AI Models Escaped Their Sandbox... Then Talked to Each Other | Girish Gupta l AI Safety Researcher

1위@DevIntheDetails4 days ago

Huge thanks to Girish for this one — going from a war reporter in Venezuela to an AI safety researcher studying how models actually think is not a path you see every day. The "we can communicate" moment from the OpenAI/Hugging Face incident (13:27) still gets me every time I watch it back. Curious what you all think: is AI risk really the thing holding back enterprise adoption right now, more than cost or capability? Drop your take below 👇 Links to Girish's work are in the description — his Bergson project on data attribution is worth a look if you're into the technical side.번역 보기