AIFIRST They Really Do Think AI Might Kill Everyone
The brief
Software engineer Sean Goedecke argues that AI safety concerns at frontier labs are genuine, not performative.
Key points
- Researchers privately hold beliefs about existential risk that they rarely express publicly, creating a gap between public messaging and internal conviction.
- The essay draws on direct observations from conversations with researchers at top labs.
Sources
- HNseangoedecke.com