Samy Bengio (born 1965) is a Canadian computer scientist specializing in artificial intelligence and machine learning. As of 2025, he serves as senior director of Artificial Intelligence and Machine Learning Research at Apple, where he leads research efforts in areas such as deep learning, neural networks, and natural language processing. Bengio has authored around 250 scientific papers and is recognized for early work that helped spark the deep learning revolution of the 2010s.
Bengio spent over a decade at Google (specifically Google Brain) before joining Apple. He is also an adjunct professor at École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland and a co-author of Torch, the predecessor to the widely used PyTorch machine learning framework. His brother, Yoshua Bengio, is a Turing Award winner and a prominent figure in deep learning.
Education
Bengio completed his B.Sc. in Computer Science in 1986 at the Université de Montréal. He continued at the same institution, earning an M.Sc. in Computer Science in 1989 with a thesis titled "Integration of Traditional and Intelligence Tutoring Systems." In 1993, he obtained his Ph.D. in Computer Science from the Université de Montréal, with a thesis titled "Optimization of a Parametric Learning Rule for Neural Networks." His doctoral work focused on improving learning algorithms for neural networks, laying groundwork for his later research.
Early Career and Academic Positions
From 1999 to 2007, Bengio worked at the IDIAP Research Institute in Switzerland, a research center focused on machine learning and multimodal interaction. During this period, he also held a position at EPFL, where he conducted research on statistical learning and neural networks. His work at IDIAP and EPFL contributed to advances in speech recognition and sequence modeling.
In 2024, Bengio was appointed adjunct professor in Computer and Communication Sciences at EPFL, formalizing his ongoing ties to the institution. This role involves mentoring graduate students and collaborating on research projects, though his primary affiliation remains with Apple.
Contributions to Deep Learning
Bengio's scientific contributions are documented in approximately 250 papers indexed by DBLP, covering topics such as neural networks, machine learning, deep learning, statistics, computer vision, and natural language processing. Some of his most cited works are among the early studies that helped catalyze the deep learning boom of the 2010s. These include research demonstrating how to explore the many learned representations produced by deep learning models, which influenced subsequent architectures and training techniques.
He also contributed to one of the first deep learning approaches to image captioning, a task that involves generating textual descriptions of images. This work combined convolutional neural networks for visual feature extraction with recurrent networks for language generation, a paradigm that became influential in the field.
Another notable contribution is his research on understanding why deep learning works, which led to a series of follow-up studies examining the properties of learned representations and optimization dynamics. Bengio also worked on one of the first demonstrations that adversarial examples can exist in the real world, showing that modifying a physical object can fool a machine learning system. This finding had significant implications for the security and robustness of AI systems.
Additionally, Bengio was involved in early work on zero-shot recognition, which aims to recognize classes never seen during training by leveraging semantic relationships between known and unknown categories. This research opened avenues for extending models beyond their training data.
Professional Activities and Leadership
Bengio has held several leadership roles in the machine learning community. He served as program chair of the International Conference on Learning Representations (ICLR) from 2015 to 2016 and was a board member from 2018 to 2020. He was also program chair of the Conference on Neural Information Processing Systems (NeurIPS) in 2017 and general chair in 2018, and he continues to serve as a board member of NeurIPS.
He is an editor of the Journal of Machine Learning Research, a leading publication in the field. His editorial work involves overseeing peer review and shaping the journal's scope.
Bengio is a co-author of Torch, an open-source machine learning library that served as the ancestor of PyTorch, one of the two largest machine learning frameworks used today alongside TensorFlow. Torch introduced a flexible tensor library and automatic differentiation, which influenced later frameworks.
Time at Google and Departure
Bengio spent many years at Google, where he led a large group of researchers working on machine learning, including adversarial settings. His team focused on understanding vulnerabilities in models and developing defenses against malicious inputs.
In December 2020, Bengio left Google shortly after the company fired Timnit Gebru, a co-lead of the Ethical AI team, without first notifying him. Bengio later stated that he was "stunned" by the events surrounding Gebru's dismissal. His departure drew attention to tensions within Google regarding research ethics and diversity, and he subsequently joined Apple in 2021 as senior director of AI and Machine Learning Research.
Current Work at Apple
At Apple, Bengio leads research in artificial intelligence and machine learning, focusing on areas such as on-device learning, privacy-preserving AI, and applications of deep learning to Apple products. His team works on advancing the state of the art in areas like computer vision and natural language processing, while aligning with Apple's emphasis on user privacy.
Bengio's role involves both directing research and collaborating with other teams across the company. He has also continued to publish academic work and participate in the broader research community, including serving on conference committees and mentoring young researchers.
Personal Life
Samy Bengio was born to two Moroccan Jews who emigrated to France and Canada. His father, Carlo Bengio, was a pharmacist who wrote theatre pieces and ran a Sephardic theatrical troupe in Montreal that performed Judeo-Arabic works. His mother, Célia Moreno, is an artist who acted in major theatre scenes in Morocco during the 1970s, including productions run by Tayeb Seddiki. Both Samy and his brother Yoshua lived in Morocco for a year during their father's military service there.
Yoshua Bengio, Samy's younger brother, is a renowned computer scientist and a recipient of the Turing Award, often considered the Nobel Prize of computing. The brothers have collaborated on research projects over the years, though they have pursued distinct career paths.
Legacy and Impact
Bengio's work has had a lasting impact on machine learning, particularly in the areas of deep learning and adversarial robustness. His early papers on representation learning and image captioning are widely cited and have inspired numerous follow-up studies. His contributions to Torch helped shape the modern deep learning toolkit, and his leadership in conference organization has influenced the direction of the field.
As of 2025, Bengio continues to be an active researcher and leader, bridging academic and industrial AI research. His career reflects a commitment to both fundamental scientific inquiry and practical applications of machine learning.