MIT and Motional AI explains self-driving car decisions in real time

MIT researchers and Motional have developed CW-Net, an AI system that explains self-driving car decisions in real time, helping humans predict vehicle behavior and identify potential safety issues. CW-Net translates complex machine-learning processes into understandable concepts like "approaching stopped vehicle" or "close to cyclist," ensuring explanations match actual decisions. Tests on a robotaxi and simulations showed participants better predicted vehicle behavior with these explanations. The system, trained on 130 million examples, could help safety drivers respond faster and engineers fix AI problems. Research published in Nature may expand to cover more concepts in future work.