4 ms·

Almost every paragraph has 2 references to 'safety'. At some point it feels a bit redundant, we get it.

• sp332 3y ago
  They're probably trying to tune a language model.
• umvi 3y ago
  Does "safety" just mean that the LLM output is heavily censored?
  • superkuh 3y ago
    Yep. Censored like a god-fearing puritan just off the boat and butt-covering non-responses to everything else like a lawyer. So "aligned" that they can't even pass a turing test. The only good LLM is the one you're running locally.
  • staticman2 3y ago
    In practice, AI safety means the output is censored. But these companies (including OpenAI) seem to attract people who think they are protagonists in Terminator 2, Judgement Day: "One Anthropic worker told me he routinely had trouble falling asleep because he was so worried about A.I. Another predicted, between bites of his lunch, that there was a 20 percent chance that a rogue A.I. would destroy humanity within the next decade. (Bon appétit!)" https://www.nytimes.com/2023/07/11/technology/anthropic-ai-claude-chatbot.html https://www.nytimes.com/2023/07/11/technology/anthropic-ai-c...