Interesting new post from OpenAI about safety. About humans being in control:

Our approach to alignment centers humans. We aim to develop mechanisms that empower human stakeholders to express their intent clearly and supervise⁠ AI systems effectively - even in complex situations, and as AI capabilities scale beyond human capabilities.

This is probably my biggest concern, AI agents running without human supervision and executing tasks that are beyond what we even know how to do. There are many positive benefits to AI, but there are also some things we shouldn’t attempt.

CrazyDogLadysezBreatheWithMe

@manton It’s all an environmental crime and should be illegal until we’ve worked out the water, energy, and other resource issues.

We’re about to have a world war over rare earth metals ffs.

They’re manufacturing superconductors in the Arizona desert, essentially f*cking every single child born in the American west during the next 30 years. Huge swaths of the region will be uninhabitable beyond then.

Bye, Phoenix! Maybe you’ll last longer than Salt Lake City, maybe not. FAFO

2025-03-05 3:34 pm

Manton Reece @manton
Lightbox Image