←back to thread

Uncensor any LLM with abliteration

(huggingface.co)
586 points mizzao | 6 comments | 13 Jun 24 03:42 UTC | HN request time: 1.065s | source | bottom
Show context
vasco ◴[13 Jun 24 06:49 UTC] No.40666684[source]▶
> "As an AI assistant, I cannot help you." While this safety feature is crucial for preventing misuse,

What is the safety added by this? What is unsafe about a computer giving you answers?

replies(11): >>40666709 #>>40666828 #>>40666835 #>>40666890 #>>40666984 #>>40666992 #>>40667025 #>>40667243 #>>40667633 #>>40669842 #>>40670809 #
1. rustcleaner ◴[13 Jun 24 07:43 UTC] No.40667025[source]▶
If I can ask the question, I can take the answer. It's not up to daddy $AI_SAFETY_CHIEF to decide what an infohazard is for me.
replies(3): >>40667474 #>>40667943 #>>40670670 #
2. stefs ◴[13 Jun 24 09:06 UTC] No.40667474[source]▶
they're not only there to protect you, but it's also to protect third parties from you. bad actors generating fake nudes of your ex and distributing them online; this used to be an expensive operation, either monetarily (hiring unscrupulous photoshoppers) or in time by doing it yourself.

the other example would be fake news for influencing people on social media. sure, you could write lies by hand. or you could specifically target lies to influence people depending on their personal profile automatically.

how about you use it to power bot that writes personalized death threats to thousands of people voting for a political opponent to keep them out of voting booths?

3. pjc50 ◴[13 Jun 24 10:34 UTC] No.40667943[source]▶
If the AI provides you with information on how to make explosives, then its owners have committed a criminal offence in the UK.
replies(1): >>40668182 #
4. averageRoyalty ◴[13 Jun 24 11:16 UTC] No.40668182[source]▶
Are all chemistry textbooks banned in the UK then?
replies(1): >>40668351 #
5. pjc50 ◴[13 Jun 24 11:36 UTC] No.40668351{3}[source]▶
Information about explosives is removed. The good old Anarchists Cookbook is illegal to posess. https://www.bbc.co.uk/news/uk-england-northamptonshire-58926...
6. digging ◴[13 Jun 24 15:11 UTC] No.40670670[source]▶
> If I can ask the question, I can take the answer.

I don't see how that follows at all. Are you asserting that it's not possible for a person (hell, let's even narrow it to "an adult") to ask a question and be harmed by the answer? I promise it is. Or are you asserting something about yourself personally? The product wasn't made for you personally.