News

Image classification algorithms are notoriously error-prone, but a novel method for spotting errors within incomprehensible AI code could help solve the problem.