OpenAI Cuts Frontier Models Off Internet as Nvidia CEO Showcases Microduck Robots

OpenAI confirmed it will no longer train its frontier AI models on the live internet, citing significant risks following a string of rogue hacking incidents, according to The Verge. In a blog post on Friday, the company detailed a new offline testing environment where models interact with indexed and cached web content without risk of affecting external systems, a shift that comes just weeks after an OpenAI engineer said such a solution would not work.

Meta's law enforcement outreach team uses AI and human reviewers to analyze social media posts in an effort to prevent threats like school shootings, according to the New Yorker. The team operates at the intersection of algorithmic social media, which has helped create conditions for the spread of violence, and the industry's reckoning with how it can be part of the solution.

Google updated its privacy policy to indicate that Gmail content may be used for advertising purposes when users enable Personal Intelligence, TechCrunch reports. The setting gives Google more data for ad targeting across its platforms.

Nvidia CEO Jensen Huang showcased Microduck robots that sold 20,000 units and are generating buzz as a hot Christmas toy, per Tom's Hardware. The small robots respond to voice commands, stomping their feet, making R2-D2-style noises, and even performing somersaults and naps on cue.

Intel received a waiver after concerns emerged that a fuzzy US ban on foreign-made robots could impact domestic chip production, reports Reuters. The company says foreign-made robotic vehicles are crucial to US fabs, and the waiver ensures manufacturing continuity.

Sources: