Q: Is dual-use alignment-complete problem?

Researchers on LessWrong are debating whether it's even possible to measure which AI research helps safety and which could accelerate AI capabilities. Some argue it's what's called an 'alignment-complete problem'—essentially something so hard it can't be answered until we've fully solved artificial intelligence safety itself. But one researcher pushes back, arguing that funding organizations and safety labs should be able to make reasonable judgment calls using the knowledge they have now, without waiting for AGI-level solutions.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton