Evidence of meeting #41 for Industry and Technology in the 45th Parliament, 1st session. (The original version is on Parliament’s site, as are the minutes.) The winning word was models.
A recording is available from Parliament.
Conservative
Conservative
Ted Falk Conservative Provencher, MB
Thank you, Madam Chair. I think you're doing an excellent job, by the way, for the record. I think Mr. Carr's job is on the line.
Mr. Buteau, I'd like to begin with you. You've talked about a very important part of AI, and that's the whole aspect of safety and controllability. Where do we go with all of this?
You talked about the arms race a little, and I liken it a bit to having a nuclear bomb. We don't have one, but we know that our best friends just south of us do, and that gives us a degree of comfort when all these other countries around the world also claim to have nuclear capabilities—and some of them do.
Is it not a good idea to have the capability that some of our potential adversaries around the globe may also have?
Consulting Program Officer, ControlAI
I think this is a question that's better answered with a concrete example to illustrate the two different threats we have to contend with.
As you said, perhaps there are some great powers we might want to acquire. Also, in the report that Anthropic put out, on page 55 it describes an incident where Mythos was confined to a secure computer inside of a secure network. They asked Mythos to escape the secure computer and send a message. Mythos succeeded, so you see the first half—the great power that we want—but this is only the beginning of the story. Mythos also developed an exploit to gain broad Internet access on its own. Then, without being asked to, Mythos bragged about the exploit on multiple public-facing websites, once again, on its own.
Yes, we have to think about not only people tasking AI to do dangerous stuff but also AI doing dangerous stuff on its own. It's not really like a nuclear weapon. It's like a nuclear weapon that sometimes goes up on its own, makes a bigger nuclear weapon and tries to take over your government.
That being said, I will repeat what I said. We don't call for unilateral disarmament. Right now, the United States is in the lead, so we don't call for the west to unilaterally disarm and cede the race to China. Wherever it is developed first, it threatens all of our lives, so we have to enter a stable regime where costs are imposed on people who are pushing forward, and they also get the reassurance that the other side is not pushing. That's the only stable equilibrium to limit the race toward superintelligence.
Conservative
Ted Falk Conservative Provencher, MB
I'm not sure how familiar you were with Bill C-27, when the Liberals attempted to pass that during the last Parliament. If you're at all familiar with it, though, what aspects of that bill should we keep, and what needs improvement?
Consulting Program Officer, ControlAI
Unfortunately, at the time, I was still in the lab, trying to find ways to make powerful AI controllable. Unfortunately, I don't have familiarity with that bill.
Conservative
Ted Falk Conservative Provencher, MB
Okay.
I know you've mentioned several times here that we get collaboration amongst stakeholders so that we have guardrails, fences or parameters that we do not exceed. How would you propose to enforce that?
Consulting Program Officer, ControlAI
Let me be absolutely clear. If China or any other country tries to develop superintelligence, we have to stop them at all costs. Superintelligence is a technology that, if developed anywhere, threatens our lives and threatens our global security, and we should treat it accordingly.
What does that mean? If we see a country threatening to develop it, Canada's reasonable reaction would be to use our full suite of deterrents to stop them. In this case, we need collaboration with as many countries as possible to give some teeth to our deterrence. The escalation ladder might begin with diplomatic pressure, followed by economic sanctions in the case of continued non-compliance.
It's also important to realize that if we don't have a trust-but-verify regime, and if the U.S. is about to develop superintelligence and China understands the situation, then we might see kinetic action from China to try to stop the U.S. Therefore, to avoid this, the only stable way is to look at the precursors to superintelligence and either have them inside the trust-but-verify regime where they're limited, or if people try to develop them outside the regime, then we have to prevent them from maintaining or acquiring them.
Conservative
Liberal
Michael Ma Liberal Markham—Unionville, ON
Thank you, Madam Chair.
My questions are for Ms. Luccioni.
The objective of the Sustainable AI Group is to help businesses square their increased AI use with sustainable goals. To support this framework, the group claims that it is already working with “North European and European clients across tech, finance, and media”.
Pillar six of Canada's national AI strategy focuses on “Building trusted partnerships and global alliances”, with an emphasis on aligning standards.
How will you work with international clients to help accelerate the establishment of international reporting standards for large AI companies to ensure maximum transparency in the sustainability domain when conducting business operations?
Co-founder and Chief Scientific Officer, Sustainable AI Group
Thank you for your question. It's a great one.
What I've been seeing so far is a real intention to regulate AI, especially in the European Union. I've gone to Brussels a couple of times and spoken to the representatives there. The issue is that AI is not a single thing. It's not a single piece of technology. For example, we've been working with them to take the AI energy score project I mentioned and make it implementable via standards. I participated in a workshop a couple of months ago with representatives from the ISO, for example, that aimed to develop metrics and functional units—all that stuff.
The tricky thing with AI is that we need to figure out specific use cases, so we're starting with, for example, text generation and image generation. What's interesting is that a lot of businesses, especially in Europe and Canada, have ESG commitments, and they're trying to understand how AI is impacting these commitments. They have their reporting guidelines. They have the things they measure. They're trying to integrate AI, but they don't have the numbers themselves. Their shareholders and even employees, in a lot of cases, are saying, “We don't want to use this technology unless you give us some numbers because we have convictions.” We're trying to help them get these numbers.
Honestly, we're seeing a lot of interest.
Liberal
Michael Ma Liberal Markham—Unionville, ON
Thank you.
You also mentioned, in an interview for BetaKit, that there is a lack of transparency around how much energy AI models consume and that the information disclosed through sustainability reports released by big tech companies is lacking and inconsistent. Pillar one of our national AI strategy is centred on building trust in AI if this technology is to deliver on its promises.
Taking into consideration these concerns, what role will the group play in de-risking AI not only to minimize business risk for stakeholders but also to foster more transparent and trustworthy relationships with skeptical Canadian citizens?
Co-founder and Chief Scientific Officer, Sustainable AI Group
We definitely see ourselves as the middlemen or “middlepeople” between users of the technology—both individuals and organizations—and developers. I've been doing my research for a while now, and we have good relationships with AI developers. Often, they want to report something. They have things they want to share, but it's about finding that sweet spot between corporate secrets, to some extent, and reputational risk. It's also about the numbers they can share that are useful.
This is where SAIG, the Sustainable AI Group, plays a role. We find the metrics that can be communicated without impacting, for example, the reputations of these companies, and we help bridge the gap between sustainability reporting and IT. Often, it's two different groups of people. On one hand, it's those who don't necessarily understand AI. On the other hand, it's those who don't necessarily understand ESG and sustainability. It's really about bridging that gap.
Liberal
Michael Ma Liberal Markham—Unionville, ON
Thank you for that.
How can we shift the national perception of AI at present in a more positive direction?
Co-founder and Chief Scientific Officer, Sustainable AI Group
The government has a unique role to play because a lot of AI companies want to work with governments. We're seeing a lot of announcements. That could be an incentive for them to be more transparent. For example, you might have a tender offer, as a city, a province or the federal government. You say, “Okay, if you want to work with us or for us, you have to be communicating these metrics. You have to be saying how much energy you're using.” Once those numbers are out, other corporate clients can follow suit.
Liberal
Michael Ma Liberal Markham—Unionville, ON
Thank you.
I'll shift this a bit. In your paper published in April of last year—
Conservative
Conservative
The Vice-Chair Conservative Raquel Dancho
Thank you so much.
Mr. Ste-Marie, you have two and a half minutes.
Bloc
Gabriel Ste-Marie Bloc Joliette—Manawan, QC
Thank you, Madam Chair.
Ms. Desrochers and Mr. Laforce, I'd like to know whether, in your opinion, you were adequately consulted about the artificial intelligence strategy that will be unveiled this week.
Ms. Desrochers, you said that you were consulted as part of negotiations on CUSMA, or the Canada—United States—Mexico Agreement, and on an addition on artificial intelligence in a UNESCO convention.
Do you think the government gives sufficient consideration to your two organizations or the cultural sector in general in its consultations?
Executive Director, Coalition for the Diversity of Cultural Expressions
We have been consulted on CUSMA, but we would like to continue to be consulted as the negotiations move forward. I just wanted to clarify that.
On the national artificial intelligence strategy, first, I can say that we submitted a brief as part of the national sprint. However, the “What We Heard” report had some grey areas on the position of the cultural sector. As a matter of fact, I think part of the report was prepared using artificial intelligence. Some subtle details and minor nuances are not fully captured.
That said, we scheduled a meeting with Minister Solomon, along with 12 members of the Coalition for the Diversity of Cultural Expressions, when we found out that the cultural industry was not represented on the working group. That was in October.
I would also like to say that at the time, we were happy to hear Minister Solomon recognize the principles of authorization, remuneration and transparency, or ART, which I spoke about earlier, and the fundamental role of culture in Canada's sovereignty.
However, since then, we have experienced challenges maintaining contact with his team, and we have not heard Minister Solomon reaffirm his commitment on that front. We are still reaching out and would very much like to continue with this conversation.
Bloc
Gabriel Ste-Marie Bloc Joliette—Manawan, QC
Thank you, Ms. Desrochers.
I have 30 seconds left.
Mr. Laforce, were you consulted?
Executive Director, Copibec
I would echo what Ms. Desrochers has said, that we are still reaching out and we hope the strategy will include the measures we recommended. If it is unveiled this week, I'm sure we will meet with Minister Solomon to discuss everything that lies ahead.
Conservative
The Vice-Chair Conservative Raquel Dancho
I sincerely thank the witnesses for their testimony today.
We appreciate your presence here. You are now free to go. Thank you very much.
Colleagues, just before we conclude, the clerk wanted me to remind you that we'll be receiving a revised version of the mould-maker report this Wednesday. We will set aside the second hour on Thursday to, hopefully, complete it then.
Thank you very much.
If it's the will of committee, we will adjourn.
The meeting is adjourned.