A leading expert in cybersecurity and two renowned AI safety researchers are set to take on leading roles at the University of Toronto’s Schwartz Reisman Institute for Technology and Society.
David Lie, who is known for his seminal work that led to modern trusted execution processor architectures, has been named the new director of the Schwartz Reisman Institute (SRI), which aims to explore and address the ethical and societal implications of artificial intelligence and other emerging technologies.
His four-year appointment, which takes effect July 1, coincides with two renowned experts in AI safety — Roger Grosse and David Duvenaud — being named Schwartz Reisman Chairs in Technology and Society for five-year terms.
“I think one of the top priorities is ensuring that SRI and U of T are the primary places in Canada — and perhaps in the world — for AI safety discussion and research,” says Lie, a professor in the Faculty of Applied Science & Engineering’s Edward S. Rogers Sr. Department of Electrical and Computer Engineering.
“My vision is to make us one of the leaders. Canada has already contributed greatly to machine learning and AI through the contributions of previous scholars like [University Professor Emeritus] Geoffrey Hinton, and I think we have a very strong role to play in this important technology going forward.”
The appointments come as inaugural director and chair Gillian Hadfield prepares to conclude her term as chair this month (she stepped down as director at the end of last year). The institute, created following a historic gift in 2019 from business leaders Gerald Schwartz and Heather Reisman, brings together experts from disciplines across U of T’s three campuses to steer AI development to prioritize safety and human welfare.
“We are thrilled to welcome David Lie, Roger Grosse and David Duvenaud to their new roles at the Schwartz Reisman Institute,” says Melanie Woodin, dean of the Faculty of Arts & Science. “Their expertise and leadership will be instrumental in fostering the interdisciplinary collaboration needed for the University of Toronto to remain at the forefront of technological innovation that benefits humanity.”
Lie, who has served as a research lead at SRI and holds cross-appointments in the Department of Computer Science and the Faculty of Law, says his decades of research on making computer systems more secure and trustworthy — including contributions to computer architecture, formal verification, techniques using operating systems and networking — have equipped him to tackle the complex issues posed by AI, which will require researchers to anticipate and adapt to the unexpected.
“As AI become more powerful, they may do things — or are already doing things — that we didn’t anticipate or expect,” says Lie. “Bringing cybersecurity skills, thinking and tools into the AI safety discussion will be absolutely critical to solving the problem.”
Lie emphasizes that interdisciplinary collaboration is key to addressing potential AI disruption, noting that it has been pivotal in his own research and other roles.
His current research focuses on securing mobile platforms, cloud computing security and bridging the divide between technology and policy. He is also an associate director at the Data Sciences Institute, a U of T institutional strategic initiative, a faculty affiliate at the Vector Institute for Artificial Intelligence and a senior fellow at Massey College.
“It’s really one of the things that I love about a place like U of T, because it's big and you have experts in every imaginable field to collaborate with,” he says. “I feel very strongly that we can always accomplish way more together than we can individually. That's true for people, but that's also true for disciplines.”
As incoming Schwartz Reisman Chairs in Technology and Society, Grosse and Duvenaud have vital roles to play in driving SRI’s research agenda and sharing its findings with the world, says Lie.
“One of the main ways universities contribute to society is through research, but we also contribute through discourse; we contribute by translating knowledge and providing that to policymakers, decision-makers and stakeholders,” he says. “I see SRI playing an important part in these roles.”
Both Grosse and Duvenaud are associate professors of computer science in the Faculty of Arts & Science, faculty affiliates at SRI, founding members of the Vector Institute and Canada CIFAR AI chairs — and both are working at San Francisco-based Anthropic, a research company focused on AI safety and alignment.
Grosse, whose research applies our understanding of deep learning to the safety and alignment of AI systems, says academia has an essential role to play in guiding AI development by looking beyond short-term incentives to ask how these technologies can be safely and ethically integrated for the long-term benefit of humanity.
“I'm very excited to be able to understand and mitigate catastrophic risks from AI, to be part of an interdisciplinary community that's especially well positioned to make progress in these issues, and I really appreciate the leadership that donors are showing and supporting this work,” he says.
“I think academia is great for being able to ask the more fundamental questions, to carry out maybe more forward-looking research that might not be directly on a company's critical path, but will contribute to safety efforts at many different organizations.”
Duvenaud’s research, meanwhile, focuses on probabilistic deep learning, artificial general intelligence governance and dangerous capabilities evaluation.
He envisions SRI as a “centre of gravity” where academics, industry members, government leaders and other stakeholders can engage with each other and shape the future of AI technologies.
“The idea is that by having this institute dedicated to this direction, we’ll be able to do things like host visitors and engage with academics from all sorts of disciplines — such as law, economics, and other parts of civil society — so that, ultimately, when policy discussions come up, we’ll be equipped and credible as people who can help governments navigate these decisions,” says Duvenaud, who is cross appointed to the Department of Statistical Sciences.
Sheila McIlraith, an associate director and research lead at SRI, professor of computer science in the Faculty of Arts & Science, and a Canada CIFAR AI Chair at the Vector Institute, underlines the importance of rallying diverse disciplinary experts from across U of T to address the opportunities and challenges that AI will wield in the coming years.
“AI is no longer the sole purview of computer scientists. It is reshaping the way we live, work, and interact with each other, and it will take experts from a broad range of disciplines to help ensure that AI is developed and deployed for the benefit of humanity, and that Canada adapts swiftly to protect our institutions," says McIlraith, who is an expert in AI safety research herself.
“Threats are already upon us; now is the time to act.”