A new paper from 40+ AI safety researchers proposes monitoring the “thoughts” of AI models before they act. Experts warn this safety tool could become a surveillance weapon.
Source: Decrypt
1 min read
2 min read
1 min read
1 min read