OpenAI is pursuing a new way to fight A.I. 'hallucinations' - Their technique called 'process supervision' rewards based on each individual correct step in a multi-step reasoned problem rather than only the final outcome. Interesting approach but sounds a bit like mind control. Who decides on the steps in the first place? Isn't the whole point of AI that it would be able to generate steps that humans would never be able to envisage? www.cnbc.com #