Anthropic warns that AI could soon escape human control

  > The industry is much closer to self-improving AI than previously expected, they warn.

Uh, no. Maybe we’re getting close to AI-assisted AI research, but nothing is going to happen unless a human presses the buttons.

  > So tech companies should consider slowing or pausing frontier AI development to allow time for researchers to gain a better understanding of the potential societal harms self-improving AI could unleash.

Ah, so they’re stuck and they want to slow their competitors down while they think of something else.