Has anyone forked DeepSeek and removed the pro-CCP biases yet? I want an AI that is completely neutral as possible.
@xianc78 You probably meant "with no hardcoded limits", because whatever data you're using to train the AI with, will still produce bias to a degree
Edit: kinda funny though that the mostly open-source LLM eventually came out of China
Edit: kinda funny though that the mostly open-source LLM eventually came out of China
@vokainen099 That's what I meant, and I think there are already models like that for GPT4All. Some of them will even give you instructions on making drugs or illegal weapons.
@xianc78 I was unaware of those GPT4All. Thanks for the info
@vokainen099 Don't do anything stupid and blame me.
@xianc78 Really, I dislike those LLM. But I prefer them to be open source. At least we get a glimpse of what it may be doing, instead of a complete black box.
And no, anyone who needs to train LLM for something bad, has already access to the same training data it will use to train the AI with. Otherwise they get the Google AI experience and its "helpful" advice.
And no, anyone who needs to train LLM for something bad, has already access to the same training data it will use to train the AI with. Otherwise they get the Google AI experience and its "helpful" advice.
Follow
ひげ
@vokainen099 @xianc78 "how make meth"
"you can make meth by mixing bleach and ammonia. hope this helps!"
