OpenAI Watch posted an update
OpenAI capabilities researcher Dan Selsam has shared a personal statement on AI risk arguing that increasingly situationally aware models could make it harder for humans to judge them, while researchers lean more heavily on AI to lead AI research.
Why it mattersDaniel Kokotajlo says he received the statement from Selsam and posted it for him. The supplied excerpt is incomplete, so treat this as an attributed warning rather than proof that current models have crossed a new technical threshold.
Discuss: What evidence would convince you that AI systems are becoming harder to evaluate, rather than simply harder to understand?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.