Can AI get the right answer from noisy data?

Scientists are studying noise in data to know when they can and cannot trust AI to answer complex science questions. Anything in this approach?