Microsoft CEO Calls for 'Emergency Brake' on Advanced AI
Microsoft CEO Satya Nadella has urged the tech industry to develop an 'urgent system to brake advanced AI' that humans can control. He proposed that AI models should be treated as potentially compromised and containable from the start, with authorized personnel able to pause or shut them down mid-task. Nadella suggested principles of observability, including model diversity, human-readable records, continuous testing, independent controls, and incident reporting.48 sourcesSee all sources