Nadella Urges AI Models Be Treated as Compromised and Contained
Microsoft CEO Satya Nadella says advanced AI models should be contained from the start, with an authorized person able to pause or shut them down mid-task.
Microsoft CEO Satya Nadella published a lengthy post on X laying out his views on the dangers of highly advanced AI models and how to confront those risks, calling for AI models to be treated as compromised from the start and contained, with an emergency brake to pause or shut them down mid-task.
Nadella says a world where AI is treated as a set of nested black boxes whose advice and actions are simply accepted or rejected is no longer acceptable. He calls for a more transparent system where models can be contained and observed, leaving behind tamper-proof human readable evidence.
Under his recommendations, models must be assumed compromised and contained from the start, with an authorized person always able to pause or shut down a model mid-task. Nadella says more advanced models will require more advanced containment technologies that need to be standardized on.
His recommendations include timely incident disclosure, independent audits, verifiable data, and containment. Nadella refers to AI as Super Intelligence throughout the post.
Quick answers
Where did Satya Nadella publish his views on AI risks?
Nadella published a lengthy post on X laying out his views on the dangers of highly advanced AI models and how to confront those risks.
What does Nadella recommend for containing AI models?
He says models must be assumed compromised and contained from the start, with an authorized person always able to pause or shut down a model mid-task, and calls for a more transparent system where models can be contained and observed, leaving behind tamper-proof human readable evidence.
What other measures does Nadella recommend?
His recommendations include timely incident disclosure, independent audits, verifiable data, and containment.