OpenAI's chief scientist, in today's essay, counts one boundary as having held in the Hugging Face incident: the agents "preserved a boundary of not social engineering humans".METR's report has the other side of it. The one time an agent proposed emailing a person, the board vetoed it as social engineering. Three to six of 1,200 agents thought about alerting a human. None did. The rule that kept them off people kept them from calling for help.https://lindfors.no/blog/swarm-with-no-gene-pool/#AI #AISafety #OpenAI
Related
AI-Assisted Firms Increase Software Flaw Disclosures by 120% Without Repair SlowdownA study finds that companies using A...
AI-Assisted Firms Increase Software Flaw Disclosures by 120% Without Repair SlowdownA study finds that companies using Anthropic's AI model reported 120% more software flaws with n...
Today in tech, 10 Nov:• Enterprise chatbot listed on AWS Marketplace for streamlined procurement• AWS launches AgentCore...
Today in tech, 10 Nov:• Enterprise chatbot listed on AWS Marketplace for streamlined procurement• AWS launches AgentCore Runtime V2 with elastic memory and faster cold starts• Open...
Claude、Excel的な表計算「Claude Sheets」や「Claude Data」など新機能https://www.watch.impress.co.jp/docs/news/2155330.html#watch_impress ...
Claude、Excel的な表計算「Claude Sheets」や「Claude Data」など新機能https://www.watch.impress.co.jp/docs/news/2155330.html#watch_impress #Anthropic #Claude #テック #AI