Anthropic - Wikipedia
… 103 Interpretability edit Anthropic carries out and publishes research on the interpretability of machine learning systems. 10 104 It has done research on "features" patterns of neural activation in a neural network that correspond to concepts . …