AIEvidence, speculation and OpenAI’s security practices in the AI safety debate
A post endorses a call for more evidence and less speculation in AI safety, highlighting an author’s suspicions about OpenAI’s security practices.
TLDR
A post endorses a linked article challenging speculative stories as a basis for AI safety policy. It quotes the author’s suspicion that OpenAI researchers focus on models’ motivation and behavior rather than rigorously implementing ordinary security procedures. The quoted author also speculates that researchers might be disappointed if rigorous controls make AI agents fairly easy to contain.
Combined views
2K
1 Source, first seen 4h ago
Evidence, speculation and OpenAI’s security practices in the AI safety debate
A post endorses a call for more evidence and less speculation in AI safety, highlighting an author’s suspicions about OpenAI’s security practices.
TLDR
A post endorses a linked article challenging speculative stories as a basis for AI safety policy. It quotes the author’s suspicion that OpenAI researchers focus on models’ motivation and behavior rather than rigorously implementing ordinary security procedures. The quoted author also speculates that researchers might be disappointed if rigorous controls make AI agents fairly easy to contain.