AI is Wreaking Havoc on Higher Education, but the System Was Flawed from the Start

If you work at a university today, you talk a lot with the dead. I don’t mean that as a metaphor – or, at least, not entirely.

Generative AI trains on people. Its algorithms learn from humans, requiring our quirks and idiosyncrasies to improve. If the models are exposed predominantly to the output from other AIs, they degenerate and, eventually, collapse.

That explains a strange new trend reported by antiquarian booksellers, who told the Guardian of bulk orders for old stock, originating, they thought, from AI companies or their surrogates. The titles were bafflingly esoteric: one bibliophile described a purchase that included a manual on soil mechanics from the 1970s, a book called Born to Thunder: Champions of New Zealand Cycling, a history of the Melbourne suburb of Hawthorn, and a copy of Early Australian Poetry.

But for AI trainers, the subjects don’t necessarily matter. Back in the day, the companies vacuumed up human content from Reddit, Wikipedia and similar sites. With data now prohibitively expensive, remaindered books provide an attractive option for “destructive scanning”, in which machines slice apart bindings to feed on the creativity contained therein.

As an academic, you encounter the process at the opposite end. When you open Speed Grader to confront another bland but competent assignment, suspiciously devoid of grammatical or spelling errors, your heart sinks. You know that, in all probability, the document you’re reading was composed not by your student (whose contribution likely extended no further than dropping the essay question into Claude’s interface and adding “make it sound human”), but by a legion of absent (and, quite probably, deceased) authors, whose forgotten texts enabled an algorithm to stitch together an essay that is at least passable and often reasonably good.

Grading induces a sense of utter futility: what could be less meaningful than offering feedback to dead people?

Cheating via algorithm

No one knows the proportion of Australian students reliant on AI for university assessments. One study of 8,000 people found 40% admitting they had cheated via algorithm. But that was in 2024 – aeons ago, given the rapidity of technological improvement.

A 2025 survey by the ARC Centre of Excellence for Automated Decision-Making and Society at RMIT identified students as particularly heavy users of AI. Nearly 80% employed the technology – though the study did not specify what for. Research on American undergraduates conducted in 2024 by UC Berkeley’s Center for Studies in Higher Education suggested two-thirds used AI. The following year, a UK analysis suggested generative AI was “near universal” among British students.

The figures don’t necessarily refer to assessment. My students use AI for all manner of tasks. Many send emails inflected with the slightly old-fashioned ChatGPT or DeepSeek obsequity, so requests for extensions read like missives from 19th century courtiers. Research suggests rates of cheating vary significantly by discipline, with certain fields faring better than others.

Still, if you believe in the integrity of traditional assessment, I have an AI-generated bridge to sell you.

“Why memorise facts or do equations or slog it out over an essay when ChatGPT or Gemini can do it in seconds?” a Macquarie University student identified as “Hayden” asked rhetorically in a recent report by the Australian’s Ros Thomas:

You get everything in the blink of an eye – any equation, any answer, any essay. It’s all being hacked by AI and the sooner people accept that the better. Stop worrying about what’s right and wrong, because everyone’s using it and you’d be dumb not to.

Thomas presses Hayden as to whether he knows any students who complete their assignments without AI. “Nup,” he replies. “Not a single person.”

The issue extends beyond students and beyond assessment. AI use proliferates in higher education because decades of neoliberal reforms have baked into universities a certain way of thinking, entirely compatible with algorithmic models.

To put it another way, the reliance of both students and staff on generative AI results from a broader crisis – a crisis which generative AI, in turn, intensifies. For that reason, AI will, I think, increasingly become a focus for the deep discontent already percolating throughout the sector.

Industrial-scale fraud

Let’s establish what is at stake.

I teach a subject on the history of journalism. In it, I quote the scholar
Charlie Beckett’s description of how the media first integrated generative AI, not through revolutionary changes, but via tools (transcription software, grammar checkers, and so on) that automated activities most reporters regarded as tedious and inconsequential.

“Artificial intelligence,” Beckett says, “is a collection of ideas, technologies, and techniques that relate to a computer system’s capacity to perform tasks normally requiring human intelligence.”

A definition that stresses the “what” rather than the “how” usefully focuses discussion on pragmatic choices in a pragmatic industry. Yet, in other contexts, it is deeply misleading.

I invite my class to visualise a university in which all students rely on ChatGPT to complete their assignments. They do not find this difficult. I tell them some universities – for example, Adelaide University, the University of Newcastle, Deakin University and RMIT – encourage staff to employ AI tools to accelerate grading, though all insist a human assessor signs off the result.

What would it mean, I ask, if both academics and students relied completely on AI?

In one sense, a university in which both parties outsourced assessment to machines would simply illustrate the process Beckett described, with tasks hitherto reliant on human intelligence instead performed automatically. Yet, in such a scenario, the substitution of technology for brain power would do more than increase efficiency. If university bots graded bots, you would no longer have a university. You’d have a tomb, an academic mausoleum in which digital phantoms murmured emptily.

The scale of the problem in Australian education became apparent in May this year, when the academic Kylie Moore-Gilbert wrote a piece for the Nine papers accusing “all of Australia’s universities” of “committing widespread, industrial-scale fraud”. She described a situation so dire that she considered advising her stepdaughter to think twice about attending university, a place where “kids are now taking on tens of thousands of dollars in debt to have a terrible campus experience while being graded on who can write the best AI prompts”.

Cath Ellis, Western Sydney University’s pro-vice-chancellor for quality and integrity, responded in the same newspapers. She reminded Moore-Gilbert that, for many young people, “university is not an optional extra”, but the “only route through to their chosen profession”. While acknowledging AI posed pedagogical challenges, Ellis insisted universities were not “turning a blind eye”, but were finding ways to encourage students to “value their own thinking”.

Her article culminated in suggestions for Moore-Gilbert’s stepdaughter and other high school students contemplating an academic future. “Don’t cut corners,” the professor urged:

Don’t outsource your thinking, however tempting it may be. […] Real learning has never been easy. That is precisely why it matters. And the graduates who do that hard work are exactly the ones we still need.

The wisdom was rather undercut by revelations that Ellis’ article relied so heavily on AI that the Age and the Sydney Morning Herald removed it from their sites.

A few months later, the same episode played out again, this time in the US. Santiago Schnell, provost of Dartmouth College, wrote for the Washington Post, explaining how educators needed to ensure they were assessing the capacity of students rather than AI-generated work, because

a degree is a public claim as well as a private benefit. Employers, patients, clients and citizens rely on it as evidence that its holder can exercise judgment.

Schnell insisted universities needed new measures to “deter cheating and help professors see what students can do without AI”.

Students might have been forgiven for wondering what professors could do without AI, as the university paper The Dartmouth suggested the provost relied on generated text, not only for the Washington Post article but scores of others, including scholarly work.

Mechanical language

I cite these incidents not to single out the individuals involved, but to caution against attributing the universities’ AI problem to the failings of callow undergraduates – or, indeed, merely to problems with assessment. Symptomatically, Western Sydney University responded to the Ellis situation not with an apology, but a justification.

“To write her opinion article,” the university explained,

Ellis uploaded 40,000 words of her own original materials into a Copilot Large Language Model (LLM). The model summarised her extensive base of knowledge, providing prompts. This was the basis of the early drafts, reflecting [Ellis’] own thinking, ideas and opinions built up over more than a decade of dedicated work as a global leader in this field.

The statement concluded by pronouncing “the AI use in this case was appropriate”.

The willingness to defend an article Nine removed for violating editorial policy hints at my broader point: the “outsourcing of thinking” has become, to employ a suitably computational phrase, not a bug but a feature of the neoliberal university.

Managerial language may well be to the information age what the machine and the assembly line were to the industrial. It is mechanised language. Like a machine, it removes the need for thinking: this essential and uniquely human faculty is suspended.

That’s Don Watson, Paul Keating’s former speech writer, in his 2003 book Death Sentence, a polemic against the rhetoric of managerialism. The next year, when Watson compiled a dictionary highlighting the replacement of “living words” with “dead ones”, he sourced many of his examples from the nation’s universities.

The proliferation of “mechanical language” in higher education represents more than a rhetorical failure. It is a linguistic manifestation of the process that, long before ChatGPT had been invented, primed the universities for AI. To understand why Western Sydney genuinely cannot grasp the problem with its pro-vice-chancellor’s intervention we need to look back at the 1980s.

In that decade, John Dawkins, Bob Hawke’s energetic education minister, set about restructuring the university sector on the basis of what was then called “economic rationalism” and we now know as “neoliberalism”. Government funding fell from something like 80% in the 1980s to around 40%, while enrolments grew substantially.

To compensate, universities turned to fees paid by local students via HECS and a vast increase in overseas enrolments. The new policies brought a new internal culture. As one account put it, reformers sought

to strengthen direct and indirect competition between universities on a system level while, at the institutional level, encouraging changes towards higher levels of managerial autonomy and control and lower levels of academic self-governance.

The Dawkins plan promised increased productivity, as the universities implemented managerial strategies familiar from other industries.

By 2024, 60% of the university workforce was precariously employed: uncertain as to whether work would be available from semester to semester, and not entitled to provisions such as sick leave and annual leave. One casual academic told the Sydney Morning Herald of how, to survive, she accepted eight different short-term contracts, juggling teaching, marking and research work across multiple institutions.

Some institutions have since committed themselves to decasualisation. Nevertheless, workloads for all staff have grown and grown and grown. In a recent survey, 88% of respondents with permanent positions said their unrecognised workload had increased since 2019.

Intelligible labour

The neoliberal turn required managers to make the extra work intelligible to the market. This is a point made by philosopher Mark Fisher in Capitalist Realism:

the drive to assess the performance of workers and to measure forms of labour which, by their nature, are resistant to quantification, has inevitably required additional layers of management and bureaucracy.

The ABC recently revealed Australian universities spend A$1.8 billion each year on management consultants. What, precisely, do these people do?

In 2024, the University of Technology Sydney gave millions of dollars to consultant firm KPMG to oversee a restructure. The result, the ABC reported, was “at least 24 KPMG staff, including directors and partners, soon had UTS email addresses, could access the university’s Microsoft Teams and SharePoint systems, and were attending staff meetings”. They were there to ascertain “which courses and academic programs were generating revenue for the university and which were not”.

Fisher explains managers engage not

in a direct comparison of workers’ performance or output, but a comparison between the audited representation of that performance and output. Inevitably, a short circuiting occurs, and work becomes geared towards the generation and massaging of representations rather than […] the official goals of the work itself.

He coined the term “market Stalinism”, a phrase that neatly captures the bureaucratic prose Watson describes: a leaden, grey language arising from a disjunction between what is being said and what is being done. The lexicon of the modern university arises in the same fashion. Reflecting on the British experience, historian Ross McKibbin remarks:

Our students are now clients, our patients and passengers customers. It is a language which was first devised in business schools, then broke into government and now infests all institutions […] it might be laughable but it is now the language shared by all those who command, and is one way they wield power.

The implications of Fisher’s argument extend beyond managerial rhetoric. The neoliberal project encountered particular challenges in relation to traditional scholarly pursuits such as research. Managers couldn’t foster productivity without objective evidence as to whether tweedy dons were actually working or simply staring into the distance.

Publication provided one obvious answer, precisely because it could be tabulated numerically. Any ledger could document the relative productivity of the scholar who published two articles and her lazy colleague who managed only one.

Entrepreneurial publishers recognised an extraordinary business opportunity in the marketisation of research. Academic journals paid nothing to authors, since researchers were not supposed to be motivated by money. They gave nothing to peer reviewers for the same reason. Yet the old traditions of collegiality did not prevent publishers levelling huge subscription fees on educational libraries and other institutions to access content acquired for free.

Such was the basis of what is now a multi-billion-dollar industry. More than half of the market in scholarly journals is now owned by five big corporations.

The grotesque commercialisation of academic publishing harms genuine research. It limits the circulation of knowledge to those able to access a well-funded academic library. It transforms peer reviewing into what management scholar Stuart Macdonald calls “little more than hollow ritual”, with overworked academics increasingly disinclined to provide free labour.

Yet corporatised universities can scarcely complain about someone else making a buck. As Macdonald says,

the modern university may still portray itself as a cloistered community of scholars, but it is actually a large multinational competing with many other such universities in selling a product. There is little of the collegial in the higher education business; professionalism has surrendered unconditionally to managerialism and performance measured in large part by publication of papers in journals.

Scholarly publishing, with its complex formulas distilling the supposed quality and influence of research into numbers and indexes, provides a vital service. It creates, in Macdonald’s words, the “indicators of academic performance on which rankings and the distribution of resources in higher education are largely based”.

Such indicators facilitate the slippage Fisher describes. The symbols of work (citation indexes) take precedence over the supposed purpose of the activity (the accumulation of knowledge). The data from commercialised research means, as Macdonald observes, “managers and administrators in both publishing and higher education can now pronounce with authority on the quality of thousands of academic journals without ever opening any of them”.

The numbers matter, in other words, more than the research they represent.

As far back as 2020, Mario Biagioli and Alexandra Lippman could introduce their collection Gaming the Metrics: Misconduct and Manipulation in Academic Research by explaining human reading was “no longer a necessary component of institutional forms of assessment”.

They weren’t referring to generative AI, which was not yet widely available. The manipulation they discussed centred on practices such as self-referencing, citation swapping between journals, fake publications and the like: an array of techniques by which enterprising researchers could juice their scores.

But in Biagioli and Lippman’s description of a system resting on “metadata and similar markers that can be picked out and processed by non-humans”, we can see how, long before AI lived in everyone’s browser, higher education was inadvertently readying itself for the subsequent tenure of Professor Claude, Dr ChatGPT and their colleagues.

Scholarly Stakhanovites

In 1935, the Donbass miner Alexei Stakhanov responded to Stalin’s calls for intensified production by extracting a record-breaking 227 tonnes of coal in a single shift. His exploits – and the increasingly fantastical achievements attributed to the subsequent Stakhanovite movement – emerged almost organically from the untethering of bureaucratic metrics from Soviet reality.

Stakhanov’s mighty output depended, it is now thought, on the unacknowledged contributions of other miners. But both he and the commissars above him had obvious motivations for massaging the truth. Stakhanov became a celebrity, and Stalin and his henchmen used his example to induce others to work harder.

Recently, the Washington Post reported on the University of Chicago statistician Nicholas Polson, who lent his name as author or co-author to more than 200 papers and 14 books in a single year. The efforts of this scholarly Stakhanovite were, as Polson acknowledged, facilitated by AI.

More profoundly, they were fostered by a university system that privileges what Fisher calls “the generation and massaging of representation”, something in which AI excels.

Earlier this year, the editorial team at Organization Science, an academic journal of management research, conducted a study of the articles they were receiving. They found, since the emergence of ChatGPT in late 2022, their submissions had risen 42%.

Forbes noted:

The editors attribute nearly all of that increase to AI. Submissions judged to be human-only actually declined. By early 2026, the majority of manuscripts submitted to the journal show some degree of AI involvement, and the fastest-growing category is papers scored at 70% or higher AI content.

“AI, as it’s being used today, is colliding with institutional incentives to create more rather than better research,” one of the editors explained. “It’s not AI on its own. It’s AI plus publish-or-perish incentives.”

Polson’s productivity might cause his institution some embarrassment, but only because of the attention garnered by its extravagance. Had he not attracted the media’s notice, he would have been hailed for his work ethic: a veritable hero of academic labour.

In his 2011 manifesto For the University, literary scholar Thomas Docherty of the University of Warwick presented an argument about modern education that has only become more relevant in the AI age.

A university, Docherty insisted, should be where “we cultivate our humanity”: a place “where we go in order to find out not what we know but rather the extent of what we do not”. Instead, the education sector had come to emphasise what was certain, familiar and replicable: “there exists in our contemporary situation a profound confusion of information with knowledge”.

We’ve discussed how, for Mark Fisher, this distinction between information and knowledge – the latter is uniquely human, the former perfectly suited to algorithmic systems – arises from marketisation. We’ve seen what that means for research. Now let’s look at teaching.

In 2027, local students undertaking degrees at Australian universities in law, accounting, business, economics, communications, history, philosophy, political science and anthropology will each pay more than $18,000 a year. For a three-year degree, students will owe over $54,000. Students in double degrees will graduate some $90,125 in debt.

International students will pay much more and much sooner. The higher education sector in Australia depends heavily on enrolments from abroad. The University of Sydney now sources 47% of its revenue from international students. At the University of Melbourne, such a student might pay as much as $187,000 to complete a Bachelor of Arts degree in 2027. For science, the bill will come to something like $200,000.

What do students get in return for these eye-watering sums? Merely asking the question raises the issues central to Docherty’s concerns. Markets, by their nature, depend on numerical equivalents, the commercial exchange of like for like. A university degree costing $200,000 must guarantee an equivalent value. In the export market, that depends on reputational worth, centred on the particular course and the university brand.

Some scholars still devote themselves to knowledge, regardless of H-indexes and other metrics. Similarly, many students genuinely want to learn. But the vast debts they acquire push them to prioritise what Fisher calls “representation” – the certificate associated with their course. They’re not paying to hear about what they don’t know.

The expectations of the universities are equally simple. Starved of public funding, they cannot function without constant new applicants. Like all businesses, they must expand. The institutions that don’t grow risk falling behind their rivals. Universities, in other words, need more and more enrolments – and to get them they must give students what they want.

The resulting system emphasises the informational certainty Docherty counterposed to genuine knowledge: set fees in return for set outcomes. As criminologist Pete Chambers says, the university system is “designed to smoothly pump through as many students as possible, not to provide a better education to a greater number, but because it is both profitable and scalable”.

The extractive model

Almost every aspect incentivises generative AI. The universities’ need for applicants creates financial pressure to lower barriers, most obviously by reducing the requirements for English-language competency. As a result, international students routinely arrive in Australia only to discover, despite what they were promised, they lack the proficiency to complete the courses for which they have paid.

In 2024, the Guardian quoted a tutor who explained how she was teaching people who could not “speak, write or understand basic English” and so “use translators or text capture to translate the lectures and tutorials, translation aids to read the literature and ChatGPT to generate ideas”.

“The thru-put model,” Chambers observes, “is highly extractive – and needs to be – to perpetuate, entrench and expand to keep pace with its own swelling.” The competitive pressures pushing universities to recruit students also leads them to reduce what they spend supporting those students.

Earlier this year, National Union of Students international officer Ariya Masud said international students “are regarded as ATMs to funnel a multibillion-dollar industry”. Masud was referring specifically to the government’s increase, without notice, of non-refundable visa fees for graduates from $2,300 to $4,600, but the point could be made more generally about a cohort marked for exploitation at every point.

As Jeryn Chang from the University of Queensland Association of Postgraduate Students says, international students, far away from friends and family, are acutely vulnerable to scams by companies offering them housing at exploitative rates or employers systematically underpaying them:

International students are often the ones who get the short end of the stick because they don’t have institutions to back them. They’re taken advantage of.

Certainly, the drive for increased productivity reduces the standard of teaching. In one recent study, higher education staff reported an adjusted annual workload averaging 3,256 hours. “Across the year,” one academic explained,

I did 25% more hours in 2025 than I was paid for. Since starting teaching at [my university] in Feb 2022, my own records show that I have completed six years worth of work in terms of 1,590 hours paid for each working year. Two years worth of free overtime for [my university] in four years. I’m knackered.

In the same survey, respondents described how that affected learning conditions: 53% reported their overall course and subject size increasing; 62% described an expansion of tutorials and labs; 47% reported an increase in numbers in workshops; 38% reported an increase in lectures; and 34% reported an increase in students attending practical demonstrations.

Reduced expectations

In the competitive higher education market, each university must keep costs down. Many are experimenting with high-tech ways of boosting student load, usually by deploying more online or pre-recorded material. The Age recently reported on undergraduates at the University of Melbourne watching clips recorded six years earlier, alongside in-person lectures.

The transactional nature of corporate education facilitates this. The students are paying for a degree, not to learn. The pedagogical experience doesn’t matter, so long as the qualifications being sold are eventually delivered – even, or perhaps especially, to students who barely know the language in which they’re studying.

To keep the fees coming, universities reduce expectations. Most on-campus lectures are now recorded, so students no longer need attend. They can choose to watch (or not) on Canvas, the online learning management system. As a result, lecture theatres are regularly almost empty.

The “thru-put” model encourages the universities not to look too closely at how students fulfil the minimal requirements expected of them. In 2024, the Guardian reported Australian academics felt they were “pressured into passing hundreds of students suspected of plagiarism and other forms of cheating in order to maintain their universities’ revenue streams”.

Students are, of course, quite aware of the situation. “I use Gemini and ChatGPT for all my assignments and projects,” an international student explained to the Australian’s Ros Thomas:

A third of my exams are online, and some units entirely, so AI is very convenient to use. […] Some units say, ‘You cannot use AI or you will fail’, but the risk is worth it because AI is so much smarter than us. So we pay for our own AI detection tools, to confuse the university’s AI detection tools. We think it’s weird no one ever gets punished, but we think we know why – the university just wants our money.

As far back as 2009, Fisher documented a system that “literally cannot afford to exclude students, even if it wanted to”. It imposed, he said, no sanctions on those who didn’t attend or who spent their classes listening to headphones.

Even before AI, a black market had developed around the provision of essays or other assignments for students who required them. ChatGPT and similar platforms now provide a similar service, at almost no cost. Describing a campus in which students “pass without learning anything”, Chambers notes

This systemic state of play predates AI by at least a dozen years, but AI is making things much worse, very quickly. […] It’s not hard to see from a student’s point of view that using AI is a prudent and rational response to the hollowing out of the universities.

In today’s university, Chambers says bitterly, “knowledge is for chumps”.

Drive-thru education

Education remains, at its core, an activity centred on human exchanges. Even today, I know colleagues almost entirely unaffected by ChatGPT, usually because they teach smaller, non-vocational subjects to a cohort passionate about the material. In their classes, students might, occasionally, deploy AI as a useful tool, but they don’t need exhortations about not outsourcing their thinking.

Why would they? If you enjoy cooking, the availability of fast food doesn’t deter you from painstakingly preparing a gourmet meal. For most students and staff, however, the university experience could be summarised by the slogan of a certain burger chain, the one that proclaims “billions served”.

That sentiment pervades a recent National Tertiary Education Union survey, in which university employees discuss their work. “I am a good teacher,” says one respondent, “and I want to keep the standards high, but everything is about McDonaldisation.”

A drive-thru education fosters AI because it sucks all the satisfaction from learning and from teaching. “It’s exhausting and demoralising,” explained another staff member, “providing students with meaningful learning opportunities in institutions that don’t care for me or them because their attention is on progressing their market position.”

Many students say the same. In interviews with the ABC last month, high school and university students explained, just because they used ChatGPT, that didn’t mean they liked it. “We’re scared,” said Elise Humphreys, a 19-year-old at Curtin University, “we don’t want it to affect us long term, and we don’t want to live in a world that’s run by robots.”

In the US, the hostility has been playing out in confrontations at commencement ceremonies. Earlier this year, Central Florida graduates booed real estate mogul Gloria Caulfield after she hailed AI as the “next industrial revolution”. Students at Middle Tennessee State University gave Scott Borchetta, an AI-promoting music industry CEO, the same treatment. At the University of Arizona, ex-Google head Eric Schmidt enthused about how AI “would shape the world”, and was jeered.

One student afterwards explained his response by saying people like Schmidt were “putting the wants and needs of billionaires over us”.

By contrast, when comedian Ronny Chieng told Harvard graduates “the mission of your generation is to destroy AI”, they cheered in delight. The clip of Chieng chanting “fuck AI, fuck AI, fuck AI” went viral.

“Gen Z already grew up in a world that they felt was shrinking,” says journalist Jasmine Sun, “where grift and shitcoins and sports gambling looked like the only paths up. Now, they’re being told AI is the reason they can’t get a job – and potentially never will.”

She describes the emergence of “AI populism”, which views the technology as

an elite political project to be resisted […] a thing manufactured by out-of-touch billionaires and pushed onto an unwilling public to achieve sinister aims like ‘capitalist efficiency’ (layoffs) and ‘population management’ (surveillance).

Sun regards this with suspicion. For AI populists, she says, the technology’s “utility as a tool is immaterial relative to the unwelcome societal change it represents.”

I think that misses the point. I don’t believe the anger correlates with an indifference to AI’s achievements. On the contrary, a perception that, in a different world, the tech could be used for ends that were genuinely useful only fuels the outrage about what’s being done with it now.

We all know AI can predict how proteins will fold; we all know it can decode Roman scrolls carbonised in Pompeii. Recently, the Financial Times speculated generative AI might facilitate a planned economy: it could, the paper suggested, overcome the complexity that neoliberal economists Ludwig von Mises and Friedrich von Hayek once claimed made socialism unworkable.

Could there be a better illustration of how the abstract potential of a technology reveals little about its real-world deployment?

The Financial Times might be correct that AI could, in theory, benevolently guide the distribution of resources so each contributes according to their ability and receives according to their need. Yet no one thinks OpenAI will use its bags of tricks to facilitate socialist planning.

The function of AI cannot, in other words, be separated from the social order in which it emerges.

An inhuman power

Many non-academics struggle to believe the stories emanating from higher education. How can it be true, they ask, that elite institutions are now passing students who complete all their assessments via ChatGPT? Aren’t they aware that, by churning out graduates who can barely read or write English, they are destroying their own reputations? Surely they recognise their business model as unsustainable?

Such questions reveal an almost wilful naivety about a 21st century economy in which “sustainability” plays no role whatsoever. The short-termism of universities burning through their social licence neatly parallels the roll out of the resource-hungry, carbon intensive data centres on which AI depends. In both cases, the underlying philosophy might be rendered as après moi, le déluge (after me, the flood) – a phrase that, in the wake of the Nepalese floods, we should understand both literally and figuratively.

Indeed, by thinking about climate change, we can put both AI boosterism and doomerism into perspective.

Earlier this year, former employees of Anthropic warned the tech being built by the company might, as one researcher put it, “kill us all by the end of the decade […] no other human activity poses this level of danger”.

You don’t need to peer into the future to see terrible deeds facilitated by AI. In Gaza, for instance, the IDF already deploys a system called “Habsora” (“The Gospel”) to generate targets automatically. It is a “mass assassination factory”, according to one former intelligence officer.

But the doomerist claims about the menace of conscious AI elide the immediate and catastrophic danger associated with global warming. This year, the UN declared the goal of limiting climate change to 1.5 degrees above pre-industrial measures out of reach.

I don’t mean the comparison as whataboutery; I mean the environmental crisis shows how our social system turns even the crudest technology against us, with even a simple internal combustion engine possessing a kind of malign agency. We don’t need to imagine computers trying to exterminate us: right now, our cars compel us to provide more and more of the oil that’s heating the planet.

Disastrous for humans, but necessary for machines, fossil fuel illustrates Karl Marx’s point about how, under capitalism, “products of the human brain appear as autonomous figures endowed with a life of their own, which enter into relations both with each other and with the human race”.

When Marx warned about human creativity becoming “other than itself” until “an inhuman power rules over everything”, he wasn’t talking against rogue AIs. He was explaining the normal functioning of a market economy.

To put it another way, the algorithmic university might be unsustainable, but that doesn’t make it anomalous, so much as exemplary. It is entirely compatible with a carbon economy that can’t continue and yet somehow does.

AI poses a particularly acute threat to academics because of the nature of academic labour. But always and everywhere capitalism restricts human autonomy; always and everywhere it makes people subservient to things, so that machines become our masters rather than our servants. Even if you don’t work in higher education, you might heed the warning Marx addressed to readers yet unaffected by the ills he described in Capital.

“De te fabula narratur!” he wrote. “The tale I tell is about you.”

The wage theft machine

In 2024, the pay packet of vice-chancellors across 38 public universities
averaged out at $1.02 million. The recent National Tertiary Education Union survey identified 300 university executives who earned more than state premiers.

These are the people who presided over what the Sydney Morning Herald called “a university wage theft machine”. Throughout the industry, precarious staff – who did not know from semester to semester whether they would have any income at all – were paid on piece rates that did not correlate with the tasks they were expected to perform. The NTEU estimates total underpayments across the sector reached a staggering $450 million.

That’s the context for the deep, deep discontent within the sector. When the faculty of arts and social sciences at Sydney University conducted an internal survey, zero percent of staff – yes, zero – in one school agreed with the statement: “I trust the University of Sydney to act in the best interests of staff, students and the broader community.”

The result was not anomalous. In 2022, researchers found academic staff in Western Australia considered their workplaces

exploitative, oppressive, toxic and fiscally driven. They felt themselves being dehumanised and demoralised by management. Most reported experiencing feelings associated with burnout, including anxiety, cynicism, depression and exhaustion.

The unhappiness pertains to the insane workloads now normalised across the sector. “I didn’t have a weekend off for five months straight last year,” one worker reported in the NTEU report. But it is also related to what the university has become.

The problems aren’t caused by AI. As historian Nils Gilman says, “the arrival of large language models is acting as a catalytic solvent, titrating out the incoherence that was always there”. We might think of AI as a nucleation site where other issues crystallise and become visible.

Because of that, it may yet prove a focus of resistance, as the concept of “AI populism” implies. In early September, staff at the University of Sydney walked out on strike. One of their demands pertained to AI. Peter Chen, the union branch president, declared:

Higher education is the canary in the coalmine. Where we go on AI […] will travel through the rest of the economy.

That’s obviously true. The state of higher education should serve as a warning about the path we are hurtling down.

But warnings aren’t enough. Less than a fortnight after the University of Sydney strike, management signed a multiyear partnership with OpenAI, intended to facilitate the use of AI agents “at all levels of learning”. In announcing the deal, Vice-chancellor Mark Scott declared the university was “proudly leading […] the widespread adoption of artificial intelligence”.

This is an old story, of course. Since the industrial revolution, employers have deployed technology to intensify the tempo of work and weaken the power of those they hire. As Marx says,

the rule of the capitalist over the worker is nothing but the rule of the independent conditions of labour over the worker […] the rule of things over man, of dead labour over living.

In a different society, AI might facilitate wonders and miracles. In this era, as the ecosocialist Andreas Malm warns in Fossil Capital, “the traditions of the dead breathe ever heavier down the necks of the living and the not-yet-born”.

Today, AI is a project of the powerful, implemented in the name of an efficiency that has nothing to do with the good. The lines could not be clearer. A struggle for the future looms, in the university as elsewhere: for genuine knowledge, for humanity, for life.