Thank you very much.
Mr. Chair, members of the committee and fellow panellists, my name is Dr. Jaron Chong. I'm the chair of the Canadian Association of Radiologists' standing committee on artificial intelligence. I'm also an assistant professor of radiology at Western University here in London, Ontario, and a body imager at Victoria Hospital.
The CAR represents Canadian radiologists and represents almost 2,900 members who provide medical imaging for millions of patients across Canada. Radiology is at the forefront of technological innovation in medicine, relying heavily on the contributions and developments of advanced technologies to enhance patient care.
These breakthroughs in imaging technology and research have led to almost an exponential growth of imaging data over the past few decades, which has then been applied back to health care questions and workflows, particularly most recently in the domain of artificial intelligence.
In 2017, the CAR established a standing committee on AI to deliberate on the practice, policy and patient care issues related to the implementation of AI in medical imaging. Through a series of highly cited white papers, contributions to scientific forums and engagement with policy-makers in Canada and abroad, the CAR has been a leader in the international conversation about AI.
I say all of this and realize that this is a session on quantum computing, and I am not an expert on mechanics or computing in that way. What I do represent, however, is what we hope will be one of the ultimate end point applications of quantum computing, particularly as it relates to AI, to help optimize health care and medical imaging.
From the health care perspective, quantum computing may not necessarily solve new classes of problems that are not currently tackled right now with conventional computing, but they may vastly accelerate the computational speeds of our most NP-hard, difficult training projects and experiments, and greatly expand the size and scope of the clinical problems we tackle. Really, we do see that conventional digital computing and quantum are mutually complementary and will almost certainly coexist for a very long time.
However, what we're most excited about is that we expect the speed at which we can train algorithms will improve by orders of magnitude. Imagine training a neural network to detect lung cancer on a CT scan in minutes instead of days to months, or—as was previously mentioned—developing a novel chemotherapy molecule for mass production in simulation, as opposed to years and years of lab experimentation.
If there's one lesson that radiology has learned about AI in the past five years, it's that the computation and the algorithms can actually change by the year and by the week, but the datasets being used to train those algorithms are a much longer-term investment, so the careful curation of datasets has remained useful from 2017 to 2022 and beyond.
Regardless of whether you're thinking about conventional or quantum computing, the amount of curated, labelled data harnessed to optimize all these patient outcomes, ensure appropriate care and enhance the efficiency of the entire system is very much a “garbage in, garbage out” metaphor. Our current work on AI right now is hindered sometimes more so by the amount of time it takes to clean and curate good data than it is by the computational capacity. I will make a metaphor: A faster car doesn't get you there faster if your roads are still full of potholes.
What we need to ask ourselves about right now is what long-term policies and investments in better data today will position Canada to be creative, competent and competitive for our health care AI needs of tomorrow, and for quantum AI, as well.
We feel that that, during the last AI revolution, investments in centres of excellence and basic science enabled Canada to play an international leadership role that was vastly disproportionate to our size and population. The real challenge is maintaining our competitive edge and retaining the benefits of our investments as these innovations are applied to various sectors. We've often seen that we invest in the short term on a cyclic manner, but the downstream benefits of those investments were oftentimes difficult to fully realize for Canadians on a population level over the long term.
From a health care perspective, we have to accept the very realistic probability that the majority of health care AI used on Canadian patients will not have been developed or trained on Canadian data. If this is the case, are we prepared to accept the consequences of imported biases, failure to perform or failure to generalize, or even the economic significance of importing them and not solely exporting applications?
In a postquantum computing landscape, we would expect that the strengths and the weaknesses of data infrastructure would be magnified. Those who have the pipelines will run faster. Those who do not will fall behind or perhaps find themselves buying from another.
If you are a decision-maker, we want you to know that we still think there's a dramatic need for investments in digitization and data collection. We need to ensure that the data we are collecting is good data that meets our current and future needs, and we need to improve our data infrastructure to facilitate data sharing to empower investigators, while also safeguarding the rights of patients and privacy.
We do need to continually invest in the basic sciences and fundamental research that will help make the promise of quantum computing in health care and real-world applications less of a far-off proposition. We've seen that with earlier efforts to advance AI that Canada definitely has the talent and the technical know-how to lead in this field and in many others. What will make a difference for Canadian patients and the health care system is if we can find a way to incentivize innovators to develop and implement their technologies here at home.
I welcome any questions you may have, and I look forward to the hopefully very interesting discussion coming up next.
Thank you very much.