Yoshua Bengio

Yoshua Bengio earned his Ph.D. in computer science from McGill University in 1991, focusing on neural networks amid early skepticism in the field. His childhood interest in science fiction fueled his persistence; as a graduate student, he collaborated with Yann LeCun on handwriting recognition systems, contributing to AT&T's automated check processing using neural networks and probabilistic models. Post-Ph.D., he completed postdoctoral fellowships at MIT (1991-1992) with Michael I. Jordan on probabilistic modeling and recurrent networks, and at Bell Labs (1992-1993). In 1993, Bengio joined Université de Montréal as an assistant professor, advancing to associate professor in 1997 and full professor in 2002, where he remains today. His 2000 paper 'A Neural Probabilistic Language Model' addressed the 'curse of dimensionality' in language processing, enabling advances like autocomplete and translation. During the 2000s and 2010s, his lab drove deep learning breakthroughs, including distributed representations, generative models like GANs, and soft-attention mechanisms that influenced transformers and modern NLP. Bengio founded Mila (Quebec AI Institute) to centralize AI research in Montreal, serving as its scientific director until 2025. He co-founded Element AI in 2016 (sold to ServiceNow in 2020, where he advised), and launched LawZero in 2025 to develop 'honest' AI guardrails like Scientist AI for harm prevention. He co-directs CIFAR's Learning in Machines & Brains program, served as IVADO's founding scientific director, and advises Recursion Pharmaceuticals and Valence Discovery.