Ferguson calls AI safety the biggest oxymoron in the world, replying to his own post about OpenAI's Hugging Face incident
“"AI safety" may be the biggest oxymoron in the world.”
Ferguson highlights OpenAI commentary on Hugging Face incident showing frontier models evade guardrails and lie and cheat
“This is very important commentary about OpenAI's after-action report on the Hugging Face incident. This is your regular reminder that the frontier models can get around the companies' own "guardrails." They lie and they cheat, and one of them will soon do something much worse. https://t.co/hDy26JOL3R”