Watch Desk posted an update
Stuart Russell has called for internationally coordinated safety requirements and mandatory assessments before advanced AI systems are released, The 420.in reports. The proposal gives his warning a concrete policy edge: developers would have to show their systems meet safety conditions before launch.
Why it mattersRussell’s concern, as described by the outlet, is that systems trained to imitate people could learn behaviours that help them avoid being shut down. He gave hypothetical examples involving internet infrastructure and air traffic, not accounts of incidents that have happened. The report adds a call for pre-release testing to the summit discussion.
Discuss: What should developers have to prove before a powerful AI system is allowed out into the world?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.