“OpenAI recently hired Alexander Madry from MIT to lead research on model interpretability and the science of predicting capabilities.”