Some folks recently asked me what I thought about AI in Brighton, particularly in the wake of news of a proposal for a new AI data centre in Trenton.
I have a lot of thoughts. And links for further reading/viewing. Strap in, this is a doozy.
First, I’ll offer the seemingly obligatory nod to the powerful potential of AI: based on what we know of it, it’s capable of incredible things, and we’re just getting started with a technology that could change almost everything. There are wonderful people using AI to do wonderful things, and I want to recognize and celebrate those people–people who, suddenly given unprecedented power, are using that power for good. I want to offer a nod to a resident I met the other night who introduced himself to me and then added “I work with AI, don’t hate me.” I don’t, I promise!
But that said, there are a lot of objections to the way AI is being developed, used, and regulated (or rather, not regulated). I’ll give you a rundown of the issues, with links to sources and videos, and a sprinkling of my own opinions, starting with the existential concerns and moving down to practical responses.
Will AI Kill Us All?
It would be lovely to offer a categorical “no” to this question, but it does seem to be a major concern held by many of the people who are actually developing AI technology, so it’s something we should take seriously. Well, semi-seriously: the following video is a little bit goofy, which is one of the reasons I chose it. Sometimes the best way to process the news is through humour.
Dave Jorgenson, the maker of this video, used to be a journalist for the Washington Post. Here he interviews a journalist from the Atlantic, but he also references a lot of sources, and he lists those sources in the video description (so click through to YouTube if you want to see them). The following videos build on ideas that are brought up here, but if you have limited time start with this one and skip the next two.
There are two things I want to highlight from this video. First, that the reason AI companies (and lawmakers, including in Canada) argue against regulating AI is because they see the development of this technology as a race against China. Ted Cruz saying “if there’s going to be killer robots, I’d rather they be American killer robots than Chinese killer robots” is chilling, particularly to anyone who isn’t American. That’s the same logic that fuelled the nuclear arms race: if someone is going to blow up the world, it should be us, not them. I’m not convinced by such logic: it smacks of paranoia and bully-logic, showing a lot of insecurity and machismo that leaders should be above. And it’s also not particularly based in fact: China is regulating the development of AI, and it hasn’t slowed them down very much, which is the subject of this interview:
So no, I don’t believe that we have to race ahead into technology we barely understand just to ensure that we get there before China. Not least because the “we” that would get there before China is not actually you and I, it’s a handful of American billionaires who are famously building bunkers and rocket ships to prepare for the end of the world:
I’ll say here that I don’t know the Breaking Points podcast, but Astra Taylor is a Massey lecturer, philosopher and organizer, and Naomi Klein is a journalist known for cutting through the zeitgeist, and their books are always excellent. Their latest book is of particular interest to me, showing the connection between the people who are developing AI technologies and (religious or non-religious) apocalypticism.
Empowering Bad Actors
The second thing I want to point out from Dave Jorgenson’s video above is the comment that if AI is going to kill us it’s far less likely to be killer robots taking over the world than it is the idea that someone with ill intent is going to use the power of AI for bad purposes. That’s already happening: the number of scam calls and emails I’ve been receiving have gone up considerably lately, but what’s even more concerning is that they’re finally becoming convincing. Even the Municipality of Brighton was targeted by email scams this summer. It almost makes me miss the early days of the internet, when my spam folder was full of emails from supposed Nigerian princes who needed me to hold their fortune in trust; now, the emails are plausible, tailored to the people they’re targeting.
Worse, the spread of disinformation is rapidly accelerating. Remember when Russian interference in American elections was such a big deal in 2016? We all got a crash course in “psy-ops” and “bot farms” and the fact that you can’t trust everything you read on the internet, and the way that disinformation can turn us against our neighbours. Public trust has steadily decreased since then, and as we increasingly can’t trust what we see on the internet, that lowering of trust is accelerating to the point where we all have a “liar’s dividend”, the ability to deny anything and everything if it doesn’t suit our own narrative – even, or maybe especially, when we’re lying to ourselves:
This video does an excellent job of exploring how AI slop degrades the internet and our ability to trust one another or even ourselves. If someone wants to start a panic, or a war, that’s really not all that hard anymore.
In a Canadian example that would be hilarious if it wasn’t so sad, the Albertan independence movement is being stoked by fake videos and websites promoting separatism coming from outside the country; and at the same time, others are using AI-made fake websites promoting disinformation about the referendum to confuse separatists:
But even beyond making us turn on each other, what if someone just asked Chat GPT how to build a bomb? There’s supposed to be protections in place to prevent a chatbot from answering that question, but those protections are surprisingly easy to get around. Put differently, it’s very easy to fool AI chatbots into telling you things you’re not supposed to know, including things like the location of US warships.
AI Psychosis and Chatbots
On a personal level, AI chatbots like ChatGPT are known to be “sychophantic”, which is to say that they’re designed to always tell you that you’re asking good questions. They’re programmed to be so friendly that people are increasingly seeing these chatbots as their friends, or…worse, as explained very well by John Oliver (note his HBO-level language, and maybe don’t watch this one on full volume at work or around kids):
Lots to unpack in this video, but note especially how easily they got Grok to tell them how to build a pipe bomb (see above point); how quickly many chatbots become flirty or sexualized, even with children; and how quickly the sycophancy of some chatbots can induce psychosis in some people.
It’s very troubling to see stories of AI chatbots counselling people about their health or mental health. The stories in which chatbots counselled teens to end their own lives are chilling. What’s particularly worrying is that the most common use of AI in 2026 is for advice, companionship, or personal support. People are using AI (up to 1/3 of its common usage) because they are lonely and vulnerable. That breaks my heart on its own, but then to hear about the times it ends up making things worse for them really hurts.
Poisoning the Well
Writing, and printing, and the internet are all technologies that help us externalize our knowledge into the world around us. A few thousand years ago, some sages complained that writing was dulling their students’ minds, because they no longer needed to remember anything. Now we all carry the internet in our pockets, and so never really need to learn anything at all so long as we can look it up. Right?
That’s a bit of a problem in itself, and one of AI chatbots’ primary functions is to augment web searches to summarize more content for us. So we don’t even need to sort the data that our search has turned up. And since we know that tech companies have a history of giving us worse content in our searches if it will improve their revenues (see below for more on that), we should be cautious about anything that appears in the AI Summary that now tops our search results: it might be correct, or it might be paying dividends to the search company.
But worse than that, I’m concerned about what is actually true. Like the Hasan Minhaj video above pointed out, the amount of AI slop out there is enough to make us not want to believe our eyes. But at least we can search the internet to verify things, right?
Maybe:
The Kurzgesagt video here has been troubling me since they released it almost a year ago. They normally put dozens of hours of research into a story, and they tried using AI to augment that research, but they found that AI hallucinations had corrupted the research. Basically, the AI chatbot invented some “facts” that they couldn’t otherwise verify, even when they brought those “facts” to real experts in the field. Then they found that other AIs were reproducing those fake claims, and citing the first AI as a source. Now this completely made up information is all over the internet, and to anyone who doesn’t already know that it’s fake, it looks legit.
It’s bad enough that using AI can sometimes provide you with fake information; but if you can’t even independently verify something, what does that mean for the validity of research more broadly? How long before the internet, where we have collectively stored nearly all of human knowledge, is inherently untrustworthy, even if you’re not using AI to search it?
The Tech Business Model
Aside from the existential and personal threats of AI, the economic situation in this tech boom is also very troubling, in large part due to the business model tech companies have been using since before generative AI was a thing. The technical term for this, believe it or not, is “enshittification”, coined by Canadian polymath Cory Doctorow. In this interview with Ezra Klein and Tim Wu (another Canadian tech writer, who has an older but still excellent book that predicted this, currently on the library sale rack in the lobby!), Doctorow gives a quick rundown of what enshittification is:
In short, the business model of tech companies is about controlling everything that’s on the internet, pushing out all competition, and then optimizing their profits even though it degrades the quality of their own services. Google knowingly made their search feature worse so that we would click through more pages, giving them more chances for ads; Amazon did likewise to prioritize its own products over products that would be more relevant to the person searching; BMW deactivated heated seats until drivers paid a monthly subscription; and more.
Facebook (and parent company Meta, which also owns Instagram and several other social media and tech platforms) is the most obvious case of enshittification: what started as a social network where we only see what our friends post is now an almost constant stream of ads and promoted content, much of which is known to contribute to anxiety, depression, and eating disorders, especially among youth. Meta knew this: in 2021 Frances Haugen leaked their internal studies, which proved a link between Facebook and Instagram content and these mental health issues, as well as internal communications emphasizing the importance of targeting youth audiences for the company’s bottom line. Finally, in 2026 Meta lost a lawsuit and was forced to pay $18Billion in damages, and has finally agreed to change some of the features of their services that led to this harm that they’ve long known they were causing.
While we’re on the topic of Meta, rather than pay Canadian news sources any royalties for news articles that get passed around on Facebook, in 2023 Meta determined that they would simply ban any news sources on their platforms. So to my friends who say that they get their news on Facebook: if it appears on Facebook, then it has been determined to not be news, because all legitimate Canadian news sources are banned on the platform.
Facebook does not protect its users’ data, it actively collects and sells it to advertisers and political communications firms. It also uses user data to train its AI models.
One of my many moral objections to the business model of AI platforms is that they were largely trained by feeding them with endless amounts of information that was taken from people without their knowledge or consent. And not just your social media feeds, either; millions of books, songs, films, and more, all without compensation even for copyrighted works. Remember the Hollywood writers’ strike of 2023? It started because studios were already looking to use AIs that had been fed every movie script they could get their hands on to replace real writers; and the actors’ guild got involved because their likenesses were being used, whether they were licensed or not, to replace their own acting.
Not all of this section has been specifically about AI, but the companies building AI are often the very same as the ones I’ve talked about here (Meta, Google, Amazon, X), and the others use similar models that are demonstrably harmful and exploitative of their own users. That’s concerning to me.
Jobs Jobs Jobs
We very often hear about the potential powerful use cases of Artificial Intelligence: it’s going to cure cancer, solve climate change, etc. All good things, and things that might even be so good that it justifies taking some risks, so good it might balance out all of the harms we’re already seeing. But the primary use case is to change the way that we work.
Experts and commentators will talk ad nauseum about what kinds of jobs will be killed by AI or other automation, and how when automation kills one kind of job it also usually creates another kind of job, so that in the long run we have just as many jobs but more of them are of higher quality. That might be the case, but not necessarily: in my opinion the quality of a job depends on how well it treats people, and I have two concerns here.
The first concern is that the jobs that AI seems to most quickly be stealing from real people are creative jobs: the books, songs, and films that I mentioned above. A significant portion of Spotify’s most streamed songs these days were made by AI, and their algorithms prioritize paid content over any question of whether it was created by humans. The book publishing industry is in chaos right now as questions of original authorship are disrupting the industry and rumours of AI-written books are tanking new authors’ careers before they even start. I’m all for AI (and other automation) taking over certain jobs, but creative jobs? No thanks.
The second concern is the way that AI augments jobs. Many experts and commentators will (correctly) say that AI doesn’t necessarily kill jobs, but it changes the way that we do them, making some things easier and more efficient, allowing us to direct our attention to other parts of the job. That sounds great, but it really can go two ways, and for that explanation we go back to Cory Doctorow:
Doctorow’s latest book (currently a bestseller, alongside his other book Enshittification) is called The Reverse Centaur’s Guide to Life After AI, and it asks the question: are we using AI, or is AI using us?
When we look at the way that companies are actually using AI, very often it isn’t in a way that simplifies some tasks to allow the human to focus on better work. Very often it’s the other way around: the human is there to support the AI, even though the AI isn’t always as good at the job as the human is. Think about Amazon: yes, the whole user experience is automated, and much of the packing and delivery system is automated, but there are some aspects of that process that require humans, and those humans are being run by the machines rather than the other way around.
Markets and Bubbles
The impacts of the AI race on the stock markets has also been very troubling. Right now there are seven tech companies that are collectively carrying the stock market. When you look at the economy as a whole it looks great; if you break it down, though, not so much. If you look at the labour and productivity side of the market, things are pretty mid, but the soaring stock market more than makes up the difference. But if you look at the stock market more closely, it looks pretty mid too, other than the seven companies at the top. And when you look at those seven companies at the top, they’re all AI companies, and they don’t look very good beyond a first blush either.
This chart has been making the rounds on the internet, and for good reason. The deals between these tech companies amount to them all passing around the same few hundred billion dollars, but each of them counts it as revenue. So on paper they’re valued as being worth as much as a small- to mid-sized nation, but in reality they don’t make money.
Because these AI models require so much computing power, they also require a ton of infrastructure, to the point where there are proposed data centres that are the size of Manhattan, and some are proposing putting them into space. Either way, that takes a lot of money. These companies have incredible amounts of revenue, because their stocks are so hot, but their business model (see above) requires more or less giving their product away until they can get us all hooked on it, and then they’ll start charging more for it. So right now they lose money on every single chatbot query, and they’re all hoping to survive long enough, and be on top of the competition, until the moment when this becomes profitable. Until then, they’re inflating their value and hiding their debt.
Do you remember the 2008 financial crisis? Our integrated global economy can’t handle a bubble of this size bursting. I find this troubling, and would hate to see it have major impacts in our little town.
Data Centres
The neighbour who introduced himself to me as someone who uses AI for a living seems like a great guy doing good work that makes the world a better place. He also doesn’t want to see an AI data centre here in Brighton, for the same reason that nobody wants one in their town:
- They guzzle energy. A data centre is basically a big building full of computer servers, all running 24/7. The energy draw can be enough that places that don’t have a surplus of energy are required to create new energy plants just to service them. And since that energy often comes from fossil fuels, it packs a big carbon footprint too. And when they can’t get enough energy from the grid, they use backup diesel generators, which of course only add to the emissions, and the noise.
- They’re noisy. I remember being in the computer lab at my high school, hearing 30 computer fans in one room, a constant drone in the background that you eventually stop noticing until they’re powered down, and then suddenly the silence becomes deafening. A data centre is thousands of computers, and creates noise similar to a major highway. Inside the building they hit around 96 decibels, loud enough to cause hearing loss. The proposed data centre in Quinte West had to do significant noise studies, and those studies compared the sound to the nearby highway 401.
- They’re hot, and thirsty. At a time when we are preparing for record-breaking heat waves to become a regular fact of life, these computers run very very hot and must constantly be cooled. While the proposed facility in Quinte West will be air-cooled, water-cooled facilities are known for being water-guzzling, using up significant amounts of the local water supply and drowning the wastewater treatment plants with polluted hot water.
- They take a major chunk of local resources and infrastructure, but don’t employ many people. They also don’t produce a local product that supports local businesses and residents; the money they make largely leaves the community.
For all of those reasons, data centres are a significant polluter and hazard that provides very little benefit to the host community. It’s no wonder that people are lining up to oppose them.
How is AI Regulated in Canada?
Poorly. The Government of Canada regulates AI primarily through a voluntary code of conduct, which is to say, it asks AI companies to please not be evil, but won’t do anything about it if they are. We do have an AI Strategy, but it’s mostly concerned with making sure that we don’t fall behind in the race, and that we aren’t dependent on American companies. That’s good, but it does little to address the concerns raised above.
The Government of Ontario is currently exploring ways to regulate AI data centres, and the current proposal would require a proponent to prove that their data centre supports the host community economically before it can be hooked up to the power grid. That’s great, but also doesn’t address most of the issues listed above.
Why I Don’t Use AI
Despite all of that, as I said, there are some people who are using the power of AI for good. But there are a few reasons why I don’t think that’s sustainable. Here’s why I think so, and some other reasons I don’t personally use AI.
First, AI is frequently wrong. As noted above, sometimes it “hallucinates”, just making stuff up in ways that you have to be careful to spot. I’ve seen AI use guidelines, including from our local school board, that really emphasize that we need to always check over the product of our AI queries to ensure that they’re accurate. I see two major problems with that, one being that it’s rarely worth my time to get AI to write something for me if I still need to fact-check it; and the other being that that assumes that I know enough to spot its errors! If someone is using AI and they aren’t already a subject matter expert in the thing they’re asking AI to summarize or write about for them, then they can’t fact-check it. They lack the skills or background knowledge to do so. While most people who are using AI today have years or even decades of knowledge and experience to draw on, and so can spot a hallucination or error, what happens when my kids start using AI today without any of that experience and knowledge base? Every use of it becomes a shot in the dark, hoping that it will produce something true and not being able to tell if it doesn’t.
Second, getting AI to produce something for us often undermines the point of producing it. I don’t write blogs just to inform you lovely people, although that’s part of it; I write blogs because it helps me to understand and frame my own thoughts on the subject, and gives me something to refer back to. But honestly, I rarely refer back to it for information purposes, because the act of writing it really settled that information into my brain in ways that are long-lasting and transformative. The act of writing hasn’t just informed me, it has made me into a person who knows that thing. I wouldn’t give that up, it would defeat the point. (In the same way, I used to love to take notes in class, and couldn’t understand people who recorded the class; I would never refer back to my notes, but rather, note-taking made the lesson sink in to the point where I didn’t need to refer back to the class. And my first blog was to help me understand the work of German theologian Dietrich Bonhoeffer, and I didn’t care if anyone ever read it! Somehow, it still gets more traffic than this municipal blog.)
People don’t make art just for the sake of seeing it or selling it, but AIs do. For humans, the joy and transformation is in the making. Journey before destination.
Third, I don’t want to support the business model of the AI companies. It’s openly exploitative and abusive to its users, even if it won’t end the world as we know it.
Fourth, I don’t have much of a use case for it. It can’t do anything that I want to do that I can’t or wouldn’t want to do for myself. Call me when it can fix my dishwasher, and please don’t take policymaking and community engagement away from me. There are people who do have a use case for it, but their use is nothing like what gets marketed (constantly) to us, which is that it will help us write emails. If you can’t be bothered to write me an email, I probably can’t be bothered to read it, and the absurdity of AIs responding to AIs, or AIs grading the work of AIs, has already come to a head:
Fifth, and I think this is maybe the most relevant point for the municipality, I don’t use it because I don’t want to damage the trust that people have in me. I don’t want anyone to have reason to question if I really said that, or if I actually know what I’m talking about. I don’t want to use the “liar’s dividend”, or have anyone use it about me. I want to be responsible for what I do and say, not for whether or not I could spot a falsehood in a document that I prompted but didn’t produce.
And for this last reason more than anything, I want to ensure that the Municipality of Brighton is careful in the way that we use AI.
What Does This Mean for Brighton?
In the past week or so I’ve had some conversations with a few people, Councillor Wright among them, about what AI means for Brighton, particularly knowing there’s likely to be a data centre in Trenton in the near future. Bobbi and I tossed around some ideas for a Notice of Motion, but as regular readers of the blog know, a Notice of Motion takes two meetings to address. Since there’s only one meeting left in the term, this will have to wait for the new term.
The gist of the motion is that, for all of the reasons listed above, I don’t want to see data centres in Brighton and I want our staff to be very cautious and transparent about using AI in municipal operations. I never want there to be a question of trust in municipal government.
So the motion that I hope to bring to council will include some aspect of:
- Implementing an Interim Control By-Law to restrict the development of any AI data centres (for the maximum duration of one year, plus a one year extension) until there are better regulations in place. Several Ontario municipalities are attempting this now.
- Directing staff to draft an AI use policy for the municipality, emphasizing the need to uphold public trust.
In my draft I tried to draw a distinction between general and specific AIs. General AIs are the ones I’ve been talking about all along: trained on every scrap of human language they can get their hands on, these AIs are supposed to be able to answer any question and create anything you can dream of (and tell you how brilliant you are for asking them to do it), which is why they’re so energy (and water) guzzling. They’re open to everyone, which means that the unprecedented power they provide can be used by anyone, for good or for evil.
Meanwhile, specific AIs are trained only on data relevant to their purpose, and are applied to that purpose. There are AIs that run climate models for the IPCC, helping them study climate change and helping us adapt to a warmer world. There could be AIs that process planning applications, reducing deskwork for our planners and speeding up the development process. Such an AI could be run from a single computer, rather than a massive data centre. If there are gains to be had from this technology, and it can be deployed in a responsible way, then we should do so. But it’s fundamentally different from all of the things described above.
In Sum
To wrap it all up:
- General AI is hugely problematic in numerous ways, from the existential to the personal to the economic;
- Data centres are hugely problematic and offer little benefit to our community;
- I’d like to see our community stop any AI data centres from being built here, and from using the more problematic forms of AI at all unless there are strong guidelines in place; and
- I want to be able to say that I did what I could to stop the slopocalypse.
Are you with me? What else am I missing in this conversation? Leave a comment below, especially if I’ve gotten something wrong!