|
|
| Name |
Bengio, Yoshua |
| Location
|
Universite de Montreal |
| Primary Field
|
Computer and Information Sciences |
|
|
 Research Interests
|
|
Dr. Bengio’s research in machine learning and artificial intelligence focuses on uncovering the principles of intelligence through the development of novel learning algorithms and deep neural architectures. His work investigates representation learning, generative modeling, and probabilistic reasoning as key avenues for enabling machines to acquire and generalize knowledge from limited data. In early 2023, Dr. Bengio shifted focus from enhancing AI capabilities to prioritizing AI safety, with particular emphasis on safe AI by design to avoid the emergence of self-preserving and deceptive behaviors. This has led to the development of non-agentic AI frameworks such as “Scientist AI,” which emphasize understanding, explanation, and prediction rather than autonomous action. These systems are conceived as safety layers that probabilistically evaluate the reliability of other AI models and block potentially harmful outcomes. His group is studying methods to embed calibrated uncertainty, humility, causal reasoning, and interpretability into deep learning architectures, with the aim of building models that are robust, transparent, and aligned with human values. These directions seek to ensure that artificial intelligence develops in directions that are beneficial, trustworthy, and safe for humanity. |
|
|
|