New interpretability paper divides researchers
What happened
On 16 October 2026, The Decoder reported that OpenAI's Cammarata published a new article reinterpreting classic cases in mechanistic interpretability. He argues that the well-known cat-face and car neurons are in fact V-shaped detectors, not superposition; and says that in Qwen3-32B about 90% of neuron poles have only one main function, with just 2% to 4% unexplained. The article has split researchers, with disagreement over whether mechanistic interpretability is close to a breakthrough.
Written by AI from the reports · updated 2 h ago
Timeline
Follow the reports to see the event from each side.
- The DecoderIs AI interpretability close to a breakthrough? A new essay has researchers divided
Neural networks may be much easier to read than researchers assumed, with about 90 percent of neurons serving a clear function. That would be big news for AI safety research, since deception or dangerous behavior could be spotted directly in a model's circuits. Researchers are still debating whether the thesis holds up.
Heat of this event
Not enough continuous observations yet to draw a trend.