There's a new AI on the block

[00:00:10] Karine Morin: What happens when the future arrives in your neighbourhood? For many Canadians, artificial intelligence feels abstract: it lives on our phones, in our workplace, or somewhere in the cloud.  

[00:00:23] But behind AI are enormous physical infrastructures called data centres that require land, energy, water, and that can have very real consequences for the communities around them. 

[00:00:35] As Canada pushes to lead the way on AI, communities across the country are beginning to ask what that ambition means for them. Who benefits? What might be lost? And what happens when residents want to have a say, or simply say no? 

[00:00:54] As a country, becoming an AI leader requires technology, power, place, and participation, so how can Canadians shape the future of AI being built around them, and what are we willing to compromise along the way? 

[00:01:05] Today, I'm joined by Renée Sieber, Associate Professor in the Department of Geography at McGill University, and together we'll explore these questions. My name is Karine Morin, and this is the Big Thinking Podcast. 

[00:01:26] Karine Morin: Dr. Sieber, it's a pleasure to have you as our guest today. I want to start with your background, which spans urban planning, geography, public administration, and you also have some hands-on experience in community activism.  

[00:01:41] So how did you come to this type of work, broadly speaking, but also how did this background shape your approach to citizen engagement relative to technologies and AI and data centres? 

[00:01:57] Renée Sieber: Oh, I did an undergraduate at an all-science college at Michigan State, and then I went off and bought a machine shop with people and started making toys, science fiction toys.  

[00:02:11] We were in an impoverished neighbourhood, and I got more active in community engagement. It is problematic cause here I'm a white woman getting engaged in a poor minority neighbourhood, but I did make a load of friends, and we talked a lot about unequal police protection, and we talked about ways that the technology could be used to support our activities, mostly through computerized mapping. 

[00:02:40] And that's what led me on a track to look at public engagement in computerized mapping. So I did that as part of my PhD, I looked at computerized mapping in the environmental movement at the time when the dominant technologies, both the hardware and the software, were outrageously expensive.  

[00:03:00] So the question was, can environmental organizations use this technology? Can they access this technology? And what do they do with this technology? And, increasingly I got involved in other kinds of technology, climate modeling, and then AI, just looking at public engagement in AI.  

[00:03:22] And AI is really cool because it's both used as a tool of public engagement, so it sits between the public and government sometimes. And also people may have feelings about the technology being used on them, with them, etc., etc. 

[00:03:40] Karine Morin: What a fascinating road you've been traveling so far. So we did want to talk more specifically about data centres, and yet it seems to me that we weren't thinking so much about big buildings appearing into these landscape. 

[00:03:56] We had been hearing a lot about the cloud. So how can you explain this sort of sudden appearance, it seems, of major infrastructure that relies on resources extensively; how do they start appearing and becoming problematic according to some? What can you tell us about this? 

[00:04:20] Renée Sieber: Data centres existed before generative AI, and we're explicitly talking about generative AI here, of chatbots, of ChatGPT and Claude. They were used for Netflix, the data has to be stored somewhere, the movies you're watching.  

[00:04:35] They have been used for Bitcoin and crypto, and, that's where we started to see some of the problems about energy usage and water usage with those kinds of data centres.  

[00:04:48] Now, because data centres have become the way to advance concepts like digital sovereignty, or they're being promoted as ways to create jobs in poor areas, we have data centres popping up everywhere. 

[00:05:05] So that, and what is a data centre? A data centre is an enormous box, and it has a lot of GPUs. It has a lot of AI processing units, or graphic processing units - GPU - and AI processing units now, Silica.  

[00:05:27] So all of a sudden it's like we've got to build data centres everywhere. And then people say, "Wait a minute, what is this data centre doing in my neighborhood? And why is it being powered by gas turbines?"  

[00:05:41] Or, "What's happening to my electricity bills?" or, "What's happening to my, the water I used to, like, just take for granted?" 

Karine Morin: So you've just touched on a few important points that I wanted to pick up on. Data centres in Canada, it seems to me that I wasn't reading a whole lot of headlines about this just a year ago or so, and suddenly I seem to be reading a headline about this weekly. What would've been driving Canadians or Canada to have data centres?  

[00:06:11] Renée Sieber: This worry that everything AI, which is inevitable and the arms race, is being controlled outside of Canada, it's being controlled predominantly in the United States. There's also China in the mix. 

[00:06:28] And there is a concern amongst middle powers, like Canada, that we don't want to be dominated by the mechanisms of AI, both the way that generative AI subtly shapes the narratives that we grow accustomed to 

[00:06:49] And we don't even think about it when we ask ChatGPT, "Make a party flyer for my kid's birthday," or, "I've got these ingredients in my fridge. What can I make for dinner?"  

[00:07:00] So we don't even think about how that is going to the US, and data may be stored in the US, and there may be other uses of your data, for example, by data brokers. But there is this concern that there is this overpowering force, and we know a lot about generative AI's ability to shape how we think about the world. 

[00:07:26] So I was at the Paris AI Action Summit, and all of a sudden with issues like the golden child of AI in France is Mistral. All of a sudden, Mistral, this large language model, was sold to Microsoft. 

[00:07:47] So early on you had people, in France, in the European Union, even before that, worrying about, "How much capacity do I have in my country to do my version of all of these foundation models, all of these classification models?"  

[00:08:07] And interestingly enough, Canada punches above its weight in terms of AI. Because of our three research institutes, and predominantly because of Mila, the Quebec Machine Learning Institute, and Yoshua Bengio, we have a presence nationally, from Montreal nationally, we have a presence, and globally too, as well as other countries, in how we think about AI, for example, in terms of AI safety, so that matters too. 

[00:08:41] Karine Morin: So then, if one is a municipal politician, or perhaps even a provincial politician, and we're thinking of sovereignty and data sovereignty, it may seem like we'll be looking to the possibility of benefiting economically from data centres, and we hear that there can be construction jobs, and we hear that there can be longer term economic benefits. 

[00:09:06] Is there somewhat of a case to be made for data centres here in Canada, or do you see in the balance more that sort of weighs against than that weighs for? 

[00:09:20] Renée Sieber: The interesting thing about being in the social sciences as opposed to being in STEM is we often don't have to make a case for data centres. But you can see where some of this is going. 

[00:09:33] Municipal officials say, "I have this unused piece of land, and I need jobs." And the promotion of these data centres and the wealth that it will bring and the economic prosperity that it will bring is kind of overwhelming.  

[00:09:48] So if you're looking at your next election or you're not being so cynical and you're actually, you want to serve your public, then you say, "I don't have anything else," or, you know, "This looks good, and this allows me to be a part of the AI economy, the AI ecosystem if we have a local data centre." 

[00:10:09] But there is plenty of empirical evidence now about how the economic benefits do not accrue with data centres. So you may get a few construction jobs, they may or may not be local people building this.  

[00:10:27] And then you build it, gets populated with virtual machines, with physical boxes, and it doesn't take many people to keep this thing working. Over and above that, it requires an enormous amount of electricity. 

[00:10:47] Where is that coming from? You know, we want to be environmental stewards, and yet the quick and dirty way, and indeed it is dirty, to power data centres is to build gas-powered plants next to a data centre.  

[00:11:08] Now, gas-powered plants are very noisy. It's natural gas. Very, very polluting. So you might have a problem if this data centre is right next door to you, and that's a big part of the proliferation and the pushback to data centres is because they are right next door. 

[00:11:29] Karine Morin: So we'll just come to what you were sort of putting in the balance here, the economic case. You pointed to some shortcomings to it, but what else would you be watching for when there's a proposal for a data centre to perhaps come into your municipality?  

[00:11:45] What do you see in terms of transparency, if at all, as to what might be other impacts on a community, or precisely lack of transparency on what might be some other impacts? 

[00:11:58] And I think we're talking about environmental, but do you see others in terms of the political realm, social realm, how community can organize around a proposal before it might be built? 

[00:12:11] Renée Sieber: It would be nice to know if your council people signed NDAs -nondisclosure agreements - which is often a part of data centre construction. And you talked about, you asked about transparency.  

[00:12:28] There's a big issue. Yeah, there, that's one thing that we could use transparency, but transparency is always a means to an end, right? You need the means attached to some outcome, accountability.  

[00:12:41] Do you have the ability to get rid of a politician? Citizens have the ability to push back against those kinds of decisions because it can be very tempting for a council member to be swayed by the proposed benefits of these data centres.  

[00:13:04] Accompanying that, and speaking once again about transparency, is the incredible creative financing underlying the construction, paying for these data centres, and also leasing arrangements for the data centres. 

[00:13:20] There's a big question, first of all, of whether municipalities are building these data centres with the help of a construction company in anticipation of an occupant, and there's no occupant at the moment, so they're therefore competing with other municipalities, and they may lose if their data centre is too small, as people go towards hyperscale data centres. 

[00:13:49] So there's one thing about who, who's actually financing this, and how durable are those finances if one partner goes under, for example, does the whole data centre construction stop?  

[00:14:06] And it's also these big tech companies are creating creative leasing arrangements so that you often do not know who is an occupant, a tenant of your data centre. We don't know if Flock will be storing the CCTV, the surveillance cameras the police are using, if they're an occupant of the data centre.  

[00:14:33] We don't know if ICE and other US agencies are occupants of data centres. You can see why this could be very important, because most of us live really close to the border with the United States, and the United States and US companies have a vested interest in having endogenous data centres here. 

[00:14:58] Karine Morin: May I ask, if you are raising all of these concerns, it does seem that in some instances there are communities that have organized and have been able to really push back, and maybe there are some Canadian examples or otherwise that you'd like to highlight of they learned enough soon enough to take action. 

[00:15:16] How is that happening? What are you seeing that gives us a sense that maybe there's a way to get to our politicians before the deal has been signed? 

[00:15:25] Renée Sieber: One notable example was the city of Sherbrooke, Québec. Their data centre was built: it was very noisy, it was on one side of a river, close access to water. Now, the place it was built was an industrial zone, but it was across the river from a well-to-do neighborhood.  

[00:15:50] And the neighbors complained, and they were able to shut down the data centre. Now, of course, that also speaks to, and this is where we shine, and we need to be a party to the conversation because we can talk about the political economy, we can talk about empowerment and disempowerment and differential political power of communities and how that plays out with race, ethnicity, Indigenous populations. 

[00:16:22] We know of a case in Alberta now that they're fighting Wonder Valley, Kevin O'Leary, he has a US data centre plan and a Canadian data centre plan. In the Canadian data centre example, the Indigenous people won their first court case against the construction of the data centre where they said, "No, you're not absolved of doing an environmental impact assessment with us, you have to." 

[00:16:53] Which also suggests that there's a long road, there's a long courtroom road to successfully resisting data centres and successfully resisting AI. But there are many different kinds of initiatives and efforts to resist AI, not just data centres. 

[00:17:15] Karine Morin: That brings me to, we've purposely been focusing on data centres, but we've seen broadly speaking, relative to some AI policies, that there's been government consultation. 

[00:17:27] And here I'm curious to hear how you see those processes unfolding, whether you've seen instances where citizens seem to be heard, or whether you're concerned that more often than not, these consultations are going to be tokenistic.  

[00:17:43] And how can you sort of encourage those concerned to take up those opportunities when there is consultation? 

[00:17:49] Renée Sieber: First of all, remind me to talk about the jurisdictional issue. But let me then say that Canada has a poor track record in terms of engaging the public around AI and generally.  

[00:18:03] You rarely hear in the United States that there is a public consultation. People talk about public engagement, civic engagement, public participation, they do not talk about public consultation.  

[00:18:18] And indeed, if you look at Sherry Arnstein's Ladder of Citizen Engagement, which is perhaps the most cited publication in urban planning, and definitely the most cited diagrammatic way of thinking about the intensity of public participation, and this was done in 1969, and we see it now in the International Association of Public Participation, they built on top of that ladder, that public consultation was considered tokenistic. 

[00:18:53] Because a lot of it was just informing the public, in this case, of how good AI is, so you should socially accept AI. Or it is, "Okay, we heard you. Now go, because we're technocrats and we know what's best for you."  

[00:19:11] Public consultation always needs, once again, that accountability piece. That piece that says, if we say no as a public to certain things, then you don't do it. So of course then the question is, how do you balance when you argue that you really do need some of it with a public that says no? 

[00:19:40] But we are not doing a good job in listening to people and doing any balancing. We are just going full speed ahead and saying, "Look, sorry we heard you, but we got to build these things for digital sovereignty." 

[00:19:53] Karine Morin: Now, is this where I put in, is there a caveat about jurisdictions? Are there some might be doing a little better than others? 

[00:20:00] Any shining example of getting the public to engage and understand and weigh the pros and cons? There are trade-offs all over that we have to make in personal lives and public lives. 

[00:20:11] Renée Sieber: A couple of things. Edmonton is great. And Edmonton has done a lot to engage the public. They're the gold standard for municipal AI development in the country, and they've also had a long history of citizen engagement, too. 

[00:20:31] You know, part of this is that a lot of people in this space are risk-averse, and they've been burned by public engagement. And public engagement takes a long time, and it's messy, and it's like, "No, but we have to build this data centre right now. We have to build this algorithm right now."  

[00:20:51] Another part of jurisdiction is level. So a lot of the public consultation, citizen assemblies, citizen juries, they happen at national levels. And in Canada, we haven't sufficiently recognized how much AI is going on municipally and locally, and how much public participation needs to be heeded locally. 

[00:21:20] Karine Morin: Let's find something a little bit more optimistic. If governments are to respect citizens' refusal or hesitancy relative to AI technology and data centres, how would you like this governance to be reframed, and what is status quo and not quite going the way you would like it to go, as I hear you speak of this, how should we reset? 'Cause it's not disappearing. 

[00:21:43] Renée Sieber: I challenge you on that because I think that feeds into the assumption that AI is inevitable and that AI is unstoppable, and that we must embrace or we will invariably fall behind and be a backwater.  

[00:22:06] And I don't think that's the case, and there are other people who argue that's not the case. So, I think one of the things that we can do, especially in the humanities, is look at the metaphors and look at the assumptions of AI. 

[00:22:23] But AI governance in this period is really fraught because people in the AI governance space thought they had it all figured out, and they built all of these models. Now we have AI governance maturation models.  

[00:22:40] But then all of a sudden, the public said, "Whoa, hold my beer." And now governments are presented with enormous opposition to AI. And how that gets built into governance in a meaningful way, how people are meaningfully listened to, is, to me, the currency of the day. 

[00:23:06] Now, governments, especially national governments, can railroad people and can say, "Look, we don't really care because we have this imperative. We have this economic imperative that we must do this or else." And, you know, it's up to us as researchers to figure out what the or else is, and whether the or else is as dire as the, unfortunately, the tech bros think it is. 

[00:23:29] Karine Morin: You're just about to make explicit a little bit about what would be a call to action to your humanities and social science colleagues to sort of take on at this time, and what sort of further insights could come about?  

[00:23:43] Where do you see the role of humanities and social sciences? You've alluded to it a little bit, but what more do you think could come from our community at this time? 

[00:23:49] Renée Sieber: Well, I think it's very important for us to talk about AI governance, to intrude upon those policy spaces, and to say, "Wait a minute, we need to think about blank and blank. We need to think about the racial, the economic implications of this. We need to think about the societal implications of this." 

[00:24:11] Sure, you can, in a municipality, you can be sold this idea that AI will make your life convenient, but what's the downside of that? What is the cost of that? And that's where we can play a huge role.  

[00:24:25] So political power, obviously, a huge, huge issue, issue. Political economy, huge issue. In terms of humanities, the assumptions, the metaphors, the ethics of it. And I would add to that, because I'm an applied AI ethicist, that I ask the question of whether we as researchers have a moral obligation to act.  

[00:24:51] And that can be very uncomfortable, especially in the social sciences, probably less so in the humanities. But we've already been faced with this vis-à-vis climate change, and we've had to ask ourselves what is our own, as researchers, our own moral obligation not to just critique, but to actually act and support and advocate for. 

[00:25:16] So AI presents yet another transformation, another crisis, and how will we respond? I think is a huge challenge for us. I don't have a good answer to how we balance that, but there we are. 

[00:25:33] Karine Morin: That has covered a lot of ground, and I just want to give you that opportunity to maybe leave our audience with one big thought of the day regarding our collective AI future. 

[00:25:44] Renée Sieber: We're supposed to be a risk-averse country. What happened to that? Why aren't we taking the slow road here? And I think that whether it's a municipality or whether it's a province or whether it's the federal government, we have the chops to do AI right now, but why do we have to destroy our environment? 

[00:26:06] Why do we have to immiserate our publics to do this? I talked about things besides data centres. Why do we just accept that facial recognition technology will be a part of our lives, and it will be surveilled from morning to night? And of course, surveillance is differentially allocated to people. So if you're homeless, unhoused, if you're a minority, you suffer much more surveillance than I do or you do. 

[00:26:39] So I think that's really important. We need to look at the way that algorithms are just dominating every part of our lives. Think about all the conferences that we can't get people to attend because they can't get visas into our country, and because the federal government is using AI to determine who gets a visa and who doesn't. 

[00:27:02] So we can be more risk-averse, and we got lots of people here who need jobs, so we can do more human in the loop. 

[00:27:10] Karine Morin: Well, it's not every day that we hear risk aversion as a bit of a compliment, but we take that and certainly stay tuned. I think there are some strong forces, and we'll have to keep watching this space to see what happens. 

[00:27:25] Dr. Sieber, thank you so much for this lovely tour of the landscape of AI data centres and more. 

[00:27:33] Renée Sieber: Thank you so much. I appreciate it. 

[00:27:42] Karine Morin: Thank you for listening to the Big Thinking Podcast, and thank you to my guest, Renée Sieber. I also want to thank our friends and partners at the Social Sciences and Humanities Research Council, whose support helps make this podcast possible. 

[00:27:54] Finally, thank you to CitedMedia for their support in producing the Big Thinking Podcast. New episodes are released each month, so be sure to follow us on your favourite streaming platform. À la prochaine!