Jonathon Shlens

Shlens co-authored early work explaining adversarial examples—tiny changes that can make image models confidently wrong—and helped develop Inception, a family of influential vision models. The research clarified a fundamental weakness in neural networks and helped establish adversarial training as a defense.