation, summarisation, translation (with context sensitivity), rewriting styles etc.