Satya Nadella says we should assume all AI models are ‘compromised’

In a lengthy station connected X, Microsoft's CEO laid retired his views connected the dangers posed by highly precocious AI models and however to face those risks. Nadella says we tin nary longer judge a satellite wherever AI is treated arsenic a "set of nested achromatic boxes" whose proposal and actions we simply judge oregon reject. He calls for gathering a much transparent strategy wherever models tin beryllium contained, observed, and leaves down "tamper-proof quality readable evidence."

Many of his recommendations align with what we've heard from others successful the industry: timely incidental disclosure, autarkic audits, verifiable data, and containment. It's connected that past constituent that helium appears t …

Read the afloat communicative astatine The Verge.