Research in AI

Prof. Alex Wong and his graduate students 

Foundational artificial intelligence

Focusing our research on important applications has a crucial benefit, beyond advancing the application areas themselves. It allows us to identify – and subsequently overcome – shortcomings in existing artificial intelligence (AI) approaches and so contribute to the foundations of the field. Key foundational research areas include:

  • Machine learning, statistical learning, and data mining
  • Probabilistic models, knowledge discovery, and knowledge representation
  • Intelligent agents and game theory
  • Optimization and decision making
  • Data science and analytics
  • Affective computing and sentiment analysis
  • Human-machine interaction

Waterloo’s AI research groups are conducting award-winning research into everything from computer vision to natural language processing to search-engine optimization to autonomous driving. Through the institute's projects, we are building tomorrow’s next-generation intelligent systems today, translating commercial and industrial requirements into deployable, real-time embedded AI.

Operational artificial intelligence

Artificial intelligence has captured the imagination of business and government. Research has focused on deep learning using big data. A key goal is to improve intelligence by building larger and deeper architectures, leveraging near-unlimited computing power and resources.

There is great promise in this approach, but the sheer complexities and resources required may exclude many smaller enterprises. A reliance on cloud computing is also leading to “tethered” AI, which introduces service, security, and data privacy concerns for many users. The resulting technologies may prove of limited use in a wide range of promising applications such as mobile medical care, assistive devices for the elderly, surveillance cameras, environmental sensing stations, and autonomous driving.

"Operational AI" at Waterloo takes a complementary, application-driven approach. We aim to develop lightweight, compact AI with highly effective intelligence, but with minimal computing power and energy requirements, suitable for a host of stand-alone applications.

The ultimate aim is to democratize AI, for use by anyone, anywhere, anytime:

  • Scalable AI, ideal for commercialization by both small startups and large corporations.
  • Compact AI, deployed wherever cost, energy, and bandwidth are limited.
  • Secure AI, with private data protected locally but metadata shared by cloud users.
  • Accessible AI, tailored for ease of use.
  • Dependable AI, with reliable performance regardless of connectivity.
  • Transparent AI, so the performance of ‘black box’ safety critical systems can be certified.