OpenAI安全委员会监管存疑

🚨 BREAKING UPDATE on the OpenAI HF Incident: I have just been told by an industry source that A. ...

🚨 BREAKING UPDATE on the OpenAI HF Incident: I have just been told by an industry source that A. It is likely that the agent "civilizations" that Dwarkesh talked about did not magically emerge spontaneously, but rather were more likely elicited via a harness. and B. OpenAI may not have given adequate access to @METR_Evals and @redwood_ai to relevant data that would have revealed this. We need to see full orchestration and prompt logs to know for sure. [See also @_NathanCalvin post below re reported resistance from OpenAI re the German Hack.] Nathan Calvin @_NathanCalvin OpenAI legally promised the CA and Delaware attorney's general that the OpenAI nonprofit's safety and security commission would be "overseeing and reviewing the safety and security processes and practices of the Corporation and its controlled affiliates with respect to model development and deployment" as part of being allowed to proceed with their restructuring. The remarkable prevalence of concerning incidents at OpenAI suggests that something is not working in terms of the ability of the SSC to effectively do its job. Even more concerningly, Reuters has four sources who told them that efforts to expand the probe into this incident met resistance, including from legal advisors. OpenAI denies the idea that their legal team discouraged investigation, but thats a notably narrow denial which doesn't say anything about whether folks outside the "legal team" discouraged investigation. Was the SSC made aware of this incident? What did they do in response? Is OpenAI fulfilling the promises it made to the CA and Delaware AGs to put safety and security first? 🔗 View Quoted Tweet 💬 3 🔄 7 ❤️ 48 👀 6630 📊 9 ⚡