@aurazonefarm: As the race toward self-improving AI speeds up, what should the leading labs actually be doing to keep it safe? In conversation with Reid Hoffman at Masters of Scale Summit, Google DeepMind founder Demis Hassabis lays out the practical steps. Labs should give trusted cyber defenders early access to models fine-tuned for defense so they stay ahead of attackers. They should build more hardened sandboxes. And they need monitoring strong enough that no AI agent can run for weeks without anyone noticing. He argues that many recent cyber issues could have been prevented with better engineering. If the labs don’t improve quickly, he warns, an accidental incident, the AI version of an oil spill, is bound to happen. #MastersOfScale #ArtificialIntelligence #AITrends #TechInnovation #FutureOfTech #GenerativeAI #ReidHoffman #TechLeaders #BusinessStrategy @mastersofscale

3673

There are no more comments for this video.

To see more videos from user @aurazonefarm, please go to the Tikwm homepage.