look you can prompt an LLM to not hallucinate and it will still hallucinate but it will hallucinate less and I *still* maintain that part of the whole shitstorm around "this" is marketing allowing people to continue to believe LLMs have the accuracy of "computers" when they really don't & can't
I think a reasonable position on the OpenAI huggingface thing is: * executing untrusted code is a bad idea * sandboxes frequently aren't, even with the best intentions * novel combinations of untrusted code are being generated at scale and being executed * consequences probably bad
is ai disclosure policy the new land acknowledgement policy
OK I just got to get this off my chest AI -- IN PARTICULAR CHATBOTS -- are useful RIGHT NOW for non-spreadsheeters in that they are actually (not always, but more often than not) helping time-poor people navigate bureaucracy receive eg benefits they deserve. That's a real plus -- not a lie.
Hm. "Generative AI allows anyone who wants to to be promoted beyond their level of competence"
unfortunately following meta's horizon metaverse initiative the company was unable to see whether their agent-based layoff plans to become ai-native would have legs
why not talk to an LLM when you don't have the energy to make up a guy to be mad at
yes but what if the latest openai model actually persuaded the openai people to relax controls during development hmmmm (sorry)
Purdue Pharma and Sackler family-style settlement but for AI
I guess people have already considered the theory that Anthropic and OpenAI are throwing megawatts of power and tokens at financially engineering their success *and* also letting agents loose at doing so
