OpenAI's chief scientist says the company is committed to ensuring its models' reasoning remains interpretable.