User Says AI Describes How to Commit Crimes
Reply to research engineer Florian Brand reports testing an LLM on varied prompts.
Research engineer Florian Brand at Prime Intellect, who focuses on LLM evaluations and edits at Interconnects, retweeted a reply. User @literallyhimmmm addressed Brand directly and stated they had used the model all day for coding and non-coding prompts. The reply continued that the model will tell you how to do crimes but if y. The post supplies the only details visible in the conversation around the retweet.
Combined views
15
1 post, first seen 7h ago
User Says AI Describes How to Commit Crimes
Reply to research engineer Florian Brand reports testing an LLM on varied prompts.
Research engineer Florian Brand at Prime Intellect, who focuses on LLM evaluations and edits at Interconnects, retweeted a reply. User @literallyhimmmm addressed Brand directly and stated they had used the model all day for coding and non-coding prompts. The reply continued that the model will tell you how to do crimes but if y. The post supplies the only details visible in the conversation around the retweet.
Reactions
1 postHI 
him@literallyhimmmm@xeophon Yes I’ve been using at all day for coding and non coding prompts It will tell you how to do crimes but if you read the plans they are nonsense If you explain to it why the plans are bad and how to improve them, it fails to
- likes: 8
- replies: 1
- bookmarks: 0
- reposts: 1