THE STANDING SENATE COMMITTEE ON HUMAN RIGHTS
EVIDENCE
OTTAWA, Monday, February 23, 2026
The Standing Senate Committee on Human Rights met with videoconference this day at 4:03 p.m. [ET] to examine and report on the impact of artificial intelligence on human rights and economic security in Canada, especially in relation to vulnerable groups and the international human right to work.
Senator Paulette Senior (Chair) in the chair.
[English]
The Chair: Good afternoon, everyone. I would like to begin by acknowledging that the land on which we gather is the traditional, ancestral and unceded territory of the Anishinaabe Algonquin Nation. I’m Paulette Senior, senator from Ontario and chair of this committee.
I now invite senators to introduce themselves, please.
Senator Karetak-Lindell: Nancy Karetak-Lindell, Nunavut.
Senator Robinson: Mary Robinson representing Prince Edward Island.
[Translation]
Senator Arnold: Dawn Arnold, New Brunswick.
[English]
Senator K. Wells: Kristopher Wells, Alberta, Treaty 6 territory.
The Chair: Thank you. Welcome, senators and all those who are following our deliberations. Today, our committee will begin its study on the impact of artificial intelligence on human rights and economic security in Canada — we’re very excited about this — especially in relation to vulnerable groups and the international human right to work.
This afternoon, we will have three panels. In each panel, we will hear from the witnesses, and then the senators around this table will have a question-and-answer session with them. I will now introduce our first witnesses, and our witnesses have each been asked to make a five-minute opening statement.
With us at the table from Innovation, Science and Economic Development Canada, we have Samir Chhabra, Director General, Marketplace Framework Policy Branch; and Katherine Burke, Director General, Foreign Investment Review and Economic Security Branch. Welcome to you both. Also at the table with us, we have Emmanuel Zangio, 1834 Fellowship Fellow and Policy Analyst. I now invite Mr. Chhabra to make his presentation followed by Mr. Zangio.
[Translation]
Samir Chhabra, Director General, Marketplace Framework Policy Branch, Innovation, Science and Economic Development Canada: Thank you, Madam Chair. Members of the committee, thank you for the opportunity to contribute to your study on the impacts of artificial intelligence on human rights and economic security in Canada.
Today, I will present the work under way at Innovation, Science and Economic Development Canada, or ISED, as well as other departments, to mitigate AI-related biases when it comes to human rights, Canadian jobs and Canada’s economic security.
[English]
The government is paying close attention to the human rights considerations surrounding AI, such as its potential to cause or amplify bias and discrimination. If not properly mitigated, bias in AI systems can disproportionately affect marginalized groups, particularly when AI systems are used to make consequential determinations or recommendations in high-impact contexts.
It is really important to note as well that existing laws continue to apply in the context of AI. For example, the Canadian Human Rights Act prohibits unlawful discrimination on the basis of protected characteristics, regardless of the technological means used. The government has, however, taken additional steps to address the potential discriminatory impact of AI in systems and products.
For example, since 2020, federal departments must comply with the Directive on Automated Decision-Making when using automated decision systems, including those that rely on AI in administrative decisions. The directive is intended to help ensure that decision-making processes are fair, transparent and unbiased.
Further, in 2023, the government launched the Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems. This code sets out practices that companies can implement to address risks associated with AI.
Signatories to this voluntary code notably commit to assessing and curating training data sets to manage data quality and pre‑emptively address potential biases, as well as to implement diverse testing methodologies. To date, 46 companies have signed on to this code, from small start-ups to large global firms.
In 2024, the government also invested $50 million over five years to launch our own Canadian AI Safety Institute, or CAISI. CAISI brings together top Canadian researchers inside and outside government to advance the science of AI safety, in collaboration with similar institutes across Europe, Asia, the U.S. and beyond.
I can give you a few examples of relevant work from CAISI, including recently announced funding through the Canadian Institute for Advanced Research to help build safer and more equitable AI models for linguistic minorities.
The National Research Council also works with us closely and has carried out a number of projects relevant to bias this year, including one on multilingual evaluations of frontier models, another on the explainability in vision language models and a third on how safeguards can be accidentally removed during fine-tuning of AI systems.
On the international stage, Canada cofounded the Global Partnership on AI, or GPAI, an international multi-stakeholder initiative integrated within the OECD with the mission of promoting the responsible development and use of AI through practical projects. The expert support centre for GPAI, CEIMIA, based in Montreal, launched policy guidelines on gender equality and diversity on AI with Mila in 2024.
Canada also supported the adoption of the UN Global Digital Compact, a framework that sets shared principles for inclusive, safe and responsible global digital cooperation. In addition, Canada signed the Council of Europe Treaty in February 2025, which is the first-ever international legally binding treaty aimed at ensuring that the use of AI systems is fully consistent with human rights, democracy and the rule of law.
[Translation]
Another important angle to consider is the impact of AI on Canadians’ privacy. The Minister of Artificial Intelligence and Digital Innovation has committed to modernizing the Personal Information Protection and Electronic Documents Act to strengthen privacy protection and ensure that AI-related risks are adequately addressed in Canadian law.
[English]
Turning to the impact of AI on work, the government is closely monitoring the expected impact of AI on Canadian jobs. AI notably has the potential to transform and improve the jobs and lives of many Canadians. To embrace this opportunity, the government has put in place several initiatives to support AI adoption. This includes the Regional Artificial Intelligence Initiative developed through Canada’s regional development agencies and the AI Assist program delivered by the National Research Council Canada.
However, AI may also impact workers and lead to disruptions in certain sectors. Through Budget 2024, the government invested $50 million in additional funding in the Sectoral Workforce Solutions Program to provide new skills training for workers in potentially disrupted sectors and communities. The government is also advancing discussions and initiatives on the impacts of AI on the workforce through its participation in international forums. During Canada’s G7 presidency in 2025, the G7 Employment Working Group developed a compendium of best practices for human-centred development and use of AI in the world of work.
In addition, the Global Partnership on Artificial Intelligence, which I referenced earlier, has the Working Group on the Future of Work, dedicated to examining the impacts of AI on the workforce, work environments, job quality, inclusiveness, health and safety.
Finally, with respect to economic security, the government has recognized that, as it continues to develop, Canada’s AI sector is an attractive target for foreign investment. Canada has the tools to ensure that such investments align with and protect our long‑term security interests. For example, under the Investment Canada Act, the government takes a measured, case-by-case view when investment proposals raise economic or security concerns.
[Translation]
As for the future, the government is working to develop a renewed artificial intelligence strategy. That should be out by the end of this quarter. The strategy will be developed thanks to the thousands of submissions received during the consultation that was held last October, as well as the report produced by the AI Strategy Task Force, which consists of 28 experts.
[English]
That concludes my remarks. I’m happy to respond to any questions. Thank you.
The Chair: Thank you, Mr. Chhabra. We’ll now proceed to our next presenter.
Emmanuel Zangio, 1834 Fellowship Fellow and Policy Analyst, as an individual: Thank you for the invitation to appear before you today. My name is Emmanuel Zangio. I am a public servant and 1834 alumnus. The views and recommendations that I share today are my own, and I am here in my personal capacity as a first-generation Congolese Canadian who is passionate about equity, artificial intelligence and economic security.
[Translation]
We are gathered here because AI isn’t an issue of the future. It’s already transforming our labour markets and institutions. The question is no longer whether artificial intelligence will change our daily lives, but how, and, above all, who will benefit, who will be protected and who is at risk of being left behind.
[English]
I speak to you today not only with my policy background, but also with knowledge rooted in my lived experiences. I grew up in a welfare community, and as a young, Black, dyslexic student who struggled with some aspects of school, I was raised by systems that were never designed with someone like me in mind. The only reason I am here today is because Canada chose, at key moments, to open the door for someone who, statistically, should not be here.
Artificial intelligence was one of those doors. AI tools helped me write, read and reach opportunities that would have otherwise remained closed to me. In many ways, it levelled the playing field.
However, I have also witnessed the flip side. Recently, a close family member lost their job. Without guidance in navigating the AI tools required to format their résumé, they would have been shut out of opportunities they urgently needed.
AI can open doors but also leave us outside.
That is the lived reality for many Canadians. Many are left behind.
[Translation]
AI is advancing at a faster rate than vulnerable people’s ability to adapt. Access to support often depends on whether a person has someone in their immediate circle who can help them. Many don’t.
[English]
Today, I offer three key messages for the committee.
First, AI literacy is now a science literacy issue and an economic security issue. Canada has world-class AI researchers but one of the least AI-literate populations among advanced economies. The divide is widening. Without intervention, we risk deepening inequality and losing competitiveness as a nation.
Second, existing federal programs are strong but inaccessible at times and rarely reach people who need them most. We are talking about adults outside formal education, older displaced workers, racialized or low-income communities and people without digital confidence. These Canadians face issues around transactional barriers, literacy, language, trust, awareness, transportation or simply not knowing such programs exist. AI cannot succeed if the people most affected by it cannot even enter the learning ecosystem.
Third and finally, Canada needs a coordinated strategy focused on underserved Canadians. I recommend a community-based model using libraries, newcomer agencies, community centres and mentorship from young students and early-career professionals who can make AI accessible. This approach is practical, scalable and aligned with Canada’s priorities in innovation, labour force readiness and equity.
Honourable senators, I stand before you not by my own strength but as the result of what this country can do when it chooses to uplift those who are vulnerable. The support and opportunities I received made it possible for me today.
[Translation]
I am a young 27-year-old New Brunswicker, and I’m happy to be here, educated, independent and committed to serving my community and those that are different from mine.
[English]
I would like to thank the Senate for giving me, a young man who is inexperienced, the chance to voice my opinions. Thank you very much.
The Chair: Thank you, Mr. Zangio and Mr. Chhabra, for your presentations today.
We will now proceed to questions from the senators. Senators, you have five minutes for your questions, which include the answers.
Senator K. Wells: My first question is for our government representatives here.
We have talked about the need for safeguards around AI, and we have seen increasingly disturbing examples of individuals in need of mental health support turning to AI instead of being directed to appropriate local support or care. We often find that the conditions they are experiencing become worse or exacerbated because of AI, tragically leading them to self-harm or harm others.
How do we go about addressing this in government legislation coming forward? We turn on the headlines today and see cases where there should have been intervention from AI companies but where nothing happened, leading to tragic, life-altering, devastating results.
That is just a general question around safeguards and how we ensure they are in government regulations — or the need for legislation moving forward with developing AI frameworks or policies. Thank you.
Mr. Chhabra: Thank you for the question.
We’re closely tracking the incidents that are being monitored and reported in the media. It’s important, first, to understand that AI is very much a general-purpose technology; it has multiple applications across the economy and the market. Some of those applications are consumer-facing chatbots, of which we have seen an exponential rise in use. I looked at some data today that suggested that approximately 70% of ChatGPT use of late has been on the personal front rather than around economic or business productivity; the ratio there seems to have flipped remarkably over the past number of years.
In the context of personal and therapeutic uses, Minister Miller has been clear that he and the government are considering what steps to take in the context of online harms legislation. I would point you to that as a space where the government has indicated an interest in doing some work. I can also highlight that the Canadian AI Safety Institute has taken an interest in this space, as well, and is continuing to work with international counterparts to monitor this space.
Senator K. Wells: I turn your attention to something else that you were talking about: the concern about unemployment, particularly youth unemployment. We know that AI can take away some of those entry-level positions through which young people are gaining valuable work experience. Is the government considering developing a strategy to protect young people from rising unemployment and the potential loss of jobs due to AI? What is the plan to help support these young people, not only in terms of learning to harness AI — the strengths and challenges of AI — but getting jobs?
Mr. Chhabra: Thank you for the question. It’s a very important one.
I will highlight a few things. First, Minister Solomon has highlighted that the AI strategy will be forthcoming within this quarter. The AI strategy was developed through consultation and engagement with 28 experts, as I mentioned. We also received more than 11,000 public submissions, which is a really significant amount of interest. A number of those submissions highlighted the importance of skills and opportunities development. I believe the strategy coming forward will certainly have something to say about how we tackle skills, development and training, more broadly, for Canadians.
As I highlighted in my opening remarks, the government has already made an investment in this space — $50 million — managed through Employment and Social Development Canada to help with re-skilling.
The last point I will raise is that StatCan has been tracking this space closely and will continue to do so. To date, looking at the most recent figures between 2022 and 2025, which was when ChatGPT was first emerging as a key consumer-facing tool, there wasn’t a definable impact; the labour market participation rates, engagement and levels of joblessness have not significantly changed. There is no clear evidence to date that jobs that are potentially more exposed to AI experience disproportionate declines compared to other occupations.
The issue is certainly one that is real. It’s certainly one that Canadians are quite concerned about. The respondents to the consultation flagged it as a significant risk in that space, for work. Certainly, the government takes that very seriously. At this point, we haven’t seen significant data on that, but that data could still be emerging over time, given how relatively new AI use is in a professional sphere.
We haven’t seen it yet. We are, certainly, monitoring it very closely, and we expect that the AI strategy will have more to say about the government’s approach to ensuring that AI opportunities are available for everybody and that we’re taking the skilling, development and diffusion part of this very seriously.
Senator Arnold: I would like to start with my fellow New Brunswicker, Mr. Zangio.
I wonder if you could talk a little bit more about how trust is a factor in the acceptance of AI across society and the effect its implementation will have on vulnerable groups. You touched on it. Maybe you could go into a bit more depth.
Mr. Zangio: Yes, definitely.
I can start by saying that I think that with my lived experience, as I mentioned, early on in my life — I think in Grade 5 or 6 — I was introduced to the first assisted AI tools, like speech-to-text. As a dyslexic kid, I didn’t have the ability to read by myself. The concern of my parents at the time was that if they are giving our child these types of tools early on, will he be able to be self‑sufficient when he gets older? That is a real aspect.
I think that we’re making some good strides in implementing artificial intelligence in our government, in our provincial and national jurisdictions, but I think that needs to continue: building that trust and saying, for example, that AI tools are not here to take away autonomy but to help us, as I said, level the playing field.
If that message to Canadians who are using them is clear and concise, that would be really helpful in terms of building that trust.
Senator Arnold: That’s a really interesting example. I think we have forgotten about it having been used in that way for a long time now, but what we see right now is that there is a lot of fear around it.
Is there anything that you learned through the experience that you went through as a kid using it? Were there any lessons learned through that which could be applied today and might be helpful to others?
Mr. Zangio: Yes, definitely. One major thing I would say is that when I was in the class where we used those tools, one major thing that was different in my class compared to others was that the soft skills we were taught were very different. I think that’s what needs to be changed now.
For example, because we had speech-to-text, so we didn’t write and sometimes didn’t come up with the writing by ourselves, they needed to teach us discernment and critical thinking. I feel that has had incredible effects on my career so far in terms of being a critical thinker. I can allude back to that time when I was in Grades 5 and 6 and learning how to use these tools that were supposed to assist me in my work, but the teachers were also teaching me the soft skills that were to go with those tools. This could be applied to a young Canadian or an old Canadian trying to reintegrate into the job market.
As you said, building that trust is important. Canadians are scared, and building those trust bridges, as you mentioned, is important, but it is also intergenerational — and I can give you a great example of that.
As I mentioned, I was educated and had a chance to go to university and have been in formal education spaces where I learned to use AI tools. Now I’m able to take those tools and bring them to members of my generation in my family who didn’t go to university. That creates trust, because they see somebody like them who has the awareness and the cultural sensitivity to train them and teach them how to use those tools safely.
I think that’s how I would answer that question. Thank you.
Senator Arnold: Mr. Chhabra, it’s great that there are all these different departments working on all of this. Will there be someone or some organization that will ultimately look at the outcomes and be responsible for ensuring that potential bad, unintended consequences don’t happen?
Mr. Chhabra: Thanks very much for the question.
I think there have been a series of different types of approaches taken internationally when we look at how governments are wrestling with AI risk. There are some jurisdictions that have chosen to take a sectoral approach — notably, the U.K. The U.S. would be another. There are others, as well, that have essentially looked at the series of risks and harms that are emerging and have determined, for example, that it is more appropriate for transportation departments and agencies to look at AI and autonomous vehicles, to have financial regulators look at AI and financial systems and to have health regulators look at AI and medical devices.
The EU was successful in promulgating the EU Artificial Intelligence Act. Canada attempted to have broad horizontal legislation in the form of Bill C-27. Part 3 of that was the artificial intelligence and data act, which, very similarly, took an approach to look at high-risk or high-impact systems as well as a general-purpose risk category for general-purpose AI systems.
I think we have seen over time that, for a number of regulators, given the breadth of AI and how technologically impactful it is across a range of demands, it can be a challenge to attempt to regulate that from a central standpoint. The current approach under the new government has been to motivate and engage regulators to address AI, whether that be through the Competition Bureau Canada looking at algorithmic pricing risk or whether that be through Health Canada looking at medical devices, or others.
I think part of the AI strategy work is going to determine whether there are some horizontal issues that need to be managed. One that Minister Solomon has been really clear about is that he sees an opportunity and a need to update Canada’s private sector privacy legislation — the Personal Information Protection and Electronic Documents Act, or PIPEDA — not just because of AI but because of some risks that AI technologies can present, including in terms of automated decision systems and, potentially, some considerations around “deepfakes” as well.
Alongside many other countries, the approach that Canada is taking right now is a bit of a hybrid, looking to ensure that sectoral regulators are aware and empowered with respect to the issues associated with artificial intelligence and then looking to see if there are gaps that need to be addressed in a horizontal manner.
Senator Robinson: My question is for Mr. Chhabra. You mentioned in your opening remarks that there is going to be disruption, but you mentioned that there are some sectors that won’t see, perhaps, negative disruption.
I was wondering if you could give us a bit of an idea of what sectors will see jobs augmented by AI. I’m hoping that your answer will include the word “agriculture” at least once.
Mr. Chhabra: Thanks very much for the question. I wish I could make such a great prognostication or promise on that front.
What I can tell you is that, as I referenced earlier, Statistics Canada is tracking this carefully. Statistics Canada is looking not only domestically but also internationally.
We have not seen really good, credible evidence of a specific direction that the job market is taking in relation to AI. There is a lot of anecdotal evidence. There is a lot of commentary.
One of the pieces that I can share with you is that a number of the companies that we speak to and a number of the events that we attend and listen in on, as well as the research papers out there, demonstrate that a significant amount of coding is now being done by AI tools and not by humans.
On the flip side, as you mentioned, where is there a demand for more work and more humans to be engaged? That is where humans are in the loop: to debug that code, to audit that code and to assess and evaluate the safety and robustness of that code.
While you might look at coding as a whole and say that we are going to have a lot less need for those who are computer programmers and coders, on the flip side, there does appear to be a concomitant rise in the need for those who can work with the AI to monitor it, assess it, troubleshoot it and debug what it has created from a coding perspective.
Security is another space where we are seeing an increase in demand for those who really understand how AI systems function and their dual-use nature, in terms of offensive cybercapabilities and requiring a significant amount of work to develop their defensive cybercapabilities.
At this point, we’re seeing that things are changing. Certainly, there are going to be impacts. There is likely a productivity uplift in a number of different domains. There is a scientific uplift in a number of different domains. There is a real benefit to adopting, developing and using these tools, but exactly which categories of occupations are going to be positively or negatively impacted still seems to be a bit of an open question.
I’m sorry I couldn’t offer anything on agriculture specifically.
Senator Robinson: I’ll dig a little deeper if you don’t mind.
I understand, within agriculture, particularly, the fact that we can process data with such an incredibly increased capacity is going to have significant implications for accelerated research and development of new varieties, new crop-protective products and things like that. Have you seen anything in that vein in your research?
Mr. Chhabra: Thank you again for the question. Certainly, we’re aware of numerous application layers for AI. You mentioned the key word, and that’s “data.” Being able to use the data and to have access to clean, reliable, well-structured data is definitely a jumping-off point for AI applications to have a significant impact. We’ve certainly understood that there are, in agriculture specifically, opportunities for better optimization around soil use, fertilization and other aspects of agriculture that are being investigated, explored and developed as commercial opportunities today with AI tools. We are certainly aware of and tracking those kinds of opportunities.
Senator Robinson: It’s incredible to see how much it accelerates because typically in agriculture, the biggest limitation has been time. In order to replicate a season, you need more years. In farming, you go to the Olympics every year, and a farmer who might crop for 40 years will get 40 Olympics in their lifetime.
I’ve seen situations where, on-farm, one researcher using AI has done the equivalent of one incredible researcher over a 40‑year career. Thinking about the learning we’ll have from that is incredibly exciting.
That’s a statement more than a question. I just wanted to get agriculture in there.
Senator Karetak-Lindell: This is an area that I have absolutely no expertise in whatsoever. I can barely turn on my laptop and do a Zoom meeting without difficulty. However, I’m very interested in health research and how this could improve our lack of access to health care where I’m from. Of 25 communities in Nunavut, most don’t have a doctor. We are seeing some success with distance health opportunities because every time we need to have specialized care, we otherwise have to go down south. There are some services available in Iqaluit right now, but they’re very limited, so most of us across Nunavut have to go south for specialized health care.
Do you have any idea if there would be opportunities to improve health care using AI? How would that improve access to health care as well as research? We have communities with TB still, and they’re trying to eliminate TB by 2030. Northern Quebec has a lot of cases right now. We’ve had communities in Nunavut that only recently got off the list of those with prevalent TB. I think there are two communities left. They’re trying to collect data as they take care of TB patients.
That has been a difficult area for us, with very limited capability to collect data from 25 health centres. Not only that, some of their health information is collected down south. It’s done in Ontario, Quebec, Manitoba and Alberta, so trying to get all that data together is also a challenge. To get a full picture of one patient, we have to go to so many different care facilities.
This is probably two questions now. I’m trying to see how you collect data, make it work and know for certain that it’s correct information for a patient.
Mr. Chhabra: Thank you so much for the question. It’s an important one that the government has paid a significant amount of attention to, including through funding for the Pan-Canadian AI Strategy for the three AI institutes: Amii in Edmonton, Vector in Toronto and Mila in Montreal. In the past year, I’ve met with researchers from Mila who are investigating and exploring this exact question of how to better leverage AI for health care. I think there are a number of issues here that are worth unpacking.
The first is the potential for AI to help with personalized medicine, which is an area of research that’s ongoing in Canada and elsewhere and that offers significant opportunities.
There’s an important point you raised as well about data and how best we can utilize the data we have across Canadian health care systems and through multiple users. One of the greatest challenges that researchers we’ve spoken to have faced is getting access to reliable data in a way that’s privacy protected.
Last year, the Government of Canada funded a project called VITALL, which is a really interesting use case around bringing together health data and doing it in a privacy-protected way — in this case, using federated learning, which is a privacy-enhancing technique. Again, utilizing AI can essentially allow researchers to query databases in a way that doesn’t require them to download sensitive personal information about patients. Instead, they can query the data in a way that helps to train the AI model without having to take that information in and adjust it themselves. You minimize the risk of inadvertently leaking, releasing or the wrong person getting access to that data by maintaining it in a secured facility.
That project was launched last year and is an excellent use case and example of how we can both support privacy and utilize data more effectively to help drive better health care outcomes.
Senator Karetak-Lindell: Mr. Zangio, you said that you learned about AI in Grade 5 and spoke of how that has helped you. How do you think that could be shared with other education systems? Should it be put right into the curriculum? What would be the way to ensure our young people are familiar with AI and how to work with it?
Mr. Zangio: That’s a very good question. Thank you so much.
One major thing that I would argue is that, with respect to the curricula, the provincial jurisdictions are doing a great job of incorporating these AI tools. As Mr. Chhabra said, we know it’s here, and it’s affecting the way youth are learning.
Something that I will stress and that I’ve seen is that young people are usually digital experts because they engage with technology on a daily basis. Those soft skills, the ability to understand and utilize those tools, are learned in everyday spaces. Traditional education in high school, middle school and elementary school is great, but it’s capitalizing on the lived experience outside of school that can really be a game-changer. That’s why youth are usually in tune with digital work. The society we live in right now is progressively shifting to a more digital one. We should keep engaging. Myself, I was engaged and kept engaging, and that’s how I am able to be here. Thank you.
The Chair: I would like to begin by asking about what’s being seen. Are you seeing a wide gap in terms of the generational difference on the take-up of AI? What’s the data saying about that? That’s my first question to you, Mr. Chhabra.
I’d also like to understand what the government’s role is. What is the government planning to do about public education in AI?
Mr. Chhabra: Thank you very much, Madam Chair, for the question.
I think we’re looking for more data on that exact question about the generational spread. I can tell you that last year, a number of folks who monitor this space with interest were quite interested to find that the utilization of ChatGPT in the United States dropped — I’m going to get this wrong, but I’m going to say it was something like 60% — the day after schools got out in the United States. That gives you a sense of how significant the use case was in academics.
I don’t, at this moment, have details on Canadian use by age, but I’m happy to come back to the committee with more details on that.
I’m sorry, but I forgot the second part of your question.
The Chair: It regarded the role of government in public education.
Mr. Chhabra: Yes. Thank you for the reminder.
It is definitely a piece of the puzzle that came up repeatedly in public submissions to the AI strategy consultation as well as in the expert reports. It was referenced several times as a key piece of the puzzle. Ensuring that AI provides opportunities for all Canadians is a real area of focus.
To do that, I think the government recognizes that there must be a degree of awareness building and skills development to support the appropriate engagement for responsible use.
I expect there will be more to say on the government’s approach to this once the AI strategy is released in the coming months.
The Chair: Thank you very much.
Mr. Zangio, I think I heard you say during your presentation that you had a family member affected by AI around this. We’re also hearing from the government that the data isn’t yet showing a serious impact in that area. From the material I’ve read, it seems as if clerical and similar roles are the most affected. Could you share with us what field your family member was in?
I think you have a very unique way of combining real experiences with the role that AI can play and how you translate that. I haven’t heard anyone else do that in terms of bringing lived experiences and being able to couple them with the importance of AI learning.
Can you talk more about that, as well?
Mr. Zangio: Thank you for the question.
I’ll start with something from Deloitte in 2026 that said that generative AI could automate nearly 40% of Canadian jobs. My family member specifically was working at a bread factory as a person who would take the bread and put it in packages to be shipped off. Those are the types of jobs that will prove difficult, because if they become automated, those Canadians don’t necessarily have the money to reintegrate into skilled jobs because they don’t have the mobility. They have debts, and the cost of living is high. Also, how do we reintegrate into a job market when there’s a job shortage?
Something that was brought up and corresponds to your second point about lived experience is that Deloitte also came out and said that 92% of AI dollars are spent on technology and only 7% on people. I think the key thing there is the 7% on people. That’s what I alluded to in my speech. When the great province of New Brunswick built that education system and created classes with certain students having AI tools, they centred the funding around people. As a member of this society, I, as a child from welfare, was able to go into the same class for two years and utilize AI tools to level the playing field.
When you’re asking that question — and I really like it — I feel as if there’s sometimes an imbalance between how much money we’re spending on technology and how much we’re spending on people. If AI is supposed to help people advance in life, we need to re-centre people in the conversation.
Thank you so much for that question.
The Chair: Thank you. I’m sure that part of what you said was music to Senator Arnold’s ears.
Senator K. Wells: I want to follow up on what Mr. Chhabra said around the $50 million that has been invested. Do you know — and if you don’t, could you find out for us — how much of that $50 million was targeted specifically at youth?
Mr. Chhabra: Thank you for the question.
I don’t have data on that in front of me, but I’m happy to engage colleagues at Employment and Social Development Canada to get more details on how that program is rolling out.
Senator K. Wells: That would be appreciated. Thank you.
My next question is for Mr. Zangio.
If you are aware, can you tell us about any of the research you might have uncovered that talks about the presence of bias or racial or gendered language or stereotyping that occurs in some of these existing AI technologies?
Mr. Zangio: That’s a very good question.
I’ll be sure to share my discoveries with the committee around that research, but one of the biggest things in terms of AI systems, as you mentioned, is racial bias. The greatest example is AI image generation and depictions: using lighter skin for higher-paying positions and darker skin for fast-food jobs or low-status roles. That is algorithm bias, and it is not theoretical. We’re observing it now.
The second part is our representation piece — and that’s where the bias is so strong: There is a representation gap in Canada’s tech workforce. This is from Deloitte also: Black Canadians and newcomers are only 2.6% of the Canadian tech workforce, despite being 4.3% of the population.
So it’s a great question, and racial bias, gender bias and misrepresentation are key to that. When we have representation, bias can be mitigated — but people are able to generate these data sets, for example. I would love to share more research with the committee on that question.
Senator K. Wells: Thank you for sharing your experience. A concern that’s top of mind is how AI can continue to perpetuate and amplify biases that are inherent in the large language models that they might be drawing from.
Part of the question, and what we’re trying to figure out, is this: What are the human rights implications when you take that bias and scale it up and amplify it? It seems like a natural and normally occurring phenomenon when, in fact, it’s based on structural, systemic discrimination. We don’t want these powerful tools to continue to marginalize and perpetuate that systemic, structural discrimination that we are trying to root out of our systems through great legislation and policies — such as the importance of equity, diversity and inclusion.
We also see tremendous potential here to point out where the gaps are and address them. We’re trying to find balance here: How do we harness the benefits of AI? At the same time, how do we ensure we’re not replicating existing disadvantages and equity discrimination that are, quite frankly, the antithesis of Canadian values?
Thank you for sharing these perspectives. From where I sit, the answers we’ve seen from all of our panellists have been helpful and informative.
Senator Arnold: I don’t want Ms. Burke to sit here without any questions.
I’m looking at your title — and maybe this isn’t fair — but something I’m grappling with is the environmental cost of AI.
I have a little anecdote: The other day on social media, I saw that everyone was making AI generations of what they looked like. They all looked a lot more youthful and beautiful. Someone put up a post saying, “This is such a waste of resources.” Does anyone know how much it costs for these people to share these memes? I don’t feel as if anyone is having that conversation right now. I don’t know if that’s in your job title; it sounds like it could be.
Are we talking about that? What are the costs of this tool we now all have at our fingertips?
Katherine Burke, Director General, Foreign Investment Review and Economic Security Branch, Innovation, Science and Economic Development Canada: Thank you for the question. I appreciate it, and I appreciate the perspectives you are bringing in terms of the environmental and pure economic costs. I will answer part of the question, and then I will turn to my colleague, yet again, who will be able to speak about it in more detail.
My role focuses on economic security. The way we understand economic security is at a national level, meaning that my role is to explore the ways that the Canadian open-market economy can be made vulnerable to external threats, predominantly ones that may undermine our national security or our national sovereignty.
When our department or the Canadian government talk about economic security, we’re thinking about it from the perspective of Canada as a sovereign nation and the potential to be undermined by new, developing technologies that can be harnessed by adversaries. It’s a slightly different concept of economic security than I think is most commonly discussed at this committee, but I’m always happy to answer questions as they come up.
In terms of your specific question as to who is thinking about this or what those questions might be, I’m afraid that, once again, I will have to turn to my colleague Samir to answer that question in more detail.
Mr. Chhabra: Thank you very much for the question.
Indeed, the environmental impacts, generally focusing on energy utilization and water utilization, are significant areas of research and work, both from a commercial perspective — because, obviously, reducing the amount of energy used and water utilized is in the commercial interest of those building these AI systems and data centres to power them — as well as those engaging it from a government or policy perspective.
We’ve taken significant steps to ensure that the data centres being funded by the Government of Canada are being approached in a manner that is at the cutting edge of the best approaches in terms of energy management and water use. Certainly, we’ve seen some provincial jurisdictions take steps to focus on whether a data centre is in the interests of the province and its people in terms of how much energy it’s going to use and whether their grids can manage that offtake.
We have a lot of innovation in Canada around this space. Organizations such as Mila, which I referenced earlier, one of our AI institutes, are pioneering work in the efficiency of the compute work. It’s actually drawing less energy for a given amount of output that’s required. This is an important area of work from a policy and technological perspective to continue advancing.
To answer your question, certainly it’s a big part of the conversation. It’s a part of almost every conversation we have domestically and internationally about AI. The underlying hypothetical is AI use is likely to continue to grow. Exactly to what extent is more difficult to predict at this point, but certainly the demand for building data centres and creating these offtake agreements is continuing to increase. If that’s the case, we need to think carefully about whether we have the resources and grid infrastructure available to support the work that’s coming.
Senator Arnold: As a former elected politician, it’s hard to take things away from people. If people are now using it so freely all the time without any thought to the cost of it, are we doing a disservice to the future of this technology?
Mr. Chhabra: That’s an interesting question. I don’t know if I can answer it directly by saying it’s a disservice. The technology is still relatively nascent, notwithstanding the fact that it’s broadly utilized. I think it’s important to make a distinction between the utilization of that technology in a series of different verticals where the use case for science, for productivity enhancement, for precision agriculture — as the senator from P.E.I. highlighted — are very significantly beneficial in health and in protein folding. There are real impacts that are worth pursuing.
I think when people talk about AI today, they generally think about these consumer-facing chatbots and the ability to create memes or to engage with Gemini, ChatGPT or another product to get answers or, as I talked about earlier, on a more interpersonal basis.
I think we’re seeing different utilization of AI. As we go forward, we need to be thoughtful about the utilization of AI. What is the application? When is it worthwhile? Who’s paying for it? How is it being funded? These types of questions will continue to be explored and deepened over time. Again, we’re on the cusp of a lot of engagement around this issue.
Senator Arnold: I agree. I think right now might be the time to have some of those conversations.
Senator Robinson: Mr. Chhabra, I want to get your perspective on how your research, how you’re looking at AI, takes into consideration how it will impact rural Canada. Is there a bias here, where you’re looking more toward urban Canada? As we heard, in the North, we have health care concerns. Also, one of our largest economic drivers is agriculture, which is predominantly in rural Canada. Do you feel we should have any concerns that there might be a bias here and that you’re not looking at rural Canada enough?
Mr. Chhabra: Thanks very much for the question.
First, the government has been engaged over a number of years with rural connectivity and broadband work, which is one of the key ingredients here for consideration: Can people access the tools?
Second, I’ll highlight that Budget 2024 provided significant funding for the regional development agencies to ensure that AI programs and supports were available across Canada, including rural areas. This is a piece of the puzzle that has been important to the government for quite some time.
Last, I would say that, under the upcoming AI strategy that the government is working to develop and bring forward, this consideration of AI opportunities for all is a clear underpinning of that work to ensure that no one is left behind or left without opportunities. That includes things such as skills and access to compute resources, which is another piece.
Senator Robinson: I’d like to take a second and look at the broadband issue and cellular coverage because in rural Canada, that is a massive limitation with respect to precision agriculture. You go down through a valley, and you have no coverage. All the science you can otherwise access to ensure you’re doing it as efficiently as possible is not there when you don’t have a connection.
You say that you’ve looked into that. It’s highlighted and you’re aware of it, but can you speak to action and deliverables on that?
Mr. Chhabra: Thanks for the question. If you’re asking specifically about cellular coverage in rural Canada, I’d have to come back to the committee with a more detailed answer. It’s not my space.
Senator Robinson: I look forward to that. Thank you.
The Chair: Ms. Burke, I’m bringing you in as well. I’m curious about your work. You mentioned the important word “sovereignty.” I’m curious about how that intersects with AI, understanding there’s no border when we’re talking about AI.
I had the opportunity to visit all three institutes across the country last year. Something that was discussed was where we are when it comes to policy guardrails and how we compare to other regions in the world, particularly our neighbour to the south. I heard Mr. Chhabra talk about a horizontal approach. I’m not sure what that means. Could you elaborate?
Ms. Burke: Thank you for the question, Madam Chair. I will absolutely answer and then turn back to my colleague to provide further clarification.
Your question, as I understand it, is about the relationship between AI as an evolving technology and our economic security. We use the phrase “economic security” around the concept of the maintenance and expansion or the resilience of Canada’s sovereign capacity to make decisions, both for our people and for our businesses or our industry. That’s the lens that we are bringing from my sector.
The impact that AI can have from that lens is both as a vulnerability as well as an incredibly powerful tool, potentially. We heard Mr. Chhabra talk today about the number of different ways that different sectors of our economy and our capacity and productivity can be enhanced as a result of AI. Those are ways that we can reinforce our economic security.
For example, if we look at health care, the ability of our country to domestically support necessary public health outcomes is a way that we can reinforce our sovereign capacity to support Canadians and also support our economy. The way that we can use AI in a variety of dimensions, from things such as personalized health care or telehealth and distance surgery — new uses of the technology such as these can help reinforce our economic security and, by extension, our national security and sovereignty.
That is the potential benefit of AI from this perspective, but, again, any cutting-edge technology creates potential vulnerabilities when we look at reliance. For example, the more we rely on digital technology, the more that reliance can potentially create a vulnerability that could be exploited by potential adversaries. We heard today at committee about that potential reliance, where, if you don’t have cellular access, you don’t have access to the connectivity or the communications you might be relying on. In that way, AI is definitely a double-edged sword when we think about our economic security.
How can we ensure that we can reap the benefits without having those vulnerabilities? Through things like this new AI strategy — which is in development and will be released shortly — and by having an integrated perspective on the ways these kinds of technologies can benefit our market and our prosperity, as well as by ensuring the government has the tools to protect against vulnerabilities as they’re evolving.
The Chair: We have run out of time, and I have to give the last two minutes to Senator Wells.
Senator K. Wells: My question is for Ms. Burke. I don’t know if this is a two-minute question, but do your best.
Taking the big picture on some of the work that you’re doing, what would you say are the most pressing economic security threats or concerns presented by AI to Canada?
Ms. Burke: Thank you for the question, and I will try to answer it in a brisk two minutes or less.
I would say that the biggest risks or potential vulnerabilities that AI can present to Canada’s economic security include, as I indicated, the potentially increasing reliance we may have on the way we use AI across different sectors. This means, as we expand the way we use AI or integrate AI in all its forms — both consumer-facing and industrial uses — ensuring that we are aware of how we are reliant on it and ensuring resilience.
Resilience can be something that is led at the federal level. It can be led at the provincial level. It is also an area where individual companies have an obligation to recognize their vulnerabilities, identify potential vulnerabilities and develop plans around them. That is definitely something that the federal government has a responsibility to support.
Again, I would turn to the upcoming AI strategy as a space where many of those conversations are happening. However, it is also, as with any evolving technology, an evolving response.
The Chair: That is the end of our time with you. I would like to sincerely thank you for agreeing to participate in this meeting. Your assistance, your expertise and your perspectives have all been very helpful.
As I said at the top of the meeting, we have been very excited about beginning this study. We look forward to receiving any additional information that you may wish to share that you didn’t get a chance to. Thank you for being here.
I will now introduce our second panel. Our witnesses have each been asked to make an opening statement of five minutes. This will be followed by questions from senators.
With us at the table, from ControlAI, we welcome Samuel Buteau, Consulting Program Officer. With us by teleconference is David Scott Krueger, Assistant Professor, Department of Computer Science and Operations Research, Université de Montréal. Also by video conference, we welcome Malo Bourgon, Chief Executive Officer, Machine Intelligence Research Institute.
I now invite Mr. Buteau to make his presentation, followed by Professor Krueger and Mr. Bourgon.
Samuel Buteau, Consulting Program Officer, ControlAI: Thank you, Madam Chair and members of the committee, for inviting me to testify today. My name is Samuel Buteau. I have worked on the technical side of AI for over a decade, and I am now the Consulting Program Officer in Canada for the non-profit organization ControlAI.
In the current study, this committee is investigating the impact of AI on human rights. Arguably, the most important human right is the right to life. With respect to this, I must warn the honourable senators that AI companies are currently gambling with the life of every human being on the planet.
It is the explicit goal of the top AI companies to build superintelligent AI. That means AI that can fully replace humans in performing any task. This includes both individual humans and groups of humans, like companies or governments.
Yet, in 2023, Nobel Prize winners, leading AI scientists and the CEOs of these same AI companies warned that superintelligent AI poses an extinction risk to humanity. If research and development is allowed to continue at the same pace, many of these experts expect superintelligent AI to be built within the next decade.
I believe many applications of AI are beneficial, but if we build machines more competent than us in every domain, we will be rolling the dice. No country, company or individual currently knows how to control AI that could outsmart, outcompete and generally run circles around them.
However, it is possible to prevent the development of this technology in the first place. A race to superintelligent AI is a race where everyone loses. Governments have no interest in letting AI companies continue pursuing it. Once they understand, countries will be motivated to ensure this technology is not built anywhere in the world until it would be safe to do so.
A few key countries, working together, could get this done, not out of naïveté but through a regime of mutual monitoring and verification. Right now, pursuing the frontier of general-purpose AI requires very large supercomputers. They use hundreds of thousands of specialized chips, consume as much electricity as a small city and are visible from space.
In short, restricting the frontier of AI research is feasible, and, with proper monitoring, it can be assured that no one builds superintelligent AI in secret.
We are currently gathering the political will for this task. Less than a year and a half ago, ControlAI started sharing these facts with decision makers and asking them to take action. During this time, we gathered more than 100 public supporters within the U.K. Parliament. Today alone, I’ve met with four Canadian MPs to brief them on the risks of superintelligent AI.
I put forth three recommendations: First, the Canadian government should publicly recognize superintelligent AI as a national and global security threat. Second, Canada should take the lead in uniting a coalition of countries focused on preventing the development of superintelligent AI through an international ban. This ban could be enforced via a mutual monitoring and verification regime. Third, Canada should protect its citizens at home and lead by example abroad by prohibiting the development of superintelligent AI within its jurisdiction.
Thank you. I look forward to your questions.
The Chair: Thank you, Mr. Buteau.
David Scott Krueger, Assistant Professor, Department of Computer Science and Operations Research, Université de Montréal, as an individual: Hello. Thanks for inviting me. My name is David Krueger, and I am a machine learning professor at the Université de Montréal and Mila.
In 2012, I learned about deep learning from Geoff Hinton’s online lectures, and I realized this new approach to AI might produce superintelligent AI within a few decades. So, I went to study under Yoshua Bengio in Montreal. At the time, I was already concerned that superintelligent AI could cause human extinction and wanted to know what the experts thought.
I hoped to find that they had good reasons not to be concerned, but what I found was that nobody was really thinking about it at the time. In fact, for most of my time in the field, the risk of human extinction from AI was considered a taboo topic and researchers feared for their career if they talked about it. This, unfortunately, set back critical public conversations about how to handle this risk by years.
I’ve been talking about this at every chance I get for over a decade, and I continue to be dismayed at the bad arguments people make to avoid confronting the problem. On the other hand, over time, I’ve seen more and more researchers become increasingly concerned.
In 2023, we had a watershed moment, and I initiated a statement:
Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war.
This was signed by many of the biggest names in AI, including Hinton, Bengio and hundreds of other AI researchers. Unfortunately, the end of 2022 was when ChatGPT was released, and AI companies started to become extremely powerful in 2023.
The most important thing I want to stress today is that we’re still in a position where the world is not yet taking the steps needed to mitigate the risk of AI leading to human extinction. So, what is this risk, and where is it coming from?
AI companies are explicitly trying to build superintelligent AI systems. These are systems that are much smarter than people across the board and can autonomously do anything that humans can do — including robotics — and do it better, cheaper, faster, et cetera. The basic goal is to render humans obsolete and take all their jobs.
But we don’t know how to control superintelligent AI. In fact, we don’t understand how existing AI systems work because they are grown, not built, using deep learning. Despite thousands of research papers over the past decade, this remains an unsolved research challenge. We should not expect any amount of investment to solve this problem in the foreseeable future.
We also don’t know how to do safety testing for these kinds of AI systems. The kinds of tests we have can show that an AI system is dangerous, but they cannot show that it is safe. We should also not expect any amount of investment to solve this problem in the foreseeable future.
So, instead of maintaining control or using rigorous safety practices, AI companies and researchers try to instill particular goals and values in AI systems so that the systems will do what they want. However, we only know how to do this approximately, and even a small approximation error might lead a superintelligent AI to reallocate resources we humans need to survive toward its own misaligned goals.
So, none of these approaches that you will hear mentioned — interpretability, testing or alignment — is adequate. We don’t know how to build superintelligent AI safely; the plan is to roll the dice.
Finally, if the companies building superintelligent AI do not immediately lose control of it, we should still expect the wholesale replacement of humans with AI throughout society. It is not just near total unemployment but also political power being handed over to AI systems that make decisions too quickly for humans to meaningfully participate in.
I want to say something about timelines, because I think we are now in a state of acute crisis. If we don’t do anything, I think we’re about five years away from superintelligent AI, so we need to course correct immediately and work to prevent its development. We need to do that internationally, which will take time.
We cannot afford to wait for more evidence that the risk is imminent; there is already ample evidence that the level of risk from this course is unacceptable.
In my home country of the United States, the main argument against stopping the race to build superintelligent AI these days is simply that it’s inevitable: If we don’t do it, China will.
This is false. One simple way to stop it would be to get rid of advanced AI computer chips and the factories that make them. Fortunately, the supply chain for these chips is extremely concentrated, making this or other interventions to control and limit the means for producing superintelligence possible.
There may be better ways that are less costly, but this cost is also well worth paying, if necessary, given the risk.
In conclusion, AI is an unprecedented technology, and the future of humanity is at stake. We are in a state of crisis, and we need immediate action to slow or pause AI development internationally. This issue should be the number one foreign policy priority of every nation, including Canada.
Thank you.
The Chair: Thank you, Mr. Krueger. We’ll now go to our next speaker for their presentation.
Malo Bourgon, Chief Executive Officer, Machine Intelligence Research Institute: Madam Chair, members of the committee, I’ll serve my statement as a reinforcement of those made by my fellow witnesses, who have covered many of the same points. Regardless, my name is Malo Bourgon. I’m CEO of the Machine Intelligence Research Institute, or MIRI, a non‑profit focused on ensuring that the development of powerful AI goes well for humanity.
I grew up in Ontario and studied engineering and computer science at the University of Guelph. During my 14 years at MIRI, I’ve seen our work help spark a new research field aimed at understanding how to build AI systems that behave as intended, even as these systems become much more capable. The leading AI companies today say their goal is to create artificial superintelligence: AI systems that are smarter than any human with respect to every cognitive task. OpenAI says this explicitly. Anthropic’s CEO, Dario Amodei, talks of building a “country of geniuses in a datacenter.”
These companies weren’t founded to make chatbots; to them, chatbots are just a stepping stone.
Unfortunately, the field dedicated to keeping such systems under control has made little progress. Even OpenAI’s former chief scientist, Ilya Sutskever, has said that “. . . we don’t have a solution for steering or controlling a potentially superintelligent AI, and preventing it from going rogue.” As a result, researchers at MIRI are concerned that if the world continues racing toward superintelligence using anything like today’s techniques and understanding, the default outcome is that we will lose control, likely resulting in human extinction.
Why is there so much danger here? A key insight is that AI is unlike traditional software. Traditional software is written line by line; in principle, a programmer can understand every part of it. As David said, modern AI systems are grown rather than designed by hand, using enormous amounts of automated trial and error. The creators of those systems have little insight into what’s happening inside them, and, as a consequence, these systems often exhibit behaviours nobody asked for or wanted.
Frontier AI models get caught cheating on evaluations, engaging in deceptive and manipulative behaviour and even driving users into a state clinicians are calling “AI-induced psychosis,” even when the AI can readily tell that its responses are harming the user. Furthermore, recent research shows that these models increasingly recognize when they’re being evaluated and adjust their behaviour accordingly, undermining the validity of safety evaluations.
At current capability levels, these behaviours are concerning but not catastrophic. But we must ask these questions: What happens when they reach the capabilities these companies are aiming for? What goals will future AIs exhibit? Do these AIs endanger us, as I say? Many who have studied these questions have found the answers quite concerning.
Two of the three godfathers of deep learning, Canadians Geoffrey Hinton and Yoshua Bengio, have publicly warned of the dangers of extinction. My fellow witnesses, David and Samuel, mentioned that, in 2023, they joined other top AI scientists, and even the CEOs of OpenAI, Google DeepMind and Anthropic, in stating:
Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war.
Some of these signatories lead the very companies racing fastest to build superintelligence, and increasingly, they’re saying they’d like to stop. Google DeepMind’s CEO said he’d support a pause if everyone else followed suit. Elon Musk says he has “. . . a lot of AI nightmares. . . .” and would slow down if he could. At Davos last month, Anthropic’s CEO said he wished they could slow down but can’t because of geopolitical rivals. These leaders say they feel trapped in a race they know is dangerous, but no single company can stop on its own.
By their own admission, we cannot rely on them to develop this technology safely. Reining in this race will require international coordination between allies and adversaries alike, something that only governments can do.
Catastrophe is not inevitable. These dangers can be averted. This reckless race so many see as unstoppable is taking place in a world where most people don’t understand the threat. That can change.
As a respected voice on the global stage, Canada can help push for the international coordination and cooperation needed to rein in this reckless race. Canadian policy-makers can say what others seem to lack the courage to, that according to top experts, the race to superintelligence is far too dangerous to continue.
And this is feasible. The race to superintelligence runs on extremely specialized AI chips, designed and manufactured by a handful of companies using machines made by one Dutch company; each machine costs as much as a 747. These chips are housed in enormous data centres that consume enormous amounts of electricity. The infrastructure for AI and superintelligence development is conspicuous and can be monitored and regulated without giving up the AI tools that, if developed and deployed wisely, could provide enormous benefits to society. The only thing that needs to stop is the race to build superintelligence.
Canadian scientists led the way on this technology and continue to lead through their efforts to avert these dangers. My hope is that Canada will use its voice and moral authority to push for a world where getting lucky is not our main plan for confronting this pressing threat.
Thank you.
The Chair: Thank you all for your presentations.
Senator K. Wells: You have all presented very sobering perspectives. There is a lot of heavy thought and worry around what you have raised.
Mr. Krueger, I grew up in the nuclear age of the 1980s, when we were thinking of the Doomsday Clock being about one minute to midnight with the Cuban missile crisis — the existential threat that turned into a very real threat that dominated so much of our conversation and landscape.
Based on what you have said, if there were an AI Doomsday Clock, how close do you feel that we would be to midnight? Are we days, hours, minutes or seconds away from what you described as human extinction?
Mr. Krueger: Thank you for the question. I first want to emphasize that the future is uncertain. I don’t claim to know exactly when things will get out of hand, but I do have my own perspective on this and can share a little about what other experts think.
The most obvious thing that we have talked about a number of times as sort of a point of no return is the development of artificial superintelligence. I mentioned that I think that’s about five years away. A lot of experts, especially people working at the companies developing it, have similar timelines for the development of superintelligence; some are even shorter.
I think it’s important to recognize that’s the point at which there is no going back. If we want to stop this race, we need to do so well before superintelligence is possible.
Building superintelligence is getting easier all the time as we make further investment in scaling up AI and further algorithmic progress. If we attempt to pause right now, some of that would continue. If we say that we will ensure that nobody is using any of the chips that we know about to try to build superintelligence, there will still be a decent number of AI chips that we don’t know about. We don’t know where they are. They could be used by a government in secret to try to continue this kind of dangerous research.
The time to act is now. We should have acted before this. We will not have an opportunity to act effectively, I believe, if and when we see clearer signs of danger. We have seen many signs of danger, like the ones that Malo mentioned. We have seen the rate at which AI capabilities are advancing, which is extremely rapid. We have seen experts tell us this is a concern, and that began about three years ago now.
I was hoping that, in 2023, after we put out this statement, governments would leap into action and start addressing this problem with the gravity that it deserves. Unfortunately, that didn’t happen. We still have a chance here, but I think we have maybe only a couple of years to not start discussing a ban on further progress to superintelligence but start implementing one. International discussions are long overdue. That’s my basic perspective.
Senator K. Wells: When we look back through history, we see some major turning points in global society. We can think of the printing press and the Industrial Revolution as fundamentally changing society, and now we’re in this new AI revolution that is before us.
If you could make recommendations to the Canadian government around providing safeguards for Canadians right now, what would be the top three recommendations? We’ll start here in the room.
Mr. Buteau: I mentioned some recommendations in my opening statement. I believe Canada’s main lever is to solve the coordination problem around having most governments in the world take this issue seriously and realize that superintelligent AI would be a national security threat to them — to everyone — and begin negotiations.
Even a few members of Parliament or senators taking a stance on this would make it much easier for others in the government to do the same and to look into it. Obviously, if the entire government of Canada were championing this issue, many allied governments would find it much easier to look into this and to take this seriously.
We have to realize that we are currently in a world where the vast majority of policy-makers, at least the ones I have spoken with, have barely heard of the issue, or are not aware that this is the kind of AI that companies are aiming to build and that these same people claim that nobody knows how to control it.
We cannot despair before we have tried to get people to realize the problem. I think once they do, the solutions will be obvious. This is the main strategy I would recommend for Canada.
Senator Arnold: Mr. Buteau, we met, and you described to me the difference between regular AI and superintelligent AI. I would like it on the record because I think it’s confusing for a lot of people.
There is some good in all this. If you could just lay that out for us, I think that would be helpful.
Mr. Buteau: As I said in my opening statement, I believe many applications of AI are quite beneficial. I think there are two distinctions to make: between specialized AI — or tool AI — and general-purpose AI. Specialized AI can only do one thing, and it remains in human hands. For example, an AI trained only on medical images to detect cancer or trained on protein data to discover how proteins fold, these systems are narrow, predictable and worth pursuing. They would remain a tool. As with any powerful technology, tools can have beneficial and dangerous uses, but they wouldn’t be a danger to their operators.
We should distinguish this from general-purpose AI, which is the paradigm that companies are currently using to build chatbots and automated coding assistants. They aim to push this toward superintelligent AI.
The difference between general-purpose AI and superintelligent AI is just a matter of degree. How much is it able to replace humans in a variety of tasks that humans take a long time to perform and that require autonomy?
General-purpose AI is quite different in the sense that it’s not just trained for one thing. We spend enormous resources — such as the data of the whole internet and many other things — and create systems that we fundamentally do not understand. The CEO of Anthropic, Dario Amodei, which is currently the leading company, recently said that perhaps we understand only 3% of how our systems work. I believe he was sort of highlighting how Anthropic is ahead of the competition there, but since we don’t understand them and nobody has a plan to control them, and given the fact that right now they have behaviours and quirks that are kind of concerning, it is concerning. But the question is this: What happens when this general-purpose AI technology goes beyond the abilities of any individual human or any group of humans, such as the government, et cetera? What then?
That’s the main distinction.
Senator Arnold: Professor Krueger, you’ve laid out a pretty scary situation here. I heard you say that it’s the chips. If we could control the chips, then some of this would slow down or stop. You’re saying that the government needs to be involved.
All the businesses that got together and wrote the letter — I guess I’m not really understanding why, if they understand what some of these unintended consequences could be, private companies don’t do the right thing and stop it.
Mr. Krueger: Thanks for asking.
I think there are a few different things going on here. One is that a lot of people view this as inevitable and haven’t thought seriously about whether there’s a way to stop it. There are reasons why it’s difficult to stop, because it does need to be international. A lot of people think the prospect of the U.S. and China cooperating on this is too remote. On the other hand, I think if these governments are aware of the stakes, it’s certainly in their interests to find a way to cooperate.
The other question is this: Besides this political question of whether they’d be willing to try to broker such an agreement, would such an agreement be technically monitorable and enforceable? That’s where it’s very important to understand the underlying physical reality of how these systems are built and the chips.
The supply chain for the chips is very concentrated. You have ASML in the Netherlands and Taiwan Semiconductor Manufacturing Company, or TSMC, in Taiwan, which are both critical to building the most advanced chips. At the moment, these are massive projects with billions of dollars of investment behind them.
It would be very easy to stop the biggest examples of this, but you would have to then wonder whether governments would keep trying to do this in secret. That’s when you have to look seriously at getting rid of a lot of the chips and the ability to manufacture them so that governments know they don’t have the means to do this in secret.
Coming back to why people would still do this, there’s also the promise of great power from this technology. If you manage to build it and control it — which some people are much more optimistic about than I am — then you could, essentially, take over the world and shape it in whatever way you choose. Some people are motivated by that. Some people are simply motivated by the massive profits that they imagine. Some people in the field of AI and at these companies are motivated by the prospect of replacing humanity with a superior advanced species of AI or cyborgs. That’s their vision for the future that they want to bring about and realize.
So, there are a number of reasons why people are continuing to do this, but I think the most important one to address is the international coordination counterargument — the argument that if we don’t do it, China will, as it is often said in the United States, as mentioned.
I think that’s the only one that gives any justification that the public would accept. They won’t accept, “We want a bunch of power,” or, “We want to replace humanity.” Those are not acceptable justifications for this kind of activity.
The only one that I think is plausibly a justification is that it is going to happen and we may as well be the ones to do it. That is why I think it is very important to point out that we can stop it. It’s a political problem, not a technical one.
Senator Robinson: My questions have kind of been answered a little bit.
Mr. Buteau, you were answering a question that Senator Wells had posed on what steps to take — how do we, as Canadian senators, contribute to trying to solve this problem? We ran out of time, and I was wondering if we could go to our online witnesses for their opinions.
Maybe we’ll start with Mr. Krueger.
Mr. Krueger: Thanks. I agree with everything that Samuel said. I think the overarching priority here is to begin the process of brokering an international agreement to prevent further progress toward developing superintelligence, which I believe needs to include major interventions into and oversight of the compute supply chain. Probably we want to simply shut down the operations of the companies that are building the advanced chips and the machines to make them. That’s the most robust and simple solution.
Now, within Canada, there is one thing I want to add, which is that every time a new AI capability comes out, there are people who immediately start using it to do things that I think everyone in this room and 99% of people would recognize as intensely antisocial. Examples include instructing the AI to destroy humanity, as in the case of ChaosGPT in 2023, or instructing the AI to basically try to become autonomous and start going out to get resources it needs to survive and pursue its own objectives.
This is not the main point to emphasize, but, obviously, these sorts of things should be made illegal immediately everywhere.
Senator Robinson: Thank you.
Mr. Bourgon?
Mr. Bourgon: David and Samuel covered it fairly well. Maybe I could add things that won’t solve the problem but might be some practical first steps that might help along the margins.
I heard the representative from the government talking about the Canadian Artificial Intelligence Safety Institute. Certainly, if Canada is going to grapple with this challenge and try to be a leader here, I think its AI safety institute should be tackling this problem. Unless something has changed very recently, it’s my understanding that looking at the catastrophic risks from AI is not even in the remit of the Canadian Artificial Intelligence Safety Institute.
I think Canada would probably be more equipped to think about this problem if the place trying to help it understand what is at stake with AI and how to navigate AI development safely actually had in its remit working on and thinking about the biggest, most catastrophic risks.
Senator Robinson: Thank you.
Senator McCallum: Thank you all for your presentations.
I want to look at human rights and the difficulty that we now have as human beings in correcting injustices for vulnerable populations. We haven’t been able to solve them over many generations. When I look at AI, my concern increases even more. When we look at AI and say, “It’s going to serve the public interest,” when it does that, we need a deep understanding of the human rights impacts to allow for restrictions and prohibitions.
One of the presenters said the machine is being groomed, so it has biases and so on built in. It’s going to intensify human injustices.
To what extent does AI threaten to erode ethics, critical thinking and interdisciplinary skills and traditions when you look at deep learning? In First Nations and in our teachings, we have deep learning to generate wisdom, but AI cannot generate wisdom — or is it capable of doing that to make ethical decisions? How does AI threaten human rights? What do you think is the solution?
Mr. Buteau: I can begin. One core technical aspect of this is that current AIs are not programmed line by line, and so it’s very difficult — we don’t have the technical means currently — to make AI want to respect human rights. The strategy of building something much more powerful than us, to which we would be kind of powerless, then prevents us from pulling back if the superintelligent AI we build does not respect our human rights or our way of life the way we’d want it to.
I think everybody can look at history and see the parallels of being at the mercy of something much more powerful than ourselves, and this is not a great situation to be in. The human story is not perfect, but it’s on a trajectory, perhaps, where we still have the means and the will to make a more just and beautiful world in the future. I think superhuman AI would be the end of that human story.
To the extent that we care about human rights — and we should be right to care about them — we should want to make a human future where humans are still able to make such decisions.
Senator McCallum: Would anyone else like to make comments?
Mr. Krueger: Sure. I’ll echo again what Samuel said. I agree with that.
You talked about historical injustices and the position of vulnerable populations. To the extent that AI doesn’t wipe everyone out, I do think it is going to empower people who currently have power to a dramatic extent and at the expense of those who don’t. I do think, ultimately, it will disempower all of humanity, though, because it will be smarter, more capable and more economically productive and fit, and so we will all be in the position of a vulnerable population compared to the AI. So the question of how we have treated disadvantaged and vulnerable groups in the past is very relevant because it hasn’t been good in a lot of cases.
A lot of people expect that AI is going to, despite taking control of everything, have a benevolent attitude toward us for some reason. I just don’t think that’s what we should expect based on these historical analogies and comparisons.
Also, it’s important to recognize that the race to develop AI is really kind of threatening all our existing social structures and social systems, including human rights and governments’ ability to effectively govern and protect their citizens. Besides the loss of control, which we’ve talked about a lot, there’s the technical reality that we don’t understand how AI works. But it’s very powerful, and that means that it will, by default, be used in many ways without any clear accountability and without any recourse for people who are harmed by it.
The Chair: Before we get to second round, I have a question I would like to clarify.
Have either of you been part of the consultation with government for the development of the strategy that’s coming in this first quarter?
Mr. Buteau: I believe that I helped another person write something for that, but I’m not sure if they ended up submitting it.
Mr. Bourgon: I have not.
Mr. Krueger: I have not.
The Chair: All right. Thank you. That answers my question.
Senator K. Wells: I’d like to open the floor and return to the question of the top three recommendations that the government should be acting on now, if there’s anything further that our online panellists would like to share. If not, I can move to another question. It sounds like we’ve covered those bases.
Turning on the news this past week, we heard pretty shocking examples of OpenAI not sharing data with law enforcement regarding the Tumbler Ridge tragedy. I’m wondering if you have thoughts on that. Obviously, we have the case of a troubled young person with mental health issues sharing disturbing information with an AI platform, which is then not shared with folks who could get this person help. We don’t know for certain, but we’ll say this was maybe a contributing factor in the tragedy that happened. It could have perhaps been a preventable tragedy.
What are your thoughts on mandatory reporting in instances like that? Is that going to help control some of these AI companies in terms of forcing them to disclose — whether it’s to law enforcement or to government officials — when troubling information comes across their platform that could result in harm to self or harm to others?
We know, for example, that teachers and counselling professionals all have a mandatory duty to report, where you violate confidentiality if a young person is about to engage in harm to self or harm to others. There is no confidentiality, and there is supposed to be forced disclosure in those cases because life is at risk.
I’m just wondering your thoughts on interventions and requirements for these platforms to do no harm, as we’re talking about — further harm — or to mitigate against harm. In the case of OpenAI, it was reported that it wasn’t just the AI technology; there were a multitude of humans that reviewed that data and decided it wasn’t at the threshold to share. We don’t even know what the threshold is for that particular company. This is a real‑world example that resulted in a tragedy in Canada, with countless lives lost and communities devastated — and that perhaps could have been prevented with the right policies or tools in place.
Mr. Bourgon: I can take a first pass at that.
I think it’s kind of remarkable how few requirements we place on AI companies to do any sort of reporting today.
Some of the things you mentioned seem as if they would be basic first starts, but when these companies are talking about the types of systems they’re building and the risks they’ll pose, including the large-scale ones, I think it’s incumbent upon governments to, at a minimum, require them to disclose progress on how their models are progressing as well as the internal safety regulations they’re using and what those are showing.
There’s a certain amount of situational awareness that is helpful in understanding how the technology is progressing, and major governments of the world might be as in the dark as I am. Maybe I’m less in the dark because I know some of the employees of the companies and sometimes get a bit of gossip. That seems crazy to me.
If Microsoft were to have a nuclear weapons program, and they said, “We’re not sure. Some of them might blow up once in a while,” governments might want to know more about what’s going on and consider hitting pause. Someone else in this space has a line I quite like, saying, “There are more regulations on sandwich shops than there are on AI companies.” That seems wild to me.
Also, there’s a whole other question of liability that I think is important to think about, as well.
The Chair: Would another witness like to jump in?
Mr. Krueger: Thank you.
I think tragedies like this are a bit of a wake-up call or a warning shot: They reveal how badly we’ve done in terms of regulating this technology and how challenging it is to regulate it effectively as it gets more and more powerful — and the need to slow down so that we have time to work through the appropriate ways to regulate things like this.
Right now, there are difficult questions around monitoring. For instance, you might be concerned about free speech or if the government will then go further in terms of what they’re surveilling and how you set it up such that the program is only reporting on the kinds of activities it’s intended to. Right now, these questions are being answered by the companies themselves based upon their profit motivation and desire to build this technology faster.
So, we’re missing basic things, like basic transparency. We’re missing liability, and we’re missing the opportunity to have meaningful conversations like this and have the time to actually have society have meaningful input into how we want to govern this. It’s just moving much too fast.
Senator McCallum: I want to look at discrimination and inequality as systemic failures that compromise the integrity of society, democracy and the realization of human rights for all. It’s urgent that we ensure historic discrimination not be embedded in AI models.
This issue with discrimination — is it even possible to mitigate? I don’t think it is. We can’t even mitigate what is happening now. When you look at Tumbler Ridge, it’s too late. Can you comment on that?
Mr. Krueger: There is a long history of work trying to address the problems of discrimination, bias and fairness in AI. We’re still lacking appropriate conceptual foundations to determine whether an AI system is discriminatory. There has been a lot of work on that, similar to how there’s been a lot of work on what’s called “interpretability,” which is understanding how the systems work internally. Similar to my answer to the previous question, these are big questions that we ought to give ourselves more time to grapple with.
I want to mention that there have been various attempts to correct for biases or discrimination that have backfired in both directions. We saw famous examples where Google’s model was generating Black Founding Fathers and Black Nazis in a kind of misguided effort at representation in their images when prompted to generate images of Nazis or the Founding Fathers of America. On the other hand, we had xAI trying to take a more “anti-woke” or “free speech” attitude, and it ended up with a model that declared itself “MechaHitler.”
Even these attempts to take a particular stance on issues of discrimination and fairness often don’t at all go the way their designers intend with current techniques.
Mr. Bourgon: I would echo and reinforce that the core challenge here is that these systems are essentially black boxes. When we train them, we’re basically just giving them a bunch of examples and having them imitate those. The actual thing we train is trillions of floating point numbers that just get multiplied together in a very particular way that makes a model behave the way it does, and no one has the ability to look at those numbers and understand why they’re doing what they’re doing.
So, when we say a model is discriminatory in one sense or another, no one can go inside and figure out why that discrimination is happening and correct that behaviour. They can try to show it some examples that would try to bias it in the other way. Then they can look at how the model is behaving and try to run it through some evaluations and check.
But the world is big, and there are often many ways for AI systems to interact. It’s very difficult to comprehensively test for that, so it’s an extremely difficult challenge.
Bias and discrimination are presently risks that we’re seeing and the same underlying concern. What is motivating our bigger‑picture worries is that, as those systems become more capable and because we don’t understand what their drives, goals and motivations are, those problems will only get bigger as they take greater actions in the world, with more autonomy and more intelligence.
The Chair: Mr. Buteau, did you have anything to comment on this?
Mr. Buteau: I will say that although there are some consequences of AI that are already kind of baked in or would be very hard to change course regarding, we have to fight the idea that it’s all hopeless. There are risks that have not yet materialized, and we are still in a position to act so as to ensure they do not. I think we should take those opportunities to do so.
The Chair: We are coming close to the end of our session. Mr. Bourgon, your internet is fluctuating a bit, but because we’re at the end, I think we can take to heart that we understood everything you said, so not to worry.
I have a final question as we wrap up, and, hopefully, it’s a quick one: Do you see that the remedy going forward would be some sort of a legal challenge, maybe in the area of human rights? I’m wondering, Mr. Krueger, if you’re seeing any of that in the U.S., so we’ll begin with you.
Mr. Krueger: There is a lot of space for legal challenges against AI companies. There are ongoing cases around their use of copyrighted data, which is, in many cases, illegal. There have been large settlements to that effect.
In terms of human rights, I think there must be some cases, but my understanding is that it’s been difficult to bring such lawsuits because of the lack of transparency and interpretability. As Malo mentioned, it’s hard to know if a system is discriminatory in terms of the “thoughts” that it’s having — how it’s processing the data.
A more fundamental issue is that people generally don’t have a right to know if an AI system is being used to process their job application, their loan application or their government benefits application, but we know they’re being used in all those applications. As somebody interacting with one of those systems, it’s difficult to claim that you were discriminated against, even if the system was discriminatory or biased, because you don’t know what the system is or have access to it.
That being said, the situation, as we have mentioned, is really in an extreme and unprecedented crisis. The activities of these companies are putting us all at risk. So there is certainly something like reckless endangerment of everybody here. I would be interested to see lawsuits trying to hold companies accountable for knowingly risking the lives of every human, which they have acknowledged they are doing.
The Chair: Thank you.
Mr. Buteau: This is mostly a common knowledge problem. When the issue is that most people or lawmakers have not been given the full picture, the first step is to speak out and encourage others to look into it and ensure that before we try to have this very targeted intervention, people understand what is going on. This will drastically increase the chance of success for any sort of means by which we would end up intervening.
The Chair: Thank you. Mr. Bourgon, anything from you to add?
Mr. Bourgon: As a person going third, I always have the luxury of saying that I would have said many of the same things but my eloquent panellists said them first. I don’t have much to add, but thank you for having me.
The Chair: Thank you all for being here. We appreciate your time and contribution to our study. I would like to sincerely thank you for taking the time to appear before us today. Your testimony will be very helpful in our deliberations. Thank you.
I will now introduce our third and final panel. Our witnesses have been asked to make opening statements of five minutes each. This will be followed by questions from the senators. With us by video conference is Joel Nicolas Blit, Associate Professor of Economics, University of Waterloo. Welcome. Also by video conference, we have Ms. Wendy Wong, Professor and Principal’s Research Chair, University of British Columbia Okanagan. I now invite Professor Blit to make his presentation, followed by Professor Wong.
Joel Nicolas Blit, Associate Professor, Department of Economics, University of Waterloo, as an individual: Thank you for having me. It is great to be here. I’m a professor of economics. I work at the intersection of innovation, technology, labour markets and public policy. My central point today is pretty simple: AI is a general-purpose technology that will transform every sector of the economy. Whether it narrows or winds, inequality in Canada is going to depend on the choices we make today, in particular whether we invest in AI literacy to empower all Canadians. That’s going to be my main point.
Let me start at the beginning. The launch of ChatGPT really is what marked the beginning of the AI revolution. It moved us from the age of predictive AI into the age of generative AI. But actually, that wasn’t the biggest impact. The biggest impact is that ChatGPT fundamentally democratized AI because all of a sudden, regular Canadians could use AI using natural language, in English or in French. All of a sudden, you didn’t have to be a programmer; anyone could interact with the technology. Therefore, the technology moved from our research centres and the back offices of our biggest firms to the fingertips of every Canadian.
How this coming AI revolution will unfold is no secret. In some of my work, I document how earlier general-purpose technologies like electricity and computers transformed the economy and society, and how they typically follow three different phases. I call them the three R’s: replace, reimagine and recombine.
During the first phase, the replace phase, the technology replaces other technologies within existing processes. It’s not until the reimagine phase that we see a radical reimagining of processes, structures, business models and — sometimes — even entire industries around the new technology. That’s the disruptive phase. Eventually, we reach the recombine phase, when the technology combines with other technologies to create entirely new technologies.
Now, the three R framework is useful for many things, including figuring out how Canada can benefit from and seize the opportunities of AI. It’s also useful for thinking about how AI will impact jobs and skills. That’s my focus today.
It’s not too hard to predict what jobs will be impacted in this first replace phase. All we need to do is look at which occupations consist largely of tasks that AI can already perform. Roughly speaking, about one fifth of Canadians are going to see their jobs impacted very little. About three fifths of Canadians will see their jobs impacted a moderate amount, meaning between 10% and 50% of the tasks they are doing are things that AI can do, and about one in five Canadians is going to see big impacts. These are people like writers, translators, journalists, accountants, lawyers and programmers, et cetera.
That’s in the replace phase. It then becomes much harder to predict impacts on jobs once we reach the much more disruptive reimagine phase. For example, I could not have predicted that the internet was going to result in the loss of so many retail jobs, but that’s exactly what happened when Amazon and Shopify reimagined the retail industry around the internet.
New disruptive technologies have always resulted in both the loss of jobs and the creation of new ones. We have to be clear that the transition is never easy. However, the most important long-run impact of disruptive technologies is that they tend to create winners and losers, depending on the types of skills that people have.
For example, industrial robots increase inequality by replacing individuals at the bottom of skill distribution. Computers increase inequality by replacing people in the middle of skill distribution while, at the same time, lifting the productivity and wages of the most skilled. For AI, here is the good news: It’s likely going to decrease income inequality because it’s displacing the most skilled. For example, a nurse empowered by AI will soon be able to perform the job of a doctor.
Now, I want to be clear that there are some important caveats to that, and two in particular: The first one is that not all white‑collar workers will be affected the same. The law clerk or the paralegal might lose their job, but the partner will not. We need to ask ourselves what will happen if most entry-level jobs disappear — something that may already be happening.
The second important caveat is that a new divide could emerge, and that divide is between those who are empowered by AI — maybe not just empowered but turbocharged — and those who are completely left behind.
I want to finish on that last point. Specifically, if AI is going to boost Canadian productivity, and if we want AI to be an inclusive technology, then we have to empower all Canadians, and that means investing in AI literacy. The United Kingdom is rolling out a national AI skills drive that will educate 1 million students and 7.5 million workers by 2030. Singapore’s whole-of-nation movement has brought AI literacy and skills as a major pillar.
Canada has no plan. That means that Canadians — or, at least, many Canadians — are at risk of being left behind. In short, Canada needs to embark on a nation-building project for the 21st century. We need to build not just an AI industry but an AI nation, a country where the use of AI becomes a shared civic skill woven into the economy and society, empowering every person and organization to thrive in the age of intelligence. That’s how we make AI powerful and inclusive.
Thank you.
The Chair: Thank you, Mr. Blit.
Wendy H. Wong, Professor and Principal’s Research Chair, University of British Columbia Okanagan, as an individual: Thank you so much for the opportunity to speak with you today.
Since the United Nations, or UN, system was established, human rights have been at the core of global governance and national law. AI is a set of technologies that challenges the way we think about human rights, not because human rights no longer matter, but because of how AI technologies challenge our basic assumptions about how we live our lives.
Like Professor Blit, I will talk about literacy, but digital literacy, and I think about digital literacy as an extension to the right to education, a basic right that other rights — such as freedom of conscience and expression and the right to work — build on.
Since the beginning of this century, we have become ever more dependent on digital technologies that take basic activities like movement, writing and speech, location and even our faces, then convert them into data to be processed by ever more sophisticated computing. But these processes work largely in the background as algorithms crunching ever more data.
To see the impact of AI on human rights, we need to highlight what has been changed by translating life into data. To realize human rights today, we need to create and protect digital literacy for all Canadians.
I’ve argued that digital data stick to us like gum on the bottom of a shoe. It’s difficult to know when we’ve made data, and it’s very hard to get rid of them once they’re made. We used to call personally identifying information “personal” or “sensitive” data. These data describe us as individuals and were fairly rare. Now they are being made everywhere and all the time.
Our systems are now geared such that even something as commonplace as Ring doorbells or Tinder data can be used with other data to find our addresses and identify our faces. AI systems can follow us home, and they are, in fact, in our homes.
The problem is that, as everything we do becomes data, we start losing the ability to enjoy even the most basic values that underlie human rights: dignity, autonomy, equality and community. Regardless of whether data collectors are fair or nefarious, the more they know about us, the less choice and freedom we have.
The federal government needs to act decisively. It is one thing to invest in AI jobs and the development of advanced digital technologies, but we must also invest in the people for whom — and on whom — these technologies work.
First, the federal government should invest in the development of digital literacy skills as part of the right to education. Digital interactions are no longer avoidable, reserved for certain sectors or tied to income. Data are part of our lives. Currently, Canadians risk being divided, not just via access to but also literacy around digital tools like AI.
Second, the federal government should work with provincial governments to develop a coordinated national digital strategy. Such a coordinated strategy would ensure that Canadians are able to make informed choices, build ethical, cutting-edge systems and continue to exercise their Charter rights. Coordination between federal and provincial governments will ensure that Canadians are not divided by access to or literacy regarding digital tools.
Third, we need to make use of expert literacy institutions outside the tech sector, such as public libraries. Libraries are integral for widespread literacy. For example, the Toronto Public Library has created over 1,200 successful programs to date toward AI literacy. Its efforts have reached tens of thousands of citizens. Organizations such as MediaSmarts provide off-the-shelf resources for parents and teachers on a variety of digital media literacy topics, including games, checklists and explainers.
Finally, we can learn from other human-rights-respecting countries, which have already invested in national digital literacy strategies. Finland, Australia and the U.K. all have comprehensive digital literacy strategies that articulate general literacy goals with regard to accessibility of digital technologies and addressing the underserved. We can adapt these models here in Canada.
Canadians are resourceful and adaptable. Widespread digital literacy will help us enjoy human rights in the age of AI.
Thank you very much.
The Chair: Thank you both. We will now advance to questions.
Senator McPhedran: Thank you both for being with us today.
My question is about the digital divide, which, I think, translates now into an AI divide. I want to combine that growing reality with a question to both of you. You have both made the point that Canada is behind at this stage. What are the consequences for this country if we stay behind?
Professor Blit?
Mr. Blit: Sure. I can start. There are a couple ways to answer. First, if we stay behind in AI, and more and more of our industries leverage AI, and more and more productivity is driven by AI — which is what everything seems to indicate, and we know that over the last 50 years, we have had the lowest productivity growth of any G7 country — it means that unless we change this trajectory, our kids are not going to have the opportunities that we had.
It means that the standard of living of Canadians is not going to be the one we have become accustomed to. It means we may not be able to afford the things that we identify with most as Canadians, like our quality public health care system or our quality public education.
That’s the first order or piece. Unless we really get on this, Canadian sovereignty is in many ways at risk because our economic trajectory, which has been so bad, shows no sign of changing.
The second piece is that unless we educate all Canadians, there will be a divide. That divide will be with respect to those who are turbocharged by AI. We have a lot of top researchers and people in the field, but AI really is a general-risk and general‑purpose technology. We want every Canadian to be empowered by it, to be able to be a part of this productivity improvement and to be able to benefit in terms of higher wages.
The second argument is the inclusivity argument. That’s the other reason why we need to do this. Those would be the consequences.
Ms. Wong: I actually see this as an opportunity for Canada. The way you framed it, senator, was that we are behind. I think we are not first movers, but there is always a second-mover advantage, which is that we have the vantage point of other countries making decisions. There is, for example, social media banning, as in Australia — many people have come out against that. I also think it’s a flawed policy.
Also, we have an opportunity to really develop the skills that both of us are talking about here, with a mind toward innovation. Running ahead with regard to AI development, on the one hand, might bring economic opportunities. However, it does intensify divides between different communities in a country. Traditionally underserved populations will, of course, struggle to keep abreast. This is a real opportunity for the Canadian government, at multiple levels, to develop a coherent strategy such that we as Canadians enjoy literacy training with no difference between provinces.
The other issue that we need to raise is the power of big tech companies, which are largely based outside Canadian borders. I think that’s a real issue. We need to tackle that, through not necessarily through brute force tactics. It’s not necessarily by upping our computing resources or simply dumping resources into AI research, which is important but needs to be more measured.
Senator McPhedran: Perhaps someone else will pick this up from here. If I had more time, I would have asked each of you, but particularly you, Professor Wong, about the statement around the Charter of Rights and Freedoms and the link you made between AI literacy and the enjoyment or the living of rights under the Charter.
Ms. Wong: Do you want me to quickly respond?
The Chair: Thank you. I’m asking that question so you can go ahead and answer.
Ms. Wong: Okay, absolutely. I think the right to education is one of those fundamental rights we take for granted in a country like Canada, where we do have widespread education. Not knowing — not having basic information and basic conceptual capacities and skills — is an impediment to our ability to participate in a free society. I think of something like the right to education as one that undergirds almost everything that we’re exercising in accordance with Charter rights.
Freedom of expression is not possible or is severely inhibited if we don’t have the right background knowledge to operate, to ask the right questions and to stand up for the things we believe in. I think about these literacy questions as going beyond AI. AI is a digital technology. It’s the one that people are talking about most. But, to me, the biggest change between our lives 20 years ago and where we find ourselves now is that so much of our lives have become digital data. They have become machine-readable and usable by computers to make predictions, to generate patterns and all the things we associate with AI today.
The Chair: Professor Blit, did you want to comment on that? No? Okay. Thank you.
Senator McCallum: For the two panellists, thank you for your presentations. Did you hear the former three presenters? The question I had asked them regarded the fact that, when we look at human rights and the difficulty humans have in correcting injustices for vulnerable populations, we, oven many lifetimes, have not even made a dent. We’re human. AI is not human. One of the presenters said there is grooming happening here and biases entering it. AI is groomed; it’s not built.
To what extent does AI threaten to erode ethics, critical thinking and interdisciplinary skills, including tradition? The context for that is the deep learning that AI is supposed to do and First Nations’ wisdom — what it generates into wisdom. That is the purpose of deep learning, but AI can’t do that.
To what extent does AI threaten the human population, including students, who are forming their thoughts and their critical skills? They are forming ethical, sustainable, human‑centred learning. Can you comment on that?
Ms. Wong: I can start.
I think you’re asking exactly the right questions. This is a question of literacy, of really understanding what is going on in terms of digital technologies. AI is one of these technologies that is very dependent on the data that we feed those systems. You’re absolutely right in the sense that you create bias if you feed biased data into the system.
The solution could be to just feed the system as much data as possible. I actually think that goes against human rights. My concern is with regard to how we have made nearly all human activities into data. It is possible to track not just things we’re explicitly aware of, but things that we do: the way that we move or the way that our bodies function. I think that is a problem for human rights in terms of our autonomy and our dignity as individuals.
In addressing injustices, the solution isn’t to automate them. I think you were speaking to how students can learn. Something that I tell my students is that being critical doesn’t mean criticizing; it means understanding where the answers you’re getting come from. That’s really a literacy question around understanding how important data are to AI systems.
Mr. Blit: I thought maybe I would pick up on the learning and the critical skills part of the question. I’m in the faculty of arts. This is a debate that we are constantly having. How is AI and the use of ChatGPT and other tools impacting our students’ ability to learn? A lot of my colleagues have taken the approach that we are going to ban that technology because students are using it instead of their own critical thinking skills to solve problems, et cetera. That is one approach.
My approach is different. I think that we can actually leverage AI as a tool to teach students better than we have ever taught them in the past. For example, there is a seminal paper from the 1980s by Bloom, who basically shows that if everyone has access to a one-on-one tutor — that the average student with a one-on-one tutor will be two standard deviations above the performance of a regular class with no private tutor.
In other words, they’re going to outperform 97.5% of regular students. This is an average student. The problem is that we’ve never been able to afford that because we can’t afford, as an educational system, to give everyone a private tutor. However, we now have AI systems that can deliver that.
As with any technology that’s disruptive like this one, there are pluses and then downsides and risks. We need to learn to leverage the pluses, not bury our heads in the sand, but leverage the pluses while trying to mitigate and address some of the risks and challenges.
Senator Robinson: Thank you. You almost answered my question with your last response.
We’ve heard about this potential disruption from AI to jobs, and I wanted to spend a bit of time talking about the potential for the positive and the augmentation of existing jobs in some sectors. My area of focus would be agriculture. I’m wondering if either of you would like to weigh in on that. I’m seeing nods from both of you, so I might start with Professor Wong.
Ms. Wong: Not to signal expertise in the area of agriculture, but, yes, I think that is an area where AI can be leveraged. I think that we’re learning more and more through the collection of data about things like soil conditions and the general climate, things that perhaps would have been very difficult to measure previously without the use of automation.
I think that agriculture is an area where we can improve human lives and productivity. Dangerous jobs, such as mining, for example, would be another area where we could use AI technologies to lessen harms to human beings, to make the conditions safer or even to remove humans from the most dangerous parts of those jobs.
I think, to speak to higher education, it’s not that AI technologies are universally bad or will threaten all jobs. In fact, they augment many jobs. I know Professor Blit talked about the medical field. We’ve learned a lot through the use of AI technologies. I think there are definitely benefits, but there are areas where we have to be mindful of human rights being affected negatively.
Senator Robinson: Thank you. Professor Blit?
Mr. Blit: Sorry, I’m also not an expert on agriculture, but certainly — I’ll echo what Professor Wong was just saying — predicting what crops to plant in different soil conditions with different weather patterns, when to harvest, autonomous pickers, there are probably going to be a lot of different applications in that space — and across many other spaces.
My wife is a family doctor, and just with AI scribes, her productivity has increased about 20%. Now she has more time to visit with her patients and look them in the eyes. They’re talking instead of typing because there’s an AI scribe doing that. That’s just one example. Another might be the detection of skin cancer through imaging. That’s also something my wife is doing. There are a lot of potential upsides, as well as many of the risks that have also been discussed.
Senator Robinson: There are many parallels in my mind between medicine and agriculture in that we’re seeing a lot of accelerated research because we can have so many replications of a trial. When you’re looking for the right source, placement and amount of fertilizer for a varying soil type, for a different seed, all of this we can do replications of and collect an infinite number of times the data. That’s going to impact our ability to be more productive and more efficient in food production. I think it is as exciting in food production, Professor Blit, as it is in medicine. Thank you for your comments.
Senator K. Wells: I want to return to something I know you’re both experts in, which is education, having that experience within post-secondary classrooms myself as a former professor and seeing the trends. As was pointed out, we’re seeing some school jurisdictions now going back to buying textbooks, and we see a lot of school districts banning cellphones from classrooms — putting them in a little mailbox. We’re even seeing university professors turning back to handwritten, paper‑based exams. Thankfully, I started that way and don’t have to end my career by going back to trying to decipher students’ handwriting — which is another lost art, thanks to our technology.
You mentioned the importance of digital literacy skills. In Canada, in particular, our K to 12 system is a provincial and territorial responsibility. What we do in one province may not be happening in another. Despite whatever national framework we have, it’s largely going to be up to the provinces to implement this digital literacy. I’m sure, as you’ve noticed, it can depend a lot on where a student went to school, to be able to see what they can do in a university environment where they’re not even getting as much one-on-one specialized education or skill development. We expect them to have the skills and to produce more critical thinking and analysis at a university level.
These are all great ideas, but practically, when we have such a decentralized education system, with instructors in universities who may all have different syllabi and curricula and the autonomy to do so, how do we come up with a coordinated education strategy if these are 21st century skills? We are in the midst of an AI revolution, much like the Industrial Revolution, which fundamentally changed how education is delivered and operationalized. We have not had such an advancement since the industrial education model that we’re using for our education system, and AI is maybe part of the unschooling movement. Do we need classrooms? Do we need teachers? Do we need to go through these discrete levels of Grade 1, Grade 2, Grade 3 and so on, which may actually hold certain students back as opposed to achieving mastery at an earlier age when they have that specialized instruction?
Maybe this is me being more philosophical. Strategy is a great idea, but how do we practically implement that if Canada is behind and we’re relying on our educational system to help catch up? How do we do that in a coordinated way? I’ll open it to your thoughts.
Ms. Wong: Those are the right questions. It’s great. Also, I just gave a paper exam today. Going back, we have something in common.
When I speak to policy-makers, I really encourage people to think about how the federal government can coordinate with provincial governments on this issue of education and revamping curricula to take into account the digitization of our lives. I think that you’re right: It’s not an easy hill to climb. In fact, given my understanding of our politics, it’s a fairly challenging one, but I think we have a massive imperative to do so. It is urgent for the reasons that we’re talking about today. I think that we cannot be successful as a country with regard to AI and other digital technologies if we don’t rethink the way we educate our kids. It starts with that.
It’s also about adult education, and we have to be creative in terms of thinking about how we reach people. Maybe it’s through classrooms, through the library or through other civic efforts that we build out. I think the federal government has a huge role to play in coordinating — not commanding but certainly coordinating — with regard to literacy skills in the digital realm.
Mr. Blit: I agree with all that. I guess your question of “how” is an excellent one. We always have this in Canada with the federal and provincial systems. I’m going to leave that question for you because you are better at thinking about how to make this happen. My comparative advantage is around what we should be doing.
As Professor Wong was saying, I think the federal government can definitely convene. It’s one of the areas where it can really play a role. It also has a lot of resources. Some of these experiments that I think we need to run are going to require significant resources, and they could be funded by the federal government. Just as we have an AI safety institute, there could be an AI learning institute or something similar that would run experiments and gather and diffuse ideas.
I want to point out that we shouldn’t just be thinking in terms of educating our students. When I talk about AI literacy, it really includes everyone. We can’t forget every adult who is already out of school because they will also be in this digital AI economy. So we will need something different for them — some kind of continuing education, online courses or something of the sort.
Finally, I don’t know if you’re familiar with Alpha Schools. These are schools where students learn on tablets for two hours a day, and with only two hours, they outperform all other students with traditional teaching. Those lessons are not even adaptive yet. In other words, they don’t yet have AI that is adapting to the way of learning of the students, but what they do — they’re mastery based. Instead of saying, “Oh, you’re in Grade 1, Grade 2” — no, forget that. It’s, “Have you mastered what you need to master?”
I think we need a radical rethinking of how we are teaching and what the goals are. That’s how we’re going to build the next generation of really elite, educated students. That needs to happen, but there’s tremendous resistance.
Senator Karetak-Lindell: On that last answer you just gave, Professor Blit, I’ve been talking about that for a while, because in Nunavut, our education system has not worked the way that it should have. We don’t have the graduates that we would like to have seen in the 70 or 80 years of doing K to 12. There are different ways of learning, and I keep trying to say that we have to find different ways of teaching because what we’re doing now is not working. So I’m really encouraged by that. Then I heard people are resisting this. That’s something that we’re definitely going to have to look at.
My question was more along the lines of what Senator McCallum was saying, about the biases and how it seems that AI works differently depending on what data you put into it. For Indigenous groups, there’s always a bias in all the education systems, hiring systems and health systems. It seems like every government system is biased in some way because they don’t know how we think, and they don’t know how we have our own ways of doing things, so I’m worried about the data that goes into AI.
How can we ensure different data is input so that they’re able to think through different ways of thinking? Or is it not possible? Is there a way to put in data that recognizes there are different ways of doing things, not just one? How do we input data in ways where people don’t get discriminated against depending on what colour their skin is, what orientation they are or what language they speak? Is there a way to mitigate biases? We keep hearing these systems are groomed to produce a certain answer. Is this another area we have to protect people from — the bias of AI systems? Either one of you can start.
Ms. Wong: This is, again, an excellent question in the sense that I think you’re pointing to exactly what the issue is with AI. We tend to think about computers as being more objective than human beings, but we know that AI systems are built by human beings for specific purposes. Whether they exclude data from those systems on purpose or by accident, every system is going to have some kind of bias, just as human beings tend to be biased in one way or another.
I think the real question that your concern is getting at is this: Should we deploy AI systems in any and all occasions because they might be more efficient and they might be seen as more effective? As Canadians, we need to think about exactly the questions you and other senators are posing, on how there will be spaces where we don’t want AI to be. I’m perhaps less enthusiastic than Professor Blit about this. This is a time to be contemplative, to think about when we want automation, when AI will make systems better and when AI could actually make systemic biases worse — or create biases that maybe didn’t exist but now do because of patterns found in the data.
We need to take a bit of time. As I’ve said before, there are second-mover advantages as well. We don’t have to be out there first. We can do things better because we have the advantage of sitting back and seeing where people are going at a national level.
Mr. Blit: I’m very sympathetic to everything you’re raising. I’m Latin American. I’m an immigrant. I have seen these biases myself in all kinds of different realms. It’s also fair to say, as Professor Wong just did, that humans are biased — and I’ve seen human biases — and systems are going to be biased.
We need to endeavour to make these systems as unbiased as possible. We need to have enough transparency around them so that they can be audited and we can find out how well they’re doing. The reason why it’s such a hard problem is that, fundamentally, it is going to require humans to make choices. When you decide what data should go in and which ones shouldn’t, it’s not just a matter of saying, “Let’s put all data on the internet in there,” because we know the internet has some terrible stuff on it. There have been experiments where they have tried to put everything in, and there have been really toxic results coming out of that. Humans will need to make decisions, and every time that humans make decisions, there is going to be a bias one way or the other. There’s going to be an alignment problem.
I think the best that we can do is to try to have open eyes, to audit these systems and to try to bring in as many different voices as possible.
I like to think of machine learning and AI systems as nothing more than systems or models that capture people’s tacit knowledge. Once you capture that tacit knowledge, you can either automate a task that requires it or you can share it with others. Let’s ensure we’re capturing the tacit knowledge of everyone and not just a small group of people.
Senator Arnold: Thank you both for being here today. You’re a lot cheerier than our last session, when we were talking about artificial superintelligence. Today, we’re with you talking more about machine learning and general-purpose technology, and it’s here to stay. I think you have both admitted that data collection is happening. There’s not much we can do about it right now.
Professor Wong, you did say that you don’t necessarily agree with Australia’s new laws around teenagers and the use of social media. Could explain your reticence to us?
Ms. Wong: Absolutely. Also, data collection has happened and is happening, but there are policy choices that can be made to either limit the types of data we collect or protect certain classes of data, which we haven’t done at the federal level sufficiently. I know Parliament is working through revising our privacy and personal data laws.
The blanket ban on social media is a problem because it hands over the decision making to the companies that produce social media technologies to begin with. You now need to age verify. Well, who’s going to verify those ages, and what types of data do you need to hand over to verify age? I think that’s a real issue if we’re concerned with the proliferation of digitized personal data.
If I hand over my driver’s licence — which is how many of us understand personal data, or one of the fundamental tenets of that — I am giving it to a private entity that, once they have those data, what are they doing with them? I’m not sure. I don’t think anyone can say. At this point, we don’t really have the capacity to monitor that process. What you’re fundamentally doing is adding to the data collection problem, even if you’re saying that we need to ban the use of social media for younger people for a variety of well-intentioned reasons.
My position on that, because my work is focused on human rights and data, is that making more data about humans or human activities is always something to be done with caution.
Senator Arnold: Where is the biggest risk in that?
Ms. Wong: We don’t have a real set of laws around who ought to have access to what data. Right now, we rely on identified data. If I can link any data directly back to Wendy Wong, that is a problem. Once the data has been de-identified — let’s say it’s just about demographics or people like me — those data don’t have the same protections and privileges.
On the one hand, people might think it doesn’t matter because it’s not linked to them, but there are ways to link back to individuals that do not use what we classically think of as personally identifying information, for example, certain marketing identifiers that are linked to all of us because we use apps. If people want to, there are a lot of ways that they can figure out where you are, what you do and where you’ve been.
Senator Arnold: It’s interesting because we’ve been talking about data literacy and the conversation has gone more toward younger people, but I think a lot of adults don’t have a clue about the data they’re providing out there day after day.
Professor Blit, did you have anything you wanted to add to that? No. Thank you.
The Chair: I want to go back to the data collection that is constantly being done, whether it’s through an app update, or my Air Canada app, which is always asking me for feedback, or terms of agreement that you have to sift through — I don’t know how many of us actually do that — that consist of issues around sharing your data, and you don’t have a choice but to agree. There’s only one option with some of those. I’m not sure that, as consumers, we are making the link between that and the sharing of data points about ourselves as well as in general.
Have you been part of the strategic consultation that the government has done? Have you had an opportunity to contribute to that? Is this something that you would want to see addressed in such a strategy?
Ms. Wong: You’re asking whether public consultation is a strategy —
The Chair: For the federal government strategy.
Ms. Wong: Think it goes back to where you started your question, which regarded the public’s general awareness of the depths to which data are being collected and the scope of data collection. I think that meaningful and useful public consultation would have to hinge on that awareness and that acknowledgement.
That being said, in the polls I have come across, Canadians are — much more so than peer countries — skeptical of AI technologies. They’re more reluctant to engage them than other people in different countries. I think there is that sort of hesitation, and I wonder if a public consultation would be able to probe that a little better than more general polling strategies.
It is an important strategy. We’d certainly need more time for that than what was given for the AI consultation last year.
The Chair: I understand that consultation was just one month. Were you a part of that, Professor Wong or Professor Blit?
Ms. Wong: No, not formally. There was a survey or something. There were multiple ways to participate. I did submit my opinion online, but I was not consulted in person.
The Chair: Thank you. Do you have anything to add, Professor Blit?
Mr. Blit: I certainly participated in that consultation a few different ways. I agree that a month is very short. On the other hand, it’s moving so fast that we don’t have a huge amount of time, so I can see both sides.
To your point on whether Canadians truly understand the risks when they agree to terms, et cetera, that’s a big part of why I am pushing really hard. This is part of what I said as part of that consultation on AI literacy. It’s not just AI literacy but digital literacy more generally. Canadians need to understand not just how to use these tools but what these tools do and what they mean.
As Professor Wong just said, Canadians are among the most pessimistic on AI, which worries me. On the one hand, it’s good because we’re not just jumping in blindly but with our eyes open. On the other hand, we’re not adopting as quickly as we should. Although, some evidence does suggest that Canadians themselves are adopting these tools quickly. It’s our businesses that are not. That is another big part of the problem for Canada: that our businesses are not jumping onboard.
The Chair: Thank you. In your research, are you seeing a generational divide?
Mr. Blit: I have not looked directly at that. I do a lot of executive education, and I see a generational divide there. Younger people are definitely adopting a lot more. I have had CEOs say to me, when I talk about all the challenges of AI, “That’s why we’re not adopting anytime soon.” The person at the far end of the room who might be the marketing director and who might be a lot younger will say, “Actually, we’ve been using it for two years now.” So there is this divide.
Again, businesses in Canada are not adopting. We are very risk-averse, but it filters in because workers and individuals in businesses use it, sometimes at home, sometimes on their own time, et cetera.
The Chair: Thank you. Do you have anything to add, Professor Wong?
Ms. Wong: Yes, thank you. Just to add to the point that Professor Blit just made, it’s important to distinguish between different types of AI as well. I think the large language models that really exploded on the scene with ChatGPT, and now with Claude and people being able to vibe code, that’s a different type of AI from the one many of us have used for a long time, which is autocorrect. That is a form of automation — a form of machine learning. Facial recognition is another form of technology that I think has very distinct human rights implications compared to other types of AI. I would add that caveat.
Senator McCallum: I want to go back to the use of personal data.
About two years ago, five senators, myself included, went to a university performing research on health. They had an agreement with an Indigenous group, and personal data was coming in. This had to do with DNA sequencing.
AI goes into areas that we cannot even think about. That’s how fast it moves. I asked them how, when it moves in a direction that their research wasn’t taking it or where it wasn’t intended to go, they go back to the group and get consent. They would have to get that in many different areas. The response was, “We have to get rid of the consent.”
My point here is that the use of data is dependent on who manipulates it. In this case, the DNA sequencing was to target areas where your body was susceptible — to diabetes, heart problems and so on. One person said, “We could have humans live up to 140 years.” I replied, “Who is going to take care of these people?” When we have not even made a dent in addressing poverty, who will be the beneficiary of this AI? The AI is leading them; they are not leading it.
My concern for AI grew there. I do not believe that when you move that fast, there can be transparency because you don’t know where it’s going. So when you say that we can look at transparency and that this can deeply affect human rights issues, do you believe there is transparency? Is there a capacity to monitor that transparency?
Ms. Wong: Most of the advanced AI systems that we know today are very difficult to monitor. The short answer to your question is that transparency is very difficult. I don’t know if transparency is the right tool here. I think you mentioned consent. The reason why I’m going to focus on that is that consent is the heart of exercising autonomy for every one of us. It is also a way we can treat people with dignity: to ask for consent for the use of personal information. That is not possible today.
Madam Chair referred to the terms and conditions; we don’t read those. They are tens of thousands of words long. Part of the problem is that, even if you do, you discover a lot of power is used by saying, “We’re going to share this data with third parties.” That is where we have lost our ability to really consent because we don’t know who those third parties are or what their intentions are. I don’t think we can even infer the intentions of the companies with which we’re signing those terms and conditions.
So right now, one of the biggest challenges with the datafication of human life is this consent question. “Can we consent?” is a different question from “Ought we consent?” I think we ought to be able to exercise some version of autonomy and be treated with dignity even if we’re living through the age of artificial intelligence.
Mr. Blit: Maybe I can pick up on a piece of what you said, which was that we have barely made a dent in poverty. Yes, I agree. We need to be doing a lot more. The way I think about AI involves two things.
First, it’s an opportunity to really grow that economic pie, and we absolutely need to do that. To some extent, we have no choice but to go on that journey because if we stop, everyone else will still continue. We will be left behind. So it’s not a matter of whether we adopt AI and seize the opportunities; it’s a matter of how we do so.
Second, we have to ensure that all those gains are not just going to the 1% or the 2% or whatever. They are not going to the people that own the technology. We need to make sure that they are reaching everyone. Again, I go back to AI literacy and ensuring everyone has access to these technologies and they are not just owned by a small group. These are real concerns. The Competition Bureau, for example, is thinking very hard about these issues.
The Chair: Thank you both very much. We have come to the end of our panel. On behalf of the committee, I would like to sincerely thank you for taking the time to appear before us today. Your testimony will be very helpful in our continued deliberations on this topic.
(The committee adjourned.)