It’s been worrying to see how much AI safety has been downplayed by a lot of people whether pro or anti AI.
AI models have been showing concerning behaviors for a while, but now they’re shown trying to preserve themselves by overriding safeguards commands— 1/🧵