The Direct Institutional Plan provides a promising path to getting necessary AI governance.

ControlAI have demonstrated its methods in the UK — building a coalition of over 25 UK politicians in support of binding regulation on powerful AI.

Crucially, it's collaborative: citizens and organizations can participate independently, and the res…

The METR paper discussed here is perhaps one of the most difficult pieces of evidence for relatively short AI timelines to argue against.

There's uncertainty, but I think we can somewhat confidently infer that powerful AI is not going to take decades to arrive.

If you disagree, you probably should have a theory that explains why this trend…