Industry Committee on Feb. 5th, 2024

[Dr. Elissa Strome]

Thank you for the question. I'll give it a shot for you.

That report was authored by Professor Gillian Hadfield, who's the director of the Schwartz Reisman Institute at the University of Toronto. She's a Canada CIFAR AI chair, and I believe she was a witness at this committee last week or the week before.

As you can understand, Professor Hadfield is an expert in regulation and in particular has developed significant expertise in understanding AI regulation in Canada and internationally.

The policy brief that we published at CIFAR represented ideas that came from Professor Hadfield and her laboratory, her research associates and her colleagues, really looking directly at the need to be more innovative as we think about regulating AI. This is a technology that's moving very quickly. It's a technology with so many dimensions that we haven't explored previously in other regulated sectors.

I believe the point that Professor Hadfield and her colleagues were making was that as we think about regulating AI, we also need to be incredibly flexible, dynamic and responsive to the technology as it moves forward.

[Tony Van Bynen, Liberal, Newmarket—Aurora, ON]

How can the findings of that report inform the consideration as we develop the act we're considering now?

[Dr. Elissa Strome]

I think the important point is the one that I made in my opening statement: It's the the need for deep flexibility. As the technology develops quickly and the world moves quickly around the development, deployment and adoption of AI, the regulations also need to be flexible and dynamic and move quickly. Innovation will be necessary in how we approach the regulation of AI in Canada.

It's things like bringing together multi-stakeholder groups to provide insight, ideas, advice and expertise. It's learning from the processes and approaches that the private sector is taking in order to comply with the regulations under the AI and data act. It's bringing together academics with government and private sector experts to learn from experiences, perspectives and approaches.

I think it's that idea of being flexible, trying new things and really trying to stay perhaps just one step behind the advancement of the technology rather than the many steps that we are behind right now.

[Tony Van Bynen, Liberal, Newmarket—Aurora, ON]

Thank you.

You've been the executive director of the pan-Canadian AI strategy since 2018. Did the work on this strategy inform the drafting of the artificial intelligence and data act, and if so, how?

[Dr. Elissa Strome]

Not directly.

The pan-Canadian AI strategy at its inception was really designed to advance Canada's leadership in AI research, training and innovation. It really focused on building a deep pool of talented individuals with AI expertise across the country and fielding very rich, robust, dynamic AI ecosystems in our three centres in Toronto, Montreal and Edmonton. That was the foundation of the strategy.

As the strategy evolved over the years, we saw additional investments in budget 2021 to focus on advancing the responsible development, deployment and adoption of AI, as well as thinking about those opportunities to work collaboratively and internationally on things like standards, etc.

Indirectly, I would say that the pan-Canadian AI strategy has at least been engaged in the development of the AI and data act through several channels. One is through the AI advisory council that Professor Bengio mentioned earlier. He's the co-chair of that council. We have several leaders across the AI ecosystem who are participants and members on that council. I'm also a member on that council. The AI and data act and Bill C-27 have been discussed at that council.

Second—

[Tony Van Bynen, Liberal, Newmarket—Aurora, ON]

Thank you—

Go ahead.

[Dr. Elissa Strome]

That's fine.

I was just going to say that the AI institutes—the really central hubs of AI research, innovation and commercialization in the country—also had the opportunity to contribute and convene their members to contribute ideas and advice on the act.

[Tony Van Bynen, Liberal, Newmarket—Aurora, ON]

Thank you.

I think I'm out of time, Mr. Chair.

[The Chair, Liberal, Joël Lightbound]

Thank you, Mr. Van Bynen.

Mr. Garon, you have the floor.

[Jean-Denis Garon, Bloc, Mirabel, QC]

Thank you, Mr. Chair.

Professor Régis, I don't want to make any assumptions or age us unnecessarily. However, when I was young, I watched films on television. After 10 minutes or so, there would be advertisements. Persuasive tactics were used to try to sell me products. It was obvious that persuasion was involved and that, if I watched these things, I was explicitly consenting to having all sorts of items sold to me.

With all the artificial intelligence or non‑artificial intelligence algorithms out there, I find that it's now getting harder and harder to identify a persuasion tactic. This issue will become increasingly widespread. We're often asked to agree to something. However, the fine print makes it incomprehensible to the average person, or even to a highly educated person.

First, do you agree that it's becoming harder and harder to consent to these tactics? Second, how can the quality of consent be improved in this situation? Third, is there any way to improve the current bill in order to enhance the quality of consent?

That's a lot of questions. You have one minute and 15 seconds left. You can answer the questions in quick succession.

[Prof. Catherine Régis]

Yes. It's easier than ever to be persuaded. That's actually one of the strategies. It can involve a very personalized approach based on your history and some of your personal data. This is indeed a problem. In the case of children, the issue gives me even greater cause for concern.

This issue not only affects consumers, as you pointed out, but democracies in general. I'm concerned about being locked into bubbles where we receive only information that confirms certain things or that exposes us to less diverse viewpoints. This issue raises a wide range of concerns, which must be taken into account. That's my answer to the first question.

Now, what more can we do? As I was saying, we need to think about this. Consumer protection is also at stake. We could do more work on the provincial component. In a recent study carried out by the Canadian Institute for Advanced Research, millions of tweets were analyzed to find out how people across Canada viewed artificial intelligence. Contrary to what you might think, people sometimes have an extremely positive view of artificial intelligence. However, they're less critical and less aware of what this technology actually does in their lives and of its limits. We often hear about legal issues, for example, but this awareness is in its infancy.

One recommendation in the Quebec innovation council's report is to encourage people to develop a critical mindset and to think about what artificial intelligence is doing, how it can influence us, and how we can create a guide for defending ourselves against it. This must start at an early age.

[The Chair, Liberal, Joël Lightbound]

Mr. Bengio, do you have any comments?

[Yoshua Bengio]

I want to add something. The recommendations that I listed briefly at the start include a line that must not be crossed. Companies should not be allowed to improve artificial intelligence technology to the point where it can influence people better than a human being could. The impact may be devastating. We must build significant barriers to avoid reaching this dreaded stage.

[The Chair, Liberal, Joël Lightbound]

Thank you.

Mr. Masse, you have the floor.

[Brian Masse, NDP, Windsor West, ON]

Thank you, Mr. Chair.

One of the things we keep hearing is that we're supposed to somehow not wait any longer but also be within the same framework as our international counterparts, many of whom have not acted yet, or we don't even know where they're going. Do you have any advice on that?

We'll do a really quick round here. I get only two and a half minutes. I'll start with Mr. Cofone again, and then go around.

Do you have any advice as to how we even err on that? That's what we're doing. We've been told, “Hurry up—wait.”

[Ignacio Cofone]

I think part of the answer to that is correctly following the risk-based approach that this act is taking. This is because with a risk-based approach based on standards, rather than trying to makes specific rules for the specific technologies we have now versus the ones we'll have in five years, we'll be able to adjust as the technology changes. Avoiding the temptation to regulate the technologies we had a couple of years ago and focusing on being technologically neutral while at the same time putting enough content into the bill will allow us to be future-proof and be aligned with these international principles.

I think part of that relates to the question that was asked just before yours about the impossibility to meaningfully consent today to most of the data processing, because it is impossible to anticipate the inferential harms from AI. I think part of the answer is again following standards and focusing on things like privacy by design, data minimization and purposeful mutation. These are independent of an individual's consent. This approach will allow our laws to adjust to the different ways in which the inferential harms will mutate in the next 10 years, and it's similar to the approach the EU is taking for artificial intelligence.

[Brian Masse, NDP, Windsor West, ON]

I think that is my time. Thanks, Mr. Chair.

[The Chair, Liberal, Joël Lightbound]

Thank you, Mr. Masse.

Mr. Vis, the floor is yours.

[Brad Vis, Conservative, Mission—Matsqui—Fraser Canyon, BC]

Thank you, Mr. Chair.

Ms. Régis, you started by talking about the commissioner's independence. On November 28, the minister sent us a letter explaining the artificial intelligence commissioner's powers and office structure. I gather that you believe that the commissioner should have an office, financial resources and independent employees.

What do you think about the idea of creating an office that would report specifically to Parliament, to ensure the independence that you referred to?

[Prof. Catherine Régis]

The idea of setting up an independent body to make recommendations to regulators or to society stakeholders at large isn't new. A number of models can serve as an inspiration. This includes the ombudsman model, which we all know, particularly at the provincial level. There's also the Competition Bureau.

To briefly answer your question, it would be good to explore the idea of creating the position of a commissioner who would report completely independently, including to Canadians. The various current models could be studied to determine which model would be best.

[Brad Vis, Conservative, Mission—Matsqui—Fraser Canyon, BC]

Thank you.

Ms. Strome, you mentioned in your opening testimony that we need to invest in subject matter experts in the Department of Industry. I'm very concerned about this. We know that artificial intelligence operates not just in Canada; it's also global. Even if we have a regulatory approach in Canada, if this bill is indeed passed, I don't think we can isolate ourselves from the potential societal and individual harms that will come from AI actors in other parts of the world.

I remember a few years ago that the Government of Canada—and I'm not trying to make a political point here—had a hard time operating its payment schedule for public servants.

[Brian Masse, NDP, Windsor West, ON]

It still does.

[Brad Vis, Conservative, Mission—Matsqui—Fraser Canyon, BC]

It still does.

How in the world is Industry Canada going to regulate online harms from AI? They can't even manage their own pay systems. I just don't know if our public service is nimble enough right now to do the job we need it to do in the format suggested thus far.

[Dr. Elissa Strome]

Let me offer an opportunity and a suggestion.

There's a lot of expertise in the public sector, in the academic sector and in the private sector in advancing and thinking about responsible AI and measuring and assessing its impact and harms. I think that's a huge opportunity for the Government of Canada to bring some of that expertise into the government, either temporarily or on a case-by-case basis, to assist the government in developing and monitoring and evaluating the risks associated with AI.

[Brad Vis, Conservative, Mission—Matsqui—Fraser Canyon, BC]

Let me just stop you right there. Are you suggesting somewhat of a hybrid model, whereby private sector actors are integrated with public officials to monitor and regulate, and maybe even make quick decisions on potential harms Canadians face?

I'll let Mr. Bengio jump in right after that quickly.