AI has already come for your job. Well, if you’re a recent graduate in the UK, at least. An editorial by The Guardian laments how entry-level tasks in the UK are being taken over by new AI technology, citing a report that shows that the job market for young people fresh out of university is tougher than at any time since 2018. Compared with 2024, the number of jobs advertised for recent graduates is down a whopping 33%.
South Africans will smile wryly at this, given that a recent report by Stellenbosch University and master’s student Hannah MacGinty shows that graduate unemployment has doubled over the past 16 years — and, to no-one’s surprise (except maybe the freedom fighters at AfriForum), women, African and young graduates have been hit the hardest. All this in a country where the current unemployment rate is already at a destructive 32.9%, or 41.9% if you use the expanded definition of unemployment that includes those who have simply just given up looking.
The Guardian editorial urges that “businesses and government must not sacrifice the next generation on the altar of tech”, pointing out that the steepest drop in graduate jobs is in occupations where the impact of AI is starting to bite, or as the newspaper puts it, the bottom rungs of the white-collar career ladder.
“According to some analysts, a growing number of companies are using AI to perform many of the collative, summarising or research-heavy kind of tasks that young graduates might previously have done. IBM’s human resources department now deploys AI to perform 94% of routine tasks, including performance reviews and development plans. Other employers may be hiring fewer graduates than in the past on the assumption that the use of AI will transform their productivity.”
In a country such as South Africa, one of the most unequal in the world, we will also want reassurance that businesses and the government aren’t rushing to adopt AI solutions at the expense of employing actual people. In a social landscape already plagued by various kinds of privilege, do we really want to add robot privilege to the list?
On the other hand — and there’s always another hand — we don’t want to miss out on the supposed benefits that AI might bring to problems that otherwise seem insurmountable. What we do want to avoid, though, is being sold a con by AI snake-oil salesmen (I can’t bring myself to say “salespersons”, it just doesn’t feel right). A new book by Emily M Bender and Alex Hanna, The AI Con, is an attempt to provide a level of inoculation against AI hype.
In a social landscape already plagued by various kinds of privilege, do we really want to add robot privilege to the list?
The authors lambaste the way Big Tech and its remoras — the AI consultants and shills feeding off the main body — drown us in AI hype, basically lying about the capabilities of the large language models that underpin the workings of AI.
They also take issue with how its praise singers anthropomorphise AI so that they can obscure the fact that it’s just a machine, and make the point that AI is mostly used by businesses as a tool to devalue human labour across various industries. Bender and Hanna describe AI as primarily a marketing term — one that doesn’t refer to a coherent set of technologies but is rather deployed for profit by making people believe technology is human-like. The authors’ focus is the US, but the examples — from Hollywood to journalism and social services — travel well.
The AI Con also shows how these automated decision-making systems, though framed as efficiencies, perpetuate existing biases and inequalities. And since the A in AI basically stands for American, you can imagine that these biases will be even more egregious in African contexts.
The authors divide beliefs about AI into two camps: AI doomsters and AI boosters, both of which contribute to the overflowing sewage that is AI hype. Doomsters are the ones who believe that, sooner or later, the development of AI will trigger a mass extinction event for humanity, with machines becoming sentient and developing their own preferences and interests that diverge dramatically from humanity’s, ultimately leading to them exterminating us.
Typically, they’re worried about AI gaining control over critical infrastructure, such as electrical grids or nuclear weaponry, or just using language to manipulate people.
AI boosters … well, they’re everywhere, the unkillable cockroaches of our information ecosystem. They’re the true believers who trumpet the idea that AI holds almost unimaginable potential (though they can imagine it just fine — for the right price) and will usher in a utopian future of ease and abundance for all.
They believe that AI technologies are general-purpose tools, such as electricity or the wheel, and are poised to advance global economic development. Life will be wonderful. Menial jobs, such as making graphics or writing columns, will be automated, and your AI buddies will even stand in for humans in meetings. In their defence, I did use my trusty AI companion to help synthesise some of The AI Con’s takeaways. But the real big lie they tell is that AI will let anyone become an artist, effectively democratising all those skills that are now too difficult for people to do without investing an unfeasible amount of time.
As Bender pointed out on an episode of the AI Inside podcast (which I heartily recommend for anyone who wants to keep up with AI but appreciates a healthy dose of cynicism), the idea that AI can help you to become creative is just a way to avoid facing society’s real problems.
Which are that most people live in a capitalist hellscape where being creative is a luxury few can afford, and it’s easier for those who should be working on solutions to pretend that AI can do it for us, rather than trying to fix the material conditions of people. Even worse, by focusing on hypothetical future benefits or existential risks, AI hype conveniently distracts attention from the real and immediate harms caused by existing automation: job degradation (such as that suffered by Amazon workers), misidentification by surveillance systems, or the huge environmental impact all AI boosters seem to gloss over.
So the primary motivation of AI boosters is the con for revenue, boosting product sales and attracting investment by making us afraid we’re missing out. They use anthropomorphic language, describing AI systems as human, or even — and this one’s a favourite — superhuman, even though these systems lack genuine understanding or consciousness. They often just make up the evidence on which they base their claims, sometimes going as far as asking AI systems if they can replace human jobs.
Whether you’re a doomster or a booster, you share the fundamental belief that AI development is inevitable and that it will be autonomous and incredibly powerful. Neither group offers a clear or justifiable definition of general intelligence. We’ll know it when we see it — apparently.
But whether you’re in camp “we’re all going to die” or camp “I for one welcome our AI overlords”, you’re really in camp “show me the money”. Both views attract huge financial investments from venture capitalists and Big Tech companies, despite having no real idea how they’re going to make money.
The AI Con is a short book, and it sometimes strays into self-help territory, but its recommendations are worth noting. To fight AI hype, the authors recommend a multifaceted approach that involves both individual actions, such as critical questioning and making fun of AI praise singers, as well as systemic changes through regulation and collective action.
They suggest that when confronted by a hype merchant selling a solution, it’s crucial to understand the concrete task being automated, the input given to the system, and the output it produces because marketing tends to obscure these facts.
What’s the evidence that there’s sufficient information in the input to determine the output? And importantly, are these systems being described as human? If they are, run a mile on your human legs.
There are plenty more recommendations, but perhaps the one that will resonate with South Africans is this: who benefits from this technology, who is harmed, and what recourse do they have? Most of the time, AI is going to be about making the rich richer and making life harder for the poor and the marginalised.
It’s not about saying AI can’t do good in the world, but it is a call to resist the AI hype that clouds decision-making in boardrooms and government offices. We want AI to benefit society, not turn society into collateral damage in the enrichment of a few.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.