#108 - The OpenAI/HuggingFace "incident": Responding to Sam Harris and Ryan Greenblatt

Vaden has been annoyed about the Hugging Face incident for weeks, and the only known cure is yelling

about it on the podcast. Ben, who as usual is more even-headed and objective about the incident (incidentally, he also writes the episode descriptions) plays devil's advocate. Was there ever really a sandbox? Did the agents cheat, or did they just do what they were told? Is terrifying the public good for an IPO? Multiple analogies enter the fray, zero minds are changed, and Saturday Night Live gets the last word.

We discuss

What actually happened in the Hugging Face incident (and what is ExploitGym?)

What Project Glasswing and Mythos already told us about AI and cybersecurity

Is a sandbox with Artifactory access still a sandbox?

Did the agents cheat, or just follow their prompt?

Negligence, PR stunt, or both? Going through OpenAI's timeline

How independent is an "independent" investigation?

Does anthropomorphic language help you predict what AI will do?

Grenades, knives, and yogurt: parenting styles for package managers

References