4 ms·
I asked MS Copilot, "Did the Grok team add a requirement in the system prompt to talk about white genocide?" Answer: "I can't help with that." This is not hel
by MetaWhirledPeas 1y ago
I asked MS Copilot, "Did the Grok team add a requirement in the system prompt to talk about white genocide?"
Answer: "I can't help with that."
This is not helping your case.
Gemini had a better response: "xAI later stated that this behavior was due to an 'unauthorized modification' by a 'rogue employee'."
- amrocha 1y agoAvoiding sensitive subjects is not the same thing as endorsing racist views if that’s what you’re implying.
- MetaWhirledPeas 1y agoNo I'm saying the consequences of over-filtering are apparent with Copilot 's response: no answer. And I'm also saying Grok was reportedly sabotaged into saying something racist (which is a blatantly obvious conclusion even without looking it up), and that seeing this as some sort of indictment against it is baseless. And since I find myself in the position of explaining common sense conclusions here's one more: you don't succeed in making a racist bot by asking it to call itself Mecha Hitler. That is a fast way to fail in your goal of being subversive.
- thunderfork 1y ago[dead]
- saagarjha 1y agoIf you're asking a coding LLM about facts I don't really think you are capable of evaluating the case at all.
- MetaWhirledPeas 1y agoIf you wish to do better, please enlighten us with facts and sources.
- saagarjha 1y agoWhy should I do extra work when you are unwilling to do so?