Goodfire's Eric Ho on Why Models May 'Think in Shapes' and How Interpretability Could Cut Hallucinations

In a video interview, interpretability startup Goodfire's Eric Ho argues better model interpretability could reduce hallucinations.