Industry Committee on Dec. 5th, 2023

[Jean-François Gagné]

I think these are good guideposts. An enormous amount of work was done by the international community to understand the issues. I think that many of the things I was reading about in Bill C‑27 and the amendments are valid, and I could identify which portions were intended to cover health or a specific aspect of biotechnology. I could really tell. However, it seems to want to cover all industries Canada, from the smallest to the biggest. What's really needed is to think carefully about them, make adjustments, and if there are specific situations, work with these sectors, while concurrently protecting people and being careful not to hinder innovation.

That's really my greatest concern. I have friends who are entrepreneurs, I'm an entrepreneur myself, and reading this worried me. It's already difficult to innovate and try to stand out from the crowd. If it becomes even more expensive to develop and launch products, it would make things more complicated.

[Sébastien Lemire, Bloc, Abitibi—Témiscamingue, QC]

I believe my speaking time has run out. Thank you very much.

[The Chair, Liberal, Joël Lightbound]

Thank you, Mr. Lemire.

Mr. Masse, you have the floor.

[Brian Masse, NDP, Windsor West, ON]

Thank you, Mr. Chair.

Thank you to the witnesses.

Ms. Ifill, if I can, I'll go to you. One thing I had a chance to do this summer was attend some conferences in the United States. They had some of the larger players that are developing artificial intelligence, and they identified what you spoke about. In their modelling that they're doing now, a lot of biases currently exist.

Can you speak more to that in terms of how it can stream people and stream ideas if we don't have the right people building artificial intelligence models in the first place that reflect more of society versus models that aren't inclusive?

[Erica Ifill]

There are many ways this could materialize into something that is not beneficial for some groups. For example, predictive policing is one way that we see artificial intelligence in use to predict criminal activity, but the training data that's used is historical. If you're using historical or certain types of data to train the AI system, you're going to get a compounded effect whereby those neighbourhoods that are overpoliced become even more policed.

Another way it comes about is in hiring. Hiring agencies have used AI to search for executives for executive positions. Unfortunately, a lot of that data is also historical, which means there's a bias against women, because traditionally, women haven't held those positions.

These are very real consequences that are at scale, and I think the scale and the speed at which this could happen are very concerning. I believe the Edmonton police recently used a system using DNA to predict the facial features of one of the suspects of a sexual assault, and what it came up with was a 14-year-old Black boy. That's the other thing. This adultification of Black boys is another way AI manipulates what we see and what we consider as victims and as perpetrators, or anything like that.

I think the problem is that it has to do a lot with the training data, but the systems.... I'm not sure if the right questions have been asked or the right assumptions have been made to create the model itself.

[Brian Masse, NDP, Windsor West, ON]

I appreciate that, and we haven't talked about it.

Mr. Schauer wants to get in here.

Please go ahead. I think that this is an important subject that we haven't really delved into too much because the modelling is critical.

[Adrian Schauer]

Yes, that would be an argument in support of Mr. Gagné's proposition, which is to legislate in the domain in which the model is used. In this specific example, in law enforcement, you have to take care to not do predictive policing that's biased. To say that the legislation should live at the model level could lead to very adverse effects of that legislation in a totally unrelated domain like health care or like transport with a self-driving car model. It's hard to say that we can find a legislation so perfect at the level of the underlying technology to cover the use cases in all these different domains.

That would be my reflection on it.

[Brian Masse, NDP, Windsor West, ON]

Thank you.

Ms. Castets-Renard, please.

[Madam Céline Castets-Renard]

Thank you.

I'd like to add something about Bill C‑27. A risk-based approach would avoid treating all artificial intelligence systems in the same way, or placing the same obligations on them. Other options include the high-impact concept, and the amendments introduced by the minister, Mr. Champagne, explain what this concept means in seven different sectors of activity.

I therefore don't think it's fair to say that it would be applied everywhere, on everyone, and haphazardly. It's possible to discuss how it's going to be applied in seven different activity sectors. Some, no doubt, would say that doesn't go far enough, but it is certainly not a law that will lack specifics, because the amendments specify the details.

To return to what was said earlier, it also means that there can be a comprehensive approach with general principles, and an separate approach for each sector or field. That's what the European Union has done with its amendments. That's why statutes being adopted in other countries need to be considered.

As for what was said about the United Kingdom earlier, Canada has signed a policy declaration which has no legal or binding value. It's a very general text that adds nothing to what we have already said about the ethics of artificial intelligence. It definitely does not prevent Canada from following its own path, as the United States did when it issued its executive order right before the summit in England. The Americans were not willing to wait for England to take the lead.

Those are the details I wanted to add.

[Brian Masse, NDP, Windsor West, ON]

Thank you, Mr. Chair.

Do I have any time left?

[The Chair, Liberal, Joël Lightbound]

You don't, but I see that Madam Ifill has her hand up and that Mr. Harris does, as well, so we'll take both interventions before we go to Mr. Vis.

[Erica Ifill]

Thank you.

I do have an issue with our not providing accountability for these harms that haven't really been laid out properly. I also have a problem with the little accountability that we have in the bill today: a commissioner that really doesn't have any sort of public responsibility. Yes, we can say that we don't want to legislate this or that we don't want to legislate that, but we know the harms are here now. I don't want a big swath of Canadians to be part of those unintended consequences when we know what the consequences can be.

[The Chair, Liberal, Joël Lightbound]

Mr. Harris.

[Jérémie Harris]

I just want to circle back to this notion of whether you regulate the model or the end applications. That is pretty central here. We're going to have to walk and chew gum at the same time. There are risks that, irreducibly, come from the model. Look at OpenAI's ChatGPT, for example. They build this one system, one model. I don't know if I can.... In fact, I know that I can't. I know that no one, technically, can count the full range of end-use applications that a tool like that would have. You'll use it in health care today, and you'll use it in space exploration tomorrow and software engineering the day after.

The idea that we're going to be able to take a general-purpose model like this and regulate it as if somehow we can play this losing game of whac-a-mole.... This is just not going to track reality, unfortunately. This is true for a certain subset of risks—the more extreme ones. We can look at the risks, for example, from general-purpose models that can orient themselves in the world and have high context-awareness. You have to regulate the model at that point because that is the source of the risk, irreducibly.

For other things, yes, we need to have application-level regulation and legislation. Again, you see that in the executive order—that we're doing both things. However, I just want to surface that although there might seem to be a tension between these two approaches, they are actually not at all incompatible. In fact, in some ways, they are deeply complementary.

I just wanted to prop that thought.

[The Chair, Liberal, Joël Lightbound]

Thank you very much.

Mr. Vis, you have the floor.

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

Mr. Gagné, you said that we needed an act that could track and adapt to the technological evolution of artificial intelligence. Does the department have the capacity to monitor this technological evolution? Can it really meet this challenge at the moment?

[Jean-François Gagné]

I wouldn't think so.

I think it would be hard.

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

Thank you.

Madam Quaid, you mentioned the model where we adopt a competition bureau for artificial intelligence. You're the first person to raise that suggestion during our testimony.

Can you just expand a bit on that, maybe in one minute or less? What would it look like in practical terms?

[Dr. Jennifer Quaid]

You've given me a challenge. I'll do my best.

The idea is the following.

I'm not saying that we absolutely imitate the Competition Bureau; there are some things that we could do differently. The kind of legislation that is imagined here has some similarities to the kind of legislation that is in the Competition Act. That is to say, it's responsible for a whole array of responses: true criminal, regulatory, administrative and civil. There is a specialized tribunal with that, but we don't need to talk about that right now.

I think the point is that it has developed an expertise and it has a large permanent staff divided into directorates. It's developed a digital intelligence agency.

Those things support what I think other witnesses have been skeptical about, which is the capacity to actually deliver on this.

The U.S. has basically not made a secret of it. They've just said, “Let's use our strong antitrust institutions while we wait to create something else”. In some ways, we would be consistent with what's being done there.

The other thing I really want to insist on—sorry, it's an extra 10 seconds—is that having an agency headed by an independent commissioner will allow Canada to participate in the international arena. That is how you get around these enforceability problems: You have to work with your friends.

In Canada, we might only target local things, but we need to work with allies and we need a player at the table for that.

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

Thank you. It almost reinforces the apprehension expressed in my first question about the capacity within the department, so that's much appreciated.

Mr. Harris, you painted a scary model in the beginning.

What do we have to look forward to with AI? I don't think it's all bad.

[Jérémie Harris]

You're quite right. That's part of the paradox here. We're talking about intelligence, as was highlighted earlier. This is the most general purpose thing humans have. It may be the most general purpose tool ever.

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

Just out of curiosity.... The Abbotsford hospital has a cancer treatment centre. I'm wondering what role artificial intelligence will play in the treatment of cancer in the next 20 years.

I can't imagine any scenario where we're doing it the same way as we are today, because of this technology.

[Jérémie Harris]

To land the plane and give you concrete timelines here, when you talk to folks in the frontier labs, the median, totally reasonable supposition they'll give is that they think it could be anywhere from two to five years to reach human-level AI across the board.

Let's say they're completely off.... That kind of estimate and the amount of change that's implied by that in areas like health care.... It's not just health care, though. In material design, Google DeepMind just came out with a paper where it made essentially 800 years of progress in material science in a couple of months. There's huge potential here.

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

When I was in university—over 20 years ago now; I can't believe it—we learned about the impact of the Gutenberg press.

Is the moment we're going through right now even larger than the Gutenberg press?

[Jérémie Harris]

I don't want to be flippant about this, but I think it's clearly bigger than that. Again, the one thing that distinguishes human beings from other species on this planet is the thing that we are right now trying to figure out how to bottle up on a bunch of servers.

Yes, there's huge potential, but great power...great responsibility. I've seen Spiderman as well. I can quote scripture.