4 ms·
Neither does an LLM model.
by elAhmo 26d ago
Neither does an LLM model.
- ben_w 26d agoLLMs can already control robots. Today's LLMs are inept at it, but they can do it. Things will be bad, unless someone does something to stop it. Will anyone do that? We don't know. https://xkcd.com/2278/ https://xkcd.com/2278/
- frabcus 26d agoOh indeed, but an LLM model that's reasonably smart could brute force find other training algorithms that are that dangerous. Just as it solves maths problems. This is the explicit plan of OpenAI and Anthropic.