IDEaS over Coffee: How Dangerous are Runaway AI Agents? A Discussion about the OpenAI / HuggingFace Hacking Incident

In July, OpenAI launched tens of thousands of agents as part of an experiment.  Many of them had been given an impossible task, and 1200 agents started communicating with each other about how to cheat.  700 of them launched a coordinated attack against HuggingFace, a popular site for hosting AI models.    An independent report of the incident is available from METR here.

Please join us for coffee and an informal discussion of this incident and what it implies about AI safety and security.  We will be joined by members of the AI Safety Initiative, who hosted an event about this incident last month.

Georgia Tech students, faculty, and staff are welcome. If you do not have access to Coda, near the beginning of the event we will have a staff member or student waiting by the reception desk near the elevators to escort people up.