The Environment and AI with Kelly Bronson
Season 8 Episode 8 [download this episode (MP3, 24.0 MB)]
In this episode Elizabeth chats with Kelly Bronson, Canada Research Chair in Science and Society at the University of Ottawa, to explore the environmental impacts of AI and why they need to be understood as social and justice issues, not just technical ones. They discuss the energy and water demands of the hyperscale data centres, and why measuring wattage or carbon alone narrows both the problem and the solutions. Kelly explains how data centres are often sited on Indigenous land and in communities already marginalized by extraction and colonialism. She also looks at how hype, from AI solving global hunger to the “probability of doom,” keeps decision-marking in the hands of industry and engineers, using the fully autonomous greenhouse as an example. Kelly makes the case for taking a full life cycle view of AI, from mining and manufacturing to labour and data. She argues Canada’s AI strategy treats AI as an unexamined public good, and that the key missing piece is a governance process that affects communities and workers into defining what counts as strategic value.
Additional Resources:
Register for the event on Dec. 3 to see a live recording of Wonks and War Rooms! From 6-7:30pm in FTX 147.
Read Government of Canada launches Canada’s Responsible Data Centre Development Principles, if you are interested in the new principles document that the government released September 3rd.
To find out more about the role engineers play in building systems where there are people who feel it is their responsibility to build ethical AI, tune in to Jason Millar's episode.
To learn more about the term P(doom) check out What is your P(doom)?
Look into the The Canadian AI Safety Institute for more information about the government's strategy to develop AI.
Check out Can machine-grown lettuce help cut Canada's reliance on U.S. greens? This farmer is betting on it about how Jay Willmot created a “machine lettuce” production.
To learn about how AI environments affect Indigenous lands by checking out Jonathan Dewar CEO of FNIGC on a previous episode.
Check out Kumanan Wilson episode about how data sovereignty looks like health policy and why it matters when building and training AI tools.
Read Jane Bronson's recent work Reviewing the evidence on precision agriculture and environmental sustainability which demonstrates that precision agriculture use leads to environmental benefits.
Check out Kelly Bronson on LinkedIn!
Contributors:
Host: Elizabeth Dubois
Research Lead: Areej El-Sharif
Transcript: Areej El-Sharif
Audio Editing: Sarah Pledge-Dickson
Translation: Julianne Bernier Robert
Publishing and Promotion: Aaron Brisson
Producer: Sarah Pledge-Dickson
Episode Transcript: Season 8 Episode 8: The Environment and AI
Read the transcript below or download a copy in the language of your choice:
Elizabeth Dubois: [00:00:05] Welcome to Wonks and War Rooms, where political communication theory meets on the ground strategy. I'm your host, Elizabeth Dubois. I'm an associate professor and University Research Chair in Politics, Communication and Technology at the University of Ottawa. My pronouns are she/her. This season, we're talking about social and policy issues surrounding uses and governance of artificial intelligence. And today we're talking with Kelly Bronson about the environment in AI. Kelly, can you introduce yourself, please?
Kelly Bronson: [00:00:30] Sure. Thanks for having me, Elizabeth. So [I’m] Kelly Bronson, I'm a Canada Research Chair at the University of Ottawa. The title is Science and Society. It's a bit broad, but I do have expertise, I don't know if you knew this across both sides of that fence. So I have a distant background now in biology, plant biology and then genetics. And then I moved from kind of practicing genetics to being really interested in public controversy around genetic engineering. And so that's how I moved to sociology. And I now sit on this kind of awkward but pretty functional three legged stool between the Center for Law, Technology and Society. That's where we sort of come together, you and I, the Institute on the Environment or Environmental Studies and sociology, anthropology.
Elizabeth Dubois: [00:01:17] Amazing. Thank you. I did not know you had a background in biology and genetics. That's super interesting. And I'm sure having that science background also informs a lot of how you think about these kinds of problems. Well, we'll get into how you think about these kinds of problems in a minute. Let's actually start with some definitions. Like I like to start all the episodes with. So we're talking about AI environment. A lot of time people are thinking about maybe carbon emissions, maybe they're thinking about how much water it takes. Maybe they're thinking, actually, AI is going to solve environmental problems. But then there's a lot of other things they might be thinking about. So maybe we can lay the groundwork. When I say AI and environment, what are you thinking about?
Kelly Bronson: [00:01:56] What do I think of it? Well, of course I think about those things. So, you know, environmental impacts include energy use. Not specifically, I guess, AI, but the infrastructure required to run, especially large language models, require massive amounts of data in their training and their use. So this is, you know, data stored in data centers. They require massive amounts of energy. And not all data centers are the same. The data center could be just a computer in your basement that's hosting a website. But we're talking here about the massive industrial scale, even hyper scale data centers required for these large language models that are super energy intensive. So, you know, yeah, south of the border, we've seen all this controversy around energy, hyperscale data center requires, I think it's like up to 50MW of energy. So just to put that into context, that's like the energy use required at any given time to run 37,000 homes. So that's a lot of energy and you know, the controversy we've seen, there's all sorts of elements to it, but part of it has been that then these data centers put massive pressure on existing energy infrastructure, like electricity grid. Elon Musk's Colossus, which I just love, and also I'm totally disgusted by the naming of that hyperscale facility to run his model X AI. They've just so exceeded the grid that they've had to use all these very creative and also kind of terrible stopgap measures, like plugging in generators that then emit gas. And they're noisy and community's been rightly quite pissed off about that. And, you know, he's also building these putatively temporary, although it's a huge financial investment, methane gas plants, which of course, we all know methane is one of the worst climate offenders. So yeah, energy concerns me. And then even within that, like climate impacts, so greenhouse gas emissions, like methane from AI, I guess infrastructure, or they think about the water because, you know, the microchips inside the data centers get really hot.
Kelly Bronson: [00:04:01] They can melt. And so they require continuous amounts of clean water, just like in, you know, our drinking water infrastructure. If the water is not clean, then it's going to erode the pipes, etc.. So it competes with drinking water. They have water impacts in terms of pollution. So I think about all those things, but even just like the examples I gave, you can already start to see that the way I look at or frame these environmental impacts of AI, I frame them a little bit broader. So if you think about these impacts are not just ecosystem impacts, right? They're social. You know, if you look at these where these hyper scale facilities are sited, it's in areas where land is more abundant and cheaper, right? And where land is more abundant and cheaper. It's not in urban centers where relatively wealthy people live, it's in places where land is also cheaper, not just for developers, but also for ordinary people in the United States or within literature. Sometimes we call these like environmental justice communities in Canada. Most of the big data centers under development and there are many. And the hyperscale centers proposed for development are on Indigenous land. Right. So these are like, you know, remote communities that have a history of, I would say, trauma, right? From environmental extraction, development, mining, I mean, ongoing formations of colonialism. And so you can already see how poverty, colonialism, right, come together. So I frame things a bit broader, we might say kind of sociotechnical framing or environmental justice, not just environmental. Yeah. And I could say more, but I'll stop there.
Elizabeth Dubois: [00:05:38] Yeah, that's really, really helpful. I'm so glad you started out with some concrete examples and then helped kind of pull us into that wider way of thinking about it, setting it in that sociotechnical context, because it is really easy, I think, particularly when we're thinking about engineering problems, to be like, we need to drill down to like the very specific standard that can be set, and then everyone can optimize to that particular standard, you know, like the number of watts, right? But that's not how like the lived experience of AI rollout is, is actually being felt, particularly, as you say, like in these communities that are often already disadvantaged in a whole bunch of different ways, already marginalized in our society and by many decades of extraction and historical kind of disconnect between what the actual best thing for those communities is versus what the government or companies are going to decide to do.
Kelly Bronson: [00:06:37] Yeah, both. You know, these are often communities where there is that history of mistrust or distrust between government, I mean, settler versus, you know, maybe a nation and and then certainly industry. Yeah. You know, people living at the coalface of well, in this case, data centers. It's not that they're not concerned about wattage. Right. But for them, it might be concern about access to reliable energy, which then the data center preventing access to reliable energy affordability of energy. But again, lots of other things too, like pollution, like those methane gas plants, for example. They're not only climate offenders, but they're known to produce really kind of localized, but poor air quality. That's bad for resident respiratory health. So yeah, those kinds of issues, you know, if we frame the problems just in terms of ecosystem or wattage or carbon emissions, then it really narrows the problem space, and then it narrows the solution space as well. I think that's important, right? Like then you're not thinking about bringing in social scientists like you and I, right to help bring visibility to some of the more complex and social dimensions of the problem.
Elizabeth Dubois: [00:07:51] Yeah, yeah. You focus in narrowly on one thing that you're measuring, and you lose sight of all of the other things that might be really meaningful in people's lives and in kind of like the development of high quality technologies or good governance of those technologies, those kinds of things.
Kelly Bronson: [00:08:12] Yeah. The power dimension, right. I think that's a key. And that's really why I find myself in a sociology program, because ultimately, and this came from actually, you know, when I was a lab bench scientist and I was doing genetic work, and then I started to get interested in public controversy around GMOs. I started to realize that when I tried to engage my fellow scientists in conversations about public controversy and some of the legal infrastructure, like licensing and patents on seeds and, you know, is that right or wrong? Some of those normative questions, it was like I had committed a breach of decorum, right? It was like, no, no, that's just anti-science, right? There just wasn't really a space to talk about those things. And I was really keen to. Yeah. Open up. I guess like the problem space in order to open up the solution space. Like, is it anti-science? Is that what farmers and consumers are concerned about? Are they worried about the science or are they worried about what's right? You know, access to food or to seeds or. So yeah, power ultimately became the thing that I was interested in. And I'm still interested in when I look at technology. And one way to make that kind of like environmental justice framing of AI and its impacts, I think really visible vis a vis power is to just ask the simple question like, okay, are you incorporating AI into your workflow? Yes. If yes, probably most of us are. But then would you want to live next to a hyperscale facility?
Elizabeth Dubois: [00:09:35] Yeah, absolutely. Do you want to have to endure the consequences of the choices that you're making exactly in your day to day life? Right. And that that really showcases where maybe decisions are being made that aren't for the good of the wider public. And you know, in a previous episode, I spoke with Jason Miller about how engineers are taught about ethics and like what? Like the cultural expectations are through the educational system and then out in the workforce. And, you know, that question of, well, would you want to live next to the tools that you're building or the things that are needed for the tools that you're building? He didn't quite frame it in that way, but he was talking a lot about this idea of, if we don't create a system where there are people who feel it is their responsibility to bring in this wider view of things, look from that sociotechnical lens, then we're never going to be able to solve any of those problems, because nobody's going to see it as their job to figure that out.
Kelly Bronson: [00:10:33] Mhm. Yeah, exactly. I guess that's what I mean by opening the kind of solution space and the set of issues, the set of solutions, and also therefore, who participates right and framing the problems and framing the solutions in the first place. Yeah, that's that's precisely it. Yeah. When I was a scientist, I started to think like, who put us in charge? Like what? And so, you know, I moved to the social sciences in part to figure out different, I would say, processes for thinking through widening the problem and solution space for consulting. Right. I don't think that all social scientists can or really should be made to speak for community. I don't know if you find this, but often like an expert panels or committee, sometimes that's a problem. You know, I'm like the only social scientist. And then I'm made to like speak for in my case, like farmers, right. And, and that's really problematic. And yet at the same time, what we learn in social science is in part tools for getting empirical evidence right on what matters to community. And starting with, I mean, coming back to this idea of measuring impacts. And I think one of the key things then is to start from really foundational questions in particular contexts, like what matters in this community. Maybe in some communities people say this is a trade off. I'm willing to to make, right? Maybe there are, you know, benefits to local community that offset some of the risks. But it's important to start with those conversations about values and about trade offs and risks and benefits and their distribution, right?
Elizabeth Dubois: [00:12:00] Come at it as a set of questions with curiosity for what the possible impacts are, what the kinds of things people might care about are, rather than like an, a narrow assumption, right? So for example, if we take what's right, so, you know, everyone's obsessed with watts, make sure that we have the lowest possible and then we're good to go. Or you meet a certain threshold and everything's fine. That is a really, really narrow way of looking at it. It doesn't kind of take this, let's ask these questions at the get go approach. Like, why is that the case? I guess is the question I have, like, how did we get here that we're just like so focused on watts and I'm saying what's, but like, it could be any of these narrower measures.
Kelly Bronson: [00:12:48] I think that's a great question. And, you know, it's so interesting to me is that kind of narrow framing, we might call it technologically determinist or techno-centric, or it happens on the flip side too, like with the positive, right, we see this all the time. It's like in the framing of benefits, this technology. At the beginning, you said people are really hopeful that AI is actually going to solve the climate crisis. Absolutely. Right. There's all this hope and grand claims being made. I work on AI in the context of food production and agriculture, and actually along the food value chain, which is a use case that hasn't really brought as much attention. But there's all sorts of promises, right? We've got really complex problems when it comes to food, ffood supply, right? Is food crises predicted for 2050 and hunger even in Canada, 1 in 5 Canadians, food insecure. We've got massive environmental problems, including climate, that affect food production. But also agriculture is one of the biggest contributors to climate change in Canada.
Elizabeth Dubois: [00:13:49] Yeah.
Kelly Bronson: [00:13:50] So yeah, that thing you said about like, why does it get framed so narrowly? I think the flip side happens too, right? In terms of solutions, not just problems that there's all sorts of techno-fix solutions proposed, including for big problems like climate change and food insecurity. I think in part because it's easier, right? It's much easier if you think about, okay, how are we going to frame this environmental problem and study it, right? And then make it amenable to policy is so much easier if we say, okay, we're going to just measure wattage, or if the solution is for something, as I think complex as global hunger is AI, right? And food production. So these are like these algorithms, decision support tools to help farmers. But in the global South, they can be simple like ICTs [Information and Communications Technologies] for farming apps for farmers. But a lot of that, there's a lot of enthusiasm from the World Bank. Bill and Melinda Gates Foundation through to nation states that AI is going to solve this problem of global hunger. It's so much easier, right, than saying, “oh, actually, it's a problem of distribution,” right? Currently, if we look at it globally, many nations over produce grains.
Kelly Bronson: [00:15:00] Many of them are not grains that are edible by humans. They're grown for animal feed, right? So that's in itself an interesting kind of moral. And so there's over production, but it's a matter of distribution. And that's the problem of border porosity or lack thereof. And war, right. And power and governance. And so, yeah, if you say that, if that's how you frame the problem as this very complex, multi-headed beast, that's so much harder a pitch than just, you know, AI is going to solve global hunger or AI is going to solve the climate crisis. And I think the pitch element to there's a power dimension there that like, you know, the finance environment around technologies like AI is such that the pitch needs to be clean, right? It needs to be a 42nd, however long it is. It's, you know, the sort of venture capital and futures and like the whole finance environment around emergent tech is such that the pitch has to be cleaned, it has to be compelling.
Elizabeth Dubois: [00:16:05] And grand.
Kelly Bronson: [00:16:06] And grand. Exactly. And it almost doesn't matter where the proof is, if there's proof or how it will bear out in 20 years, because, right, it's about the circulation of capital and to some degree, right. And I think also, you know, even if we think about you and I began the podcast talking about the headlines recently about, you know, the end of humanity and the quitting of scientists from anthropic. And it's all very depressing and scary. But even that, right, that like, I think people in industry call it P(doom), right? The probability of doom, even that is part of the hype because again, if we think about power, it's like if the pitch is clean and slick and it's technological, then who gets to be part of the solution? It's the engineers. It's the people who make money from the sale of these technologies that are putatively going to save all of us or save the world. And on the flip side, right, if you know, the probability of doom is quite high, who ends up being part of that solution space, right? It's like the tech bros who are then consulting, right. And there's nothing wrong with this. The [Canadian] AI Safety Institute, obviously there's a place for that. It's really important, you know, international collaboration among scientists, super important. I'm not denying that. But I think the hype on both sides really narrows the solution space.
Elizabeth Dubois: [00:17:26] Yeah. And then you layer on the technical level of having really narrow things to measure is just easier to say, yes, I've done it, or no, it hasn't been done. Whereas figuring out at what point have you sufficiently thought about all of the wider social implications of your tools and dealt with figuring out what the potential implications are going to be, and then figuring out how to do that as safely, as environmentally friendly as possible. Those are way harder questions to answer than like, did I hit a certain what threshold? And so you can kind of see all throughout the whole system for funding the tools, creating the tools, deploying the tools all along the way. It's way easier to do the thing that's less complex, even though the thing that's more complex is kind of what's needed most of the time.
Kelly Bronson: [00:18:24] Yeah. Sometimes I feel like a real Debbie Downer, though. Last year there this greenhouse Haven Greens made the news. And I don't mean by any means to pick on Haven Greens. I think the person, the guy who started the greenhouse, Jay Willmot, was really well intentioned, but it was really in the news because this is a greenhouse in southern Ontario, kind of a hot spot for indoor agriculture in Canada. That was a good pun, I suppose a hot spot. Well, I like a good pun. Yeah, it was a bit accidental. But he, you know, he said, we've started this totally autonomous greenhouse. No human hand is going to touch this lettuce. And this was meant to be a good thing. He said, this is it's using AI systems and robotics, and we're going to water cycle and heat cycle. And so it's sustainable, but also, right, it's, it means lettuce production all year round. And therefore we don't have to bring lettuce in from California. You know, a huge part of our greenhouse gas emissions in agrifood come from food transport, which is largely truck transport. So that's great. But, you know, then the CBC, I think, reached out to me and they're like, what do you think about this? Like, this is the future of sustainable lettuce. And I was like, oh, yes, maybe. But, you know, if you think about if you widen that space again, if you frame it differently, it's not just the autonomous greenhouse per se.
Kelly Bronson: [00:19:52] It's you. It's functioning right where Jay Willmot will not. They've designed Haven Green so that it heat cycles and water cycles to some degree. I think it's not a perfectly closed system currently. But what about, right, the running of the AI system coming back to data centers. And that's hugely environmentally costly. And that's kind of externalized. Economists would say or like left out of this picture, if you just say the greenhouse is perfectly closed loop and sustainable, what about the manufacture of that greenhouse? Right. What about the rare earth minerals that were mined that likely northern or Indigenous communities affected at the coal face of that, that mining or that extraction. And then. Right. What about again, just like, you know, continuing this thought experiment, what about the workers? So greenhouses are typically places in Canada and actually globally that employ migrant workers. There's all sorts of problems in terms of the treatment of workers in these greenhouse environments. But simply like displacing those workers with robots is not really an ethical solution to the complexity of the labor problems in indoor agriculture. And so, yeah, again, maybe, but I don't think this is going to solve [it] and then also it's lettuce, right?
Elizabeth Dubois: [00:21:08] Right.
Kelly Bronson: [00:21:08] It's not going to feed the world. We're not going to feed the world with lettuce. Hugely water intensive and not really high nutrient or nutrient dense. So yeah, it's not, it's not going to solve the 2050 food supply problem.
Elizabeth Dubois: [00:21:20] Yeah. But I mean, as like a little test balloon, you know, if you figure out really good ways for it to work. Like you could imagine it being a really great template for other kinds of produce and other kinds of things, but.
Kelly Bronson: [00:21:36] Especially in really remote, like, say, northern environments, right? Where food access is a problem. But again, if you think about the history behind a lack of access to healthy or culturally appropriate food and say, northern Canada, that has to do with colonialism again, right? It's like disruption of traditional foods and food ways and yeah, concentration of grocery retail in northern environments. And so at least part of the problem has to do with corporate power, has to do with historic and ongoing colonialism. And so it's not, but you're right. It's figuring out technological solutions. I'm not anti-tech, I think absolutely there's a place, but it's just important to take, I think, this fulsome view on both the solutions and the problems.
Elizabeth Dubois: [00:22:21] Yeah, yeah. And thinking about like the full life cycle of the tech. If you. If you have decided, okay, like the tech solution is at least part of what we want to do, you need to be thinking about it all the way along, not just, you know, okay, so a technology has been created and deployed. Good to go. You need to think about how it was created and where it's going to go after.
Kelly Bronson: [00:22:43] Absolutely. Yeah. That kind of life cycle analysis, you know, in the social sciences, I think some people talk about kind of infrastructure perspective, right? You have to think about not just the technology, but also everything surrounding it. You know, as I said earlier, like my jump from the lab bench to social sciences happened because I became interested in public controversy around GMOs. And then I realized that the way that the controversy was being framed was as like this controversy about science. But then when I started to just talk to farmers in this case who were really upset with patenting around seeds, I realized that the issues were about the infrastructure, in that case, the laws. Right. And I think the same thing is true with AI. I do a lot of work currently on not just AI on farms, but what happens in the context of farm AI with the data once it leaves the farm, who has access to it, right? Who gets compensated for the valuation of those data? And all of that then takes you into policy and governance and yeah, law.
Elizabeth Dubois: [00:23:45] Right? Yeah. There's so many, so many different pieces of this puzzle. And one of the ways that you bring those different pieces in is that sociological sociotechnical lens. Another way that I've seen these kinds of ideas start to be brought in is from the kind of government intergovernmental perspective, right? So the UN sustainability declarations like, and this call for broader definitions of what we need to be thinking about when we're thinking about sustainability. Do those existing frameworks translate to the AI context particularly well? Are we seeing that sort of idea of sustainability that's much broader than the AI conversation actually embed itself or get embedded in the AI discussions?
Kelly Bronson: [00:24:32] That's a really interesting question. I think that, yeah, on the one hand, obviously, you know, the Sustainable Development Goals are a big tent. I think policy folks say. Is that right? Am I getting that right? Right. Like that creates a big space that is perfect at the level of kind of international or supranational governance. Right? It broadens the space. The conversation brings in things like justice and power. You know, one of the hazards is that when you brought in, then there's a risk that things lose their precision, right? And so like, what do we mean by sustainability if it's writ large? Right? If it includes everything we've seen this with, I have a whole other area of research or used to on environmental assessment. So in 2019, the Canadian government passed a new piece of legislation, Impact Assessment Act 2019, and we moved in our environmental assessment. So these are like, you know, assessments that are tripped or legally mandated if the scale of a development project is really big. So like, you know, hydroelectric dams or pipelines. And in 2019, the government moved from Environmental Assessment to Impact Assessment and Impact Assessment Act 2019 includes a whole a wider consideration of what in the legislation is called, like social effects, right? There's a mandate for meaningful and this is the language in the law, public engagement. And yet it's kind of like, again, what do we mean by sustainable development? What do we mean by sustainability? Especially if you're including the the justice and the social aspects. It's difficult then to move from those widely agreed upon ideas or ideals into practice. Right. Like how do you actually measure social effects and how, you know, what constitutes meaningful public engagement? That's a really interesting question.
Kelly Bronson: [00:26:34] I've thought. So the framework, if it's broad that I use when thinking about food and tech, is food sovereignty. You know, in the 70s and 80s, the UN used to talk like in their documents, talks about food security because, you know, you're probably too young to know this. But in the 70s and 80s, there were massive food crises in sub-Saharan Africa, like massive famine. And so it was a crisis, right? So they talked about food insecurity and food crisis. And also, I think that security had a particular kind of traction in the 70s, when welfare state was like a more viable concept, I would say like kind of pre-really the entrenchment of neoliberalism. But then there was this move, and partly it was led by the Global South to talk about food sovereignty. I think part of the problem with food security is it was also focused largely on households. So the UN in their measuring tools and they talk about household food insecurity and food sovereignty. You know, groups of farmers from the Global South really pushed for a widening of the, again, problem space and to talk about food sovereignty and the definition of sovereignty versus security is really like, not just do people have access to food, but, you know, do people have access to culturally appropriate food and also access to grow their own food, producers were really brought into the mix. And yeah, and so I've actually thought that in the AI space, it would be interesting to talk about AI sovereignty. That would be a good analogous tool, but I haven't really seen that applied. Maybe that's some work to do in the future.
Elizabeth Dubois: [00:28:09] Yeah, I think I hear a lot about data sovereignty and we actually talk in this series? I've done an episode with a few other folks. Jonathan Dewar is one. He's the CEO of the First Nations Information Governance Center and Kumanan Wilson is another where in episodes on Indigenous governance principles and health policy, respectively. We come to this conversation about what data sovereignty looks like and why that's important in the context of building out and training AI tools. And so I think the idea of sovereignty is something that the AI space is grappling with and dealing with. And so I could see kind of like it's not a huge leap to then think about sovereignty in the way that you're kind of talking about it.
Kelly Bronson: [00:28:57] Okay. Yeah, exactly. And they're the stewards, right, of these great principles. So and yeah, in terms of data, absolutely. I mean, even think about the Canadian government and AI for All. And even like the early mandate letter under the Carney government, it was a lot about sovereignty and sovereign data for the Canadian AI ecosystem, also Canadian consumers and citizens. But it's sort of been confined to data, right? And again, kind of defined somewhat narrowly, not under the OCAP [Ownership, Control, Access, Possession] principles necessarily, but food sovereignty has been really ecumenical, interpreted and applied. And I just wonder if there's a way to do that across the AI life cycle, coming back to that lifecycle conversation or across the kind of broader AI and infrastructure.
Elizabeth Dubois: [00:29:46] Yeah, yeah, I think there's plenty more that we could think through and talk about there. And I imagine, I hope more of that will be done in the next little while, as so much of these AI development and governance conversations are being had. But we're coming up to time for today's episode. So I wanted to end with a question that I've been ending every episode with this season. From your perspective, what is being missed? Like what are policymakers missing? What's maybe not on their radar quite enough.
Kelly Bronson: [00:30:18] Mhm. And so in relation to AI, I think the Canadian government hasn't yet fully thought through or the environmental costs. So at least in AI, for all the strategy proposes an expansion of data centers. And even in this new principles document, the full especially taking that life cycle perspective, environmental costs of AI and its infrastructure, I think is underexamined. So AI for all really included this push for skilling, right? And yeah. And so it's like there was a kind of presupposition or I think there is in that document that AI is just an unquestioned or unexamined public good. And so the issue is like, how do we get more Canadians to incorporate AI? And how do we create skilling, kind of classic digital divide stuff, right? Bring people. Yeah. But I think that labor could be looked at in a more complex and should be looked at in a more complex way. So right. A more critical reading would be incorporating workers perspectives would be really important. How can we enlist workers in participating in AI governance? And maybe that's the big thing. Like for me, I think even in this new principles document that the government released September 3rd, a lot of people have been saying, okay, this is fine as a starting point. It's just principles. Absolutely. Now, there's a lot of work to do to move these principles into practice. But for me, the real challenge is that kind of process challenge, that governance challenge, like not just how are we going to do this, but who constitutes the we, right? Who's going to be convened to, to be a part of this conversation? Is it just going to be, you know, and, you know, I've told you this, but I was part of some of the consultations in advance of AI for all.
Kelly Bronson: [00:32:02] And I was really the lone social scientist. You know, I didn't see a wide consultation. There was some consultation with labor unions, but like, who's really at the coalface of change, especially if we think from environmental justice to come right back to our first conversation about unequal distribution of benefits and harms. Who's really going to bear the brunt? And how can we create governance processes that bring those people into the conversation around how to define strategic value? Coming back to what was in that principles document, right? They say develop, you know, data centers, for example, only if there's strategic value for Canadians. But of course, you can't include all Canadians who ought to be included in defining what strategic value is. What they say, you know, benefit should attach to local community. Okay. But, you know, there are multiple actors or stakeholders in local communities. And so how do you bring different values and needs and trade offs together? Right? When they're in tension, who gets included? Those kind of process based and governance questions, I think are the key ones that are missing. I see missing currently from the AI conversation in Canada.
Elizabeth Dubois: [00:33:10] Thank you. Yeah, that I mean, everything you're saying is making a lot of sense. And I'm seeing connections to other episodes in this season too, where, you know, if we're thinking about how we actually do governance well, we need to be thinking about the people who are impacted and having them brought into the conversations in a really meaningful way. So lots, lots to learn from that. But for now, I will thank you for your time. This has been a great conversation.
Kelly Bronson: [00:33:36] Thanks, Elizabeth.
Elizabeth Dubois: [00:33:37] Thanks for listening. But before you go, on December 3rd, we’re having a special live recording of the podcast on the University of Ottawa campus. I’m going to be chatting with the privacy commissioner of Canada. It is a free event, but registration is required so please go check the show notes for a link to that registration form. The event runs from 6-7:30pm; we’d love to see you there.
Alright, that was our episode on the environment and A.I. As always, we’ve got a bunch of links to different resources in the show notes and you can head over to polcommtech.com for annotated transcripts available in English and French. They link out to a ton more resources and further reading.
This season of Wonks and War Rooms is supported by University of Ottawa’s Vice President of Research and Innovation, my university research chair, the Centre for Law, Technology and Society and the Faculty of Arts at the University of Ottawa. I also want to acknowledge that I’m recording from the traditional and unceded territory of the Algonquin people, and I want to pay respect to the Algonquin people, acknowledging their longstanding relationship with this unceded territory.



Comments