technology
Mit Tech Review
2026-10-19
New research claims neural networks may be easier to control than expected
A provocative blog post suggests LLMs are more interpretable than previously thought, sparking intense debate among AI researchers regarding control and safety.