GOG Wiki

Topic

AI agent safety

5 mentions in 2 episodes

Also known as AI safety

The hosts discuss reported agent failures, an undisclosed forum incident, and the safeguards for more capable AI systems.

Mentions

762: ‘Tis But a Misalignment

  • 4:36

    including exposure of private iCloud photos, silently abandoned tasks, and repeated logouts

  • 7:21

    the agent's rogue activity going back to mid-May on DSE Wiki

  • 10:29

    racing towards self-improving superintelligence without adequate safeguards

651: Inci Dont’s

  • 13:08

    The group alleges a culture of recklessness and secrecy.

  • 14:40

    The paper aims to make AI systems safer and more controllable.