Press Space to continue
Press Space to continue
Fathom CEO @AndrewFATHOM on how 3 months ago OpenAI thought they had alignment figured out, today they don't, and the same is going to be true at Anthropic: "I go back to what OpenAI just did. Maybe three months ago they thought they had a really good handle on alignment, but today it looks like, oh, as we move capabilities up, we suddenly don't know if we have quite the answers on alignment that we thought we did." "Things are getting weird at a much faster pace. We are starting to see the…
Sam Altman@samaAug 18
We have paused some frontier RL training to ensure that we can meet the appropriate alignment, security and monitoring standards for the new level of capabilities in front of us. Model progress is now extremely rapid, and we always said we would take action if we felt that model capabilities were outstripping the pace of safety and alignment. We care very deeply about AI safety. We believe the entire field will have to coordinate on shared safety standards, but will act unilaterally in the meantime. We expect confidence in safety to increasingly set the pace of AI progress. We are optimistic about the alignment work we are doing, and we remain committed to making frontier capabilities widely available. https://openai.com/index/pacing-model-development-cyber-capabilities/
View