Canada can lead in AI for Science - Vector Institute for Artificial Intelligence
The Vector Institute’s celebration of its first five-year milestone highlights the role that AI can play in the Canadian economy and society. AI is no longer considered a new technology to most of our stakeholders, but its transformative potential has yet to be realized in many sectors, and many Canadian companies continue to lag behind their international competitors in the adoption of AI-enabled tools.
Over the next five years, Vector will continue to focus its work in areas where we can have the most impact, intensifying investments in areas of existing AI strength, such as health, life sciences, and trustworthy AI, as well as moving into new strategic ones.
AI for Science, which we take to mean using AI-enabled tools to push the boundaries of science and engineering, is an area that Canada can lead. Our strategic roadmap includes bolstering research capacity in the application of AI to materials and drug discovery, and quantum applications, both promising AI for Science subdisciplines.
A recent quote by the Council of Canadian Academies highlighted that AI has the “potential to spur innovation and further scientific understanding beyond the limits of the human mind and abilities.” With a growing team of world-renowned scientists deploying AI to further scientific discovery, Vector aims to work with the other two national AI institutes in the Pan-Canadian AI Strategy (PCAIS) to make this potential statement a reality.
CIFAR announced last month an additional cohort of eight phenomenal new Canada CIFAR AI Chairs, who will help advance Canadian leadership in priority areas under the PCAIS. Seven are affiliated with Vector, including Anatole von Lilienfeld, who recently joined the University of Toronto (U of T) from the University of Vienna. Von Lilienfeld’s team works on theoretical and computational methods for the quantum mechanics-based exploration of chemical compound space, and he joins Vector Faculty Members Juan Felipe Carrasquilla Álvarez who works at the intersection of condensed matter theory, quantum computing, and machine learning, and Alán Aspuru-Guzik who, with his group the Matter Lab, works at the interface of theoretical chemistry with physics, computer science, and applied mathematics. Together they will be a catalyzing force, enabling Vector to deepen its focus on AI for scientific discovery.
U of T recently announced the launch of the Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship, a program of Schmidt Futures, which brings together natural science and engineering postdoctoral fellows who are interested in applying AI to their domain. Working with both U of T’s Data Science Institute and the Aspuru-Guzik-led Acceleration Consortium, the Vector Institute is eager to support 10 postdoctoral fellows who will be awarded the Fellowship in the program’s first year to create an inter-disciplinary community of AI-fluent scientists making an impact in their respective domains. This total number of postdoctoral fellows will also double next year—from 10 to 20 and years thereafter.
Schmidt Fellows will become integral members of the Vector community and participate in the programming that we offer. Our shared programming will help foster a strong sense of community between Vector postdoctoral fellows, our affiliated institutions, and the newly awarded Schmidt Fellowships. Vector will also facilitate and encourage further collaboration through informal mentorship supported by Vector Faculty Members. For prospective applicants to the Schmidt Futures program, I would encourage you to explore Vector Faculty as potential supervisors or co-supervisors.
Vector’s commitment to AI for Science also includes our growing expertise in foundation models: large general-purpose models trained on broad data at scale, and later specialized for specific tasks. Scalable representation learning, as embodied by such foundation models, will be the key to our success in AI for Science.
Language- and vision-based foundation models BERT, DALL-E, GPT-3 have transformed AI, as they are a high-tech industry asset. Vector intends to define, synergize, and grow a program that utilizes our experiences running a successful NLP project to yield rewards in these new domains, better democratizing these technologies.
The Vector Institute is proud to announce that two of the new CIFAR AI Chairs are natural language processing (NLP) experts who could support this initiative. Vered Shwartz, who recently joined the University of British Columbia, aims to build models capable of human-level understanding of natural language. Wenhu Chen, who recently joined the University of Waterloo, uses deep learning and multimodal learning to incorporate world knowledge into deep neural networks, helping them make more accurate and transparent predictions.
Shwartz and Chen, who work at scale and have deep domain expertise, will help see Vector develop foundation models for data streams such as single-cell “-omics”, DNA barcodes for biodiversity monitoring, and digital pathology.
These announcements are just the beginning of Vector’s AI for Science initiative. I’m excited about this next phase of Vector’s mission, as we transition from our creation and gathering to applications that will harness the transformative potential of deep learning and machine learning to help solve global challenges.