Sand Castles: the hollow, market-based politics of AI and Canadian Liberalism

2026-01-15

My eyes are like telescopes /

I see it all backwards, but who wants hope? /

If I ever catch that ventriloquist /

I'll squeeze his head right into my fist /

Something comes a-tracking in /

What is it? What's the prediction? /

I'll betcha it's friction

(Television - "Friction")

The data centre push

In the uproar over the recent Canada-Alberta Memorandum of Understanding, which rolls back almost all of Canada's emissions regulations and clears the way for a new bitumen pipeline, one of the more overlooked aspects was the language around AI data centres. The MoU calls for "the construction of thousands of megawatts of AI computing power", which would almost certainly either eat up any new electricity generation from renewables in the region, or result in burning an enormous amount of gas that, in an ideal world, would be left in the ground. [1,2]

It's headscratching at best, and infuriating at worst. But for argument's sake, let's suppose you aren't bothered by the environmental and social impacts of building a bunch of hyperscale AI data centres in drought-prone northern Alberta, and you're convinced that AI is important and necessary and useful. Even then, the economics of these massive facilities don't make any sense, according to computer scientist Nicholas Weaver, who points to the dramatically reduced cost of running inference either on-device or via on-premises clusters, with the trade-off being a small drop in performance. [3]

On the other hand, there are the opinions of companies like Brookfield (Carney's employer prior to this new gig) who in August 2025 released a whitepaper [4] that casually throws around fantasy-land terms like "Artificial Superintelligence" and "Age of Abundance", and makes the eye-popping claim that $7 trillion will be spent on 75 gigawatts of new AI infrastructure over the next 10 years worldwide.

$7 trillion—that's a lot of money! The market for inference must be enormous, and the projected revenues of these facilities must be astronom—

huh, that's odd, the whitepaper doesn't mention revenues. I wonder why?

¯\_(ツ)_/¯

Regardless, Brookfield has been splashing out on AI infrastructure to the tune of billions of dollars, including co-leading a funding round in July alongside Deutsche Bank who, only a few months later, are reportedly exploring ways to short AI-related stocks as a hedge against the bubble bursting. [5,6,7,8]

Echoes of Thatcher

Our PM doesn't work at Brookfield anymore, of course—he works for the people now! So we might well ask: why is Carney's government still so aligned with his old employer when it comes to pushing the build-out of massive AI infrastructure, given the clear financial risks of doing so?

One possibility: neither Carney nor Brookfield mind taking a financial loss, if it's in the service of a larger project. Besides, it's other people's money they're putting on the line: Carney gets to play with public dollars now, and in Brookfield's case, a lot of that money belongs to Canadians' pension funds.

So maybe the idea is similar to Margaret Thatcher's 'Right to Buy' policy in 1980s Britain, intended to promote a sell-off of public housing and tie the wealth of many in the working class to rising property values, thereby incentivizing support for policies benefiting property owners, even if those benefits would mostly accrue to society's upper crust. Pushing people's retirement savings into AI-related stocks and/or data centre investments (the latter being a policy explicitly championed by Chrystia Freeland and Carney's fellow former BoC Governor Stephen Poloz) further ties individual Canadians' financial well-being to the AI industry—as well as to the surveillance industry that feeds it, the military industry that deploys it, and the fossil fuel industry that hopes to power it. [9,10]

Secondly, a massive data centre rollout, especially one without a realistic revenue stream, can make sense if the goal isn't for it to pay dividends directly, but instead to drive a massively expanded military and domestic surveillance machine, that facilitates the continued, relentless extraction of resources and profit as a second-order effect. And when the VC money and circular financing deals dry up, I expect a lot of the infrastructure now being built ends up being used for those kinds of applications: facial recognition, gait analysis, sentiment analysis, predictive policing, automated drone warfare, target selection, and the like. The "quiet consolidation of America's biometric border" in 2025 will almost certainly be replicated here in Canada, to the extent that it hasn't already been. [11]

Also important to explore, though, is how the continued buildout of this infrastructure fits alongside a parallel ideological commitment to an expansion of market dynamics into more and more aspects of our lives: where statistical prediction and cost-benefit triangulation form the basis of our moral and political thinking, perhaps even of truth itself.

As Justin Joque explores in his 2023 book "Revolutionary Mathematics: Artificial Intelligence, Statistics and the Logic of Capitalism", AI systems have a logic to their functioning that mirrors the objectification and commodification performed by traditional markets. [12] Joque, who cites Safiya Noble's 2018 book "Algorithms of Oppression" as a major influence on his work, explains this in a much deeper and somewhat different way than I'm about to (and has a way better handle on the underlying theory) but hopefully the parallels hold up. [13,14][note 1]

To get a sense of what Joque is driving at, let's first take the labour market as an example: domestic labour is vital for the continued functioning of society and the economy, yet this work mostly goes unpaid. Similarly, the work done by a schoolteacher or nurse, based on the salary of those professions, is 'worth' far less than the work done by a Bay Street day-trader, even though the latter doesn't provide any tangible benefit to society. In general, we rarely question these outcomes; 'the market' has made them appear to be natural facts of how the world works. But in actuality, the wages reflect and reproduce social relations & power structures—in this case, society's long-running devaluation of work traditionally done mostly by women.

In the case of AI, algorithmic systems are fed a pile of data and are tasked with finding statistical patterns for the purpose of making predictions. These predictions often appear to have the status of objective truth, or at least political neutrality, but the system has really just performed a similar sleight-of-hand to that done by the labour market.

With an LLM or other generative system, choices around what text or images to include/exclude in the training data are obviously massively important, since they're repeated in the generated outputs which are presented as objective facts. But the problem runs deeper, too, because even with non-generative applications of machine learning, the measurement data used for training isn't neutral: in gathering it, choices are made to measure some aspects of things/phenomena and not others. These choices are a result of our opinions about what is important to know about a thing, and therefore reflect our often arbitrary and socially-determined biases and assumptions about the world. Correlations found in this data, and predictions based on those correlations, can't help but perpetuate these social dynamics. [Note 2]

This process—whereby our bodies, actions, and interactions are measured, quantified, and turned into objects of correlation detached from their social context and causation, stripped of narrative, meaning, and moral valence—changes the way we see ourselves, undermining our agency and human potential. And this alienation ends up appearing in areas of our lives far beyond what we would ordinarily think of as the domain of markets.

Speculation on speculation

One way to begin to understand this process of alienation and detachment is, appropriately, to look at the way many traders approach the stock market itself. Once upon a time, someone might have pored over spreadsheets and earnings statements and press releases looking for undervalued gems or businesses set up for long-term growth, the idea being that a company's stock price should reflect real-world conditions and events. How quaint!

That approach to the market was supplanted by the idea that the trader's job is not to have an opinion of their own, but instead to predict the behaviour of other market participants, often by looking for visual patterns in the price charts, as opposed to keeping a close eye on the news or reading quarterly financial reports. [16]

The introduction of high-frequency trading algorithms by institutional investors resulted in retail trading techniques even further removed from notions of underlying value. Indeed, in cryptocurrency markets, where the 'assets' have no underlying value whatsoever, the trader's task is ultimately to try and figure out what visual/algorithmic patterns *other* traders are overlaying onto the price charts in order to make their trades, in order to set up your own trades: a perfect ouroboros of financial speculation. Moving averages, 'Fibonacci' numbers, trend lines and channels, even (no joke) actual spirals get drawn onto the charts. It's tea leaves. There's no relation to any sort of real-world production. You may as well sit the market participants around a table with a Ouija board. Who knows, they might even notice there's a spectre haunting the joint.

The hollow market politics of Ezra Klein

This dynamic seems to crop up all over the place: when Ezra Klein looks Ta-Nehisi Coates in the eye and tells him, "to be very clear, I think in a place like Nebraska, you should try to run some pro-life Democrats" [18], it's the exact same structure of detached triangulation as the stock or crypto trader, the exact same commodification, stripping of context and meaning, and undermining of human agency performed by big tech's LLMs. Klein is treating people's rights (and his own supposed values) as negotiable, as subject to a cost-benefit analysis, as chess pieces that can be sacrificed for some potential future advantage. But rights, of course, are either universal or they aren't really rights at all.

Klein attempts to frame Coates' broader, narrative view of history as "fatalistic" and his own granular, transactional approach as liberatory. In truth, the reverse is true: in his unwillingness to stand for anything, Klein is content to be swept along by the tides of polling. Coates, meanwhile, in being honest with himself and his readers, is able to express the genuine agency of someone acting in accordance with what they believe in. [Note 3]

I'm picking on Ezra Klein, but he's just one example of many. For decades, our Canadian media have increasingly focused on polling, public perception, and political strategy, at the expense of addressing the meaning and impact of policy decisions. Recently I even witnessed a well-known opinion columnist cite favourable polls as evidence that the PM did the correct thing in signing the Memorandum of Understanding, ignoring the fact that his own columns (in which he bent over backward to frame the MoU as an environmental *win* despite all evidence to the contrary) likely contributed to those very polling numbers. [19]

Then there's the PM himself, who regularly posts one thing on Bluesky, a completely different thing on Twitter/X, and later says something else entirely in front of a room full of corporate executives. Not that I expected better. The most frustrating part is that many of Carney's supporters applaud this behaviour.

And like Coates did with Klein, I'll concede there are occasions where politicians may feel the need to make strategic compromises or triangulate an advantageous position. But we have to remember: **that process of political triangulation requires that members of the public play their assigned role, which is to advocate for what they believe in, not to surrender their own opinions and agency to the polling results.** Especially because those very polling results are, to at least some extent, themselves the result of other folks surrendering *their* opinions to whatever they believe public sentiment is. It's that ouroboros again.

Ultimately the point is, praising a politician for their strategic manoeuvering on an issue is a self-contradictory position to take; by doing so, you've given up your own position on the issue, a position which is supposedly one of the factors being weighed by the politician. Ultimately all you've done is cede ground.

Kalshi x CNN

Prediction markets like Polymarket and Kalshi are, of course, where everything discussed up until now congeals into a steaming pile of shit, and the goal—the commodification of every aspect of life—is made absolutely explicit. Just a few weeks ago, Kalshi CEO Tarek Monsour said it pretty unambiguously, telling an interviewer that "the long-term vision is to financialize everything and create a tradable asset out of any difference in opinion." [20]

Wow, gross.

And what a disaster for human agency. Right off the bat, it's relatively easy to see how staking money on an outcome you expect is likely to happen, but that you don't *want* to happen, creates a financial incentive to act in the world to bring about an otherwise undesired outcome, pitting you against your own values in the same way Thatcher's 'Right To Buy' policy undercut class solidarity.

Even more worrying: the merging of these markets with existing media continues to close the loop between control of information flows and economic coercion. Compare prediction markets with cryptocurrency exchanges, where the biggest winners are invariably the 'whales' (the market players with the most capital, often the exchanges themselves) who can bend markets to their will. With prediction markets, not only are the platforms themselves often the market movers, but their integration with social media and mass media platforms has the potential to make the early 2020's wave of celebrity crypto pump-and-dumps look kinda adorable by comparison. Partly because of the much broader reach, but also because the media are presenting these gambling lines, which in reality exist somewhere between data-driven speculation and outright influence operation, as sources of truth. [21,22,23]

In other words, although these markets ostensibly operate solely on the idea that those with the most information are best positioned to capitalize on it, the dynamic is also self-reinforcing: those with the most capital are also best positioned to shape our collective understanding of reality—and to transform prediction market odds, that supposedly represent the "wisdom of the crowd", into a propagandistic tool. All while data on public sentiment is slurped up by the tech giants, and presumably sold to the highest bidder. The feedback loop of influence and analysis is getting tighter, and increasingly automated.

Squeeze his egghead right into my fist

The market for prediction isn't limited to those platforms explicitly called prediction markets, however. Chatbots, search engines, Spotify playlists, and social media platforms are all prediction machines at their core, commodifying our words, actions, and interactions, and mining this data for profit.

And if any of the previous section sounds overly paranoid, may I direct your attention to this nifty little essay titled "Prediction: the successor to postmodernism" by Andreessen Horowitz's new (as of Sept 2025) Editor-at-Large, Alex Danco. [24][Note 4]

Danco is probably too advanced in years to be auditioning for the role of Peter Thiel blood-boy, but he tap-dances admirably here. At one point, famed classical pianist Glenn Gould and supervillain consulting firm McKinsey are crammed into a single sentence, and later, meme coins are deemed a "kind of art" (because, you see, if people get mad when you say something is art, it just proves that "you have something interesting"). The essay is capped off with a declaration that "we have a new way to think about our purpose in the world: we create order in the universe by contributing information."

It's a weird piece. The whole thing is ultimately an attempt to frame our surrender to a totalizing form of surveillance capitalism as a categorical imperative. Bow down, mortals, to the great prediction machine in the sky!

But of course, there is no prediction machine in the sky, no information-extracting apparatus at scale that works on behalf of anyone but a handful of centibillionaires unsatisfied with anything less than complete control of everything everywhere. Much like with the whole 'effective altruism' ruse popular in Silicon Valley circa 2022, Danco is straining to back-propagate a moral/philosophical justification for the techno-totalitarianism these assholes are already building, just in case any of his pals aren't sleeping well at night.

Friction

The good news is that these systems aren't totalizing and there are a few recent victories that indicate how we might avoid that fate.

Zohran Mamdani, for instance, showed that when political figures put forward clear, achievable proposals for improving people's material conditions, they can energize mass support.

But it would have been a mistake to build that movement entirely on a narrow version of economic well-being. Because if everyone's pension funds and retirement savings are heavily invested in Uber, then the promise of free buses starts to be looked at primarily as an economic cost-benefit calculation rather than as a universal, pro-environmental, and pro-equity good. "Material conditions" doesn't just mean people having *stuff*, it also means having rights and freedoms. This is also why Mamdani's refusal to shy away from giving his full-throated support for the rights of trans New Yorkers, of racialized New Yorkers, and of immigrant New Yorkers, is a required piece of the puzzle. [Note 5]

Here in Canada, the resounding victory of the flight attendants' union over Air Canada management showed there are basic principles of fairness and workers' rights that are still held more or less universally among people in this country, but that somehow weren't understood by the Carney government's analysis of the dispute.

For all the talk of the strategic prowess of the Liberal Party machine, for all the commissioned polls and focus groups and online sentiment they're presumably crunching the numbers on—overall support for unions, willingness to tolerate flight delays/cancellations, fear of economic disruption, and so on—apparently none of those calculations managed to capture a simple principle: "workers must be paid for their time". The government was so flummoxed, caught so flat-footed by this basic truth, that they tripped over themselves back-tracking from the corporate support they've been in the habit of providing as a matter of course, pretending to be *shocked, shocked I tell you*, that flight attendants weren't getting paid for a third of hours worked. This is despite the fact that it was a key issue in contract negotiations, and despite every single striking worker's picket sign reading "unpaid work won't fly" in what looked to me like a perfectly fucking legible font. [25]

These predictive systems, and their politics of detachment and alienation, do not *and can not* account for meaning or narrative. Basic, strongly-held principles are still capable of rising up, cutting through those granular piles of data and correlation, and sweeping away their algorithmic outputs like sand castles. [Note 6] To be clear: a completely reasonable worry is that these systems may work just well enough for the people with almost all the money and power to retain and/or continue to grow their money and power, that these tightened feedback loops of propagandistic information flows, alienation and detachment, and economic coercion now represent a permanent boot on our necks. The victories of Mamdani and the flight attendants should give us some hope that this world is still capable of surprises big enough to put the oligarchs' dreams to bed.

Notes

1. Joque (assuming I understand correctly!) locates a parallel between 1) the process of objectification in capitalist markets and 2) the production of probability using Bayesian statistical methods.

2. Whether or not the particular social dynamics being perpetuated by a given system are desirable—well, that's a different question. But I think it's safe to say the AI companies receiving billions of dollars in funding from major investment firms and venture capitalists aren't busting their tails to create liberatory AI projects, if such things are even possible.

3. Two other things I'll note about this: there's arguably a similarity between Coates' intervention here with Klein's transactional politics and Daniel Dennett's compatibilist argument against the 'free will does not exist', billiard-ball determinism of Robert Sapolsky, which I wrote about in a short piece a couple of years ago; [26] the relevant point here is that both Coates and Dennett see meaning and narrative as the domains where human beings express agency. Secondly, the debate is also about ethical systems, with Klein advocating for the primacy of the consequentialist view, and Coates asking him to incorporate more of the deontological position. To me, the expected-return calculation of consequentialism mirrors those of the capitalist profit motive; see also Carney's rhetoric around "focusing on outcomes not targets".

4. Before his move to a16z, Danco, a Torontonian, was Director of Product at Shopify, whose executives have been spearheading a pro-AI, pro-fossil fuel campaign called Build Canada, which quite clearly has the ear of the Prime Minister [27] and whose "We're Not DOGE" t-shirt has been raising questions already answered by the t-shirt. [28] Also, one doesn't have to squint very hard to spot similarities between this essay from Danco, and the chapter of Mark Carney's 2021 book 'Value(s)' called "The Future of Money", for instance with the comparison of social media 'likes' to actual currency.

5. This is also why the extremely intentional shift among Canadian elites from talking about "reconciliation" to "economic reconciliation" with Indigenous peoples is so worrying! See also the thinking of Atossa Soltani (representing the Amazon Sacred Headwaters) in response to the idea of financializing the work of land defenders. [29]

6. To be clear, I agree with Ta-Nehisi Coates that these basic principles shouldn't have to be (and often won't be) positions held by the vast majority of people, although obviously it would be great if they were; Coates, for example, points out to Ezra Klein that "the line we [Black America] have drawn in general has not been majoritarian politics, unfortunately."