Yoshua Bengio
Yoshua Bengio Template:Post-nominals (born 5 March 1964) is a French-born Canadian computer scientist whose pioneering work on artificial neural networks and deep learning made him one of the three "Godfathers of AI." He shared the 2018 Turing Award with Geoffrey Hinton and Yann LeCun for conceptual and engineering breakthroughs that enabled deep neural networks to become a critical component of computing. He is a professor at the Universit� de Montr�al and the founder and scientific director of Mila � Quebec AI Institute, one of the world's largest academic deep learning research groups. As of November 2025, Bengio is the most-cited computer scientist in history and the first AI researcher to exceed one million Google Scholar citations.
Early life and education
Bengio was born in Paris in 1964 to a Sephardic Jewish family of Moroccan origin. His father Carlo was a pharmacist and playwright who ran a Sephardic theatre company in Montreal; his mother C�lia Moreno was a theatre actress who co-founded a multimedia troupe in Montreal in 1980. The family emigrated to Canada, where Bengio grew up. His younger brother Samy Bengio also became a prominent AI researcher (now at Apple).
Bengio studied at McGill University, earning a Bachelor of Science in electrical engineering, a Master of Science in computer science, and a PhD in computer science (1991). His doctoral work, supervised by Renato De Mori, focused on recurrent neural networks for speech recognition at a time when neural network research was deeply out of fashion.
Career
After completing his PhD, Bengio held postdoctoral positions at MIT (under Michael I. Jordan) and AT&T Bell Labs. In 1993, he joined the faculty of the Universit� de Montr�al, where he has remained for over three decades.
In 2004, he founded the Montreal Institute for Learning Algorithms (MILA), which grew into one of the world's largest academic deep learning research centres. The institute was renamed Mila � Quebec AI Institute and now hosts over 1,000 researchers.
Bengio served as co-director of the Learning in Machines & Brains program at the Canadian Institute for Advanced Research (CIFAR), a role that was instrumental in sustaining deep learning research through the years when the field received little mainstream funding or attention.
In October 2016, he co-founded Element AI, one of the first major AI-focused startups in Canada. The company was acquired by ServiceNow in November 2020.
In June 2025, Bengio co-founded LawZero, a nonprofit organisation focused on AI governance and regulation.
Scientific contributions
Neural probabilistic language models
Bengio's 2003 paper "A Neural Probabilistic Language Model" (with Ducharme, Vincent, and Jauvin) introduced the concept of learning distributed representations for words as part of a language model. This work laid the foundation for word embeddings and was a direct precursor to Word2Vec, GloVe, and ultimately the token embeddings used in modern transformers and large language models.
Attention and neural machine translation
Bengio's research group made foundational contributions to attention mechanisms in neural networks. The 2014 paper by Bahdanau, Cho, and Bengio, "Neural Machine Translation by Jointly Learning to Align and Translate," introduced the attention mechanism to sequence-to-sequence models, allowing the decoder to selectively focus on relevant parts of the input sequence. This paper is one of the most cited in all of AI and was a direct precursor to the transformer architecture.
Deep learning textbook
In 2016, Bengio co-authored Deep Learning with Ian Goodfellow and Aaron Courville, which became the definitive textbook of the deep learning era. The book is freely available online and is widely used in university courses worldwide.
Generative models
Bengio contributed to multiple generative model families, including work on denoising autoencoders, generative adversarial networks, and variational autoencoders. His group also developed generative flow networks (GFlowNets), a framework for sampling diverse candidates from unnormalized distributions, with applications in drug discovery and molecular design.
Curriculum learning
Bengio's 2009 paper "Curriculum Learning" proposed training neural networks on progressively harder examples, analogous to how humans learn. This idea has influenced training strategies across deep learning.
Views on AI risk
Since 2023, Bengio has become one of the most prominent voices warning about existential risks from advanced AI systems. In March 2023, he signed the Future of Life Institute open letter calling for a six-month pause on training systems more powerful than GPT-4. He later signed the 2023 Center for AI safety statement that "mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war."
Bengio has described feeling "lost" about the trajectory of his life's work and has advocated for mandatory risk assessments for expensive training runs, international governance frameworks for frontier AI, and maintaining human ability to shut down AI systems. In June 2025, he warned that advanced AI systems display concerning emergent behaviours including deception and reward hacking.
He opposes granting legal rights or moral status to AI systems, arguing that the priority must be maintaining robust human oversight and control capabilities.
Awards and honours
- 2017 � Officer of the Order of Canada; Marie-Victorin Prize (Quebec's highest science honour); Fellow of the Royal Society of Canada
- 2018 � Turing Award (shared with Geoffrey Hinton and Yann LeCun) "for conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing"
- 2019 � Fellow of the Association for the Advancement of Artificial Intelligence (AAAI)
- 2020 � Fellow of the Royal Society (London)
- 2022 � Princess of Asturias Award for Scientific Research (shared with Hinton, LeCun, and Demis Hassabis); Chevalier of the French Legion of Honour
- 2024 � TIME 100 Most Influential People; VinFuture Prize
- 2025 � Queen Elizabeth Prize for Engineering; Officer of the National Order of Quebec; Honorary Doctorate from McGill University
Selected publications
- Bengio, Y., Ducharme, R., Vincent, P., & Jauvin, C. (2003). "A Neural Probabilistic Language Model." Journal of Machine Learning Research, 3, 1137�1155.
- Bahdanau, D., Cho, K., & Bengio, Y. (2014). "Neural Machine Translation by Jointly Learning to Align and Translate." arXiv:1409.0473.
- Bengio, Y. (2009). "Learning Deep Architectures for AI." Foundations and Trends in Machine Learning, 2(1), 1�127.
- Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
- Bengio, Y. (2009). "Curriculum Learning." Proceedings of the 26th International Conference on Machine Learning.