AI’s real threat: white-collar jobs vanish faster than policymakers notice

There is nothing new about fantasy doomsday scenarios triggered by the rise of artificial intelligence (AI): robots who can think for themselves, turning on humans and killing or enslaving us. And while we are all being entertained by those, it does not seem like anybody is noticing the very real impending apocalypse. This is not sensationalist. AI is coming for our jobs.

“Yes, but you still need a human to prompt the AI,” the retort goes. The other popular soundbite is “AI won’t replace human beings; human beings using AI will replace human beings.” Both, and countless others like them, would be true if we were still talking about the current free version of ChatGPT, which is what most of us are using. AI has advanced far beyond that, and it is advancing at a blistering speed.

Anthropic CEO Dario Amodei predicts that 50% of all entry-level white-collar jobs will be wiped out by AI over the next five years. Is that just another attempt for free marketing from one of the biggest names in AI?

About 18 months ago I was at an AI conference and a professor in attendance reported that a major South African law textbook company was experimenting with training a large language model on its entire database. So, think ChatGPT, but trained entirely on all South African law, ever. Why would partners in law firms hire expensive associates to do research when they can just use an AI? Why would a company employ a general counsel when they could subscribe to a website at a modest fee instead?

“Garbage in, garbage out,” yet another retort goes, because our concept of AI is the AI we played with eighteen months ago. Go see for yourself: AIs now produce the right outputs on the basis of pretty basic inputs. Prompt engineering was touted as the next big profession a few months ago. I can’t see it any more. AIs understand English perfectly well, and are only getting better at interpreting what it is that you want.

Hassle and expense

Until a few years ago, my brother helped build a cloud-based payroll company. The idea is simple enough. You subscribe to the service of the company at a modest monthly fee, and all your payroll issues are taken care of by the app. To be clear, that company does not use AI, but it had already prompted other companies to question why they needed the hassle and expense of a whole payroll clerk or department, when they could just subscribe to the cloud-based payroll company at a fraction of the cost.

Well, AI is going to annihilate any payroll clerk jobs that are still around, and it is also going to annihilate the cloud-based payroll company. Why subscribe to a cloud-based payroll company when an AI can do the same tasks at a fraction of the subscription fee? If you follow the markets in the US, you will know that over the past few weeks investors have been dumping software company shares for this very reason. Forbes reports that not all software-as-a-service companies will die. No matter how you cut it though, it is not pretty.

Accountants, administrative staff, journalists, data capturers, sales agents, translators, proofreaders: we’re all in the firing line. I have zero programming training, but I’ve used AI to build working apps. That means computer programming is not the bright career path it was only a few years ago. Investors have long seen the logic of shifting customer care operations from countries like the UK. Why pay somebody in pounds when you can get the same service from someone who would be paid less in rands? The result has been the much-celebrated jobs in call centres, as in Cape Town.

Well, the investors’ logic has a further implication: why pay someone in rands when you can pay an AI a few cents.

And then, in a recent podcast, Elon Musk explained much about the advancement of humanoid robots. Do yourself a favour: isolate Elon’s politics for a moment and listen to what he has to say about the direction of AI, from self-driving cars to building a colony on Mars.

Five years is not a long time – the Covid-19 pandemic was six years ago. Extremely well-funded companies are racing towards building the best humanoid robots.

Think about that.

Where human factory workers are late for work, need sick leave, want pay increases, are impossible to fire when they misbehave, and so on – humanoid robots have no such issues. You charge them, service them and they work. In fact, where the humans working in a factory only work eight-hour shifts, humanoid robots can work 24/7. Why would factory owners employ human beings?

Human digital emulation

Elon says he would “be surprised if by the end of [2026] human digital emulation had not been solved”. And his plan is to build and sell millions of these robots that can emulate human beings so that they are cheaper to buy than a small car. Elon is not the evil genius behind all of this. This is going to happen whether it’s Elon doing it or someone else.

So, if you thought Covid-19 six years ago was bad, wait and see what the next five years are going to be like. The scale of disruption is going to be apocalyptic.

What makes things especially bad is that nobody is even talking about this in SA. Our government is notoriously and perpetually distracted. The ANC is fixated on President Cyril Ramaphosa’s successor, oscillating between the hopeless and the pointless. The DA is trying to convince everyone that its Uber Eats line item is not a monument to the much bigger crisis in the party that everyone thinks it is. And then there’s former president Jacob Zuma and his band of merry men, Julius Malema, Herman Mashaba… None of them are going to give us the leadership we need.

And yet we need to take this seriously. We need a Just Transition 2.0 that can ensure that, as with the shift to a green economy, the shift to an AI-driven economy is human-centred and just. We need to ensure that AI adoption in companies is regulated with clear policy and legislation – that workers are trained to use AI before their jobs are made redundant because of AI, for example.

Moral consciousness

We need the same kind of moral consciousness that we see directed at protecting the environment being directed towards protecting people’s jobs. We need special tax deductions for companies that employ human beings where they could have used cheaper AIs instead. And so on.

We need to wake up our policymakers and legislators and get them to give this the attention it started demanding five years ago, a demand all of us ignored. The City of Cape Town – the so-called light on the hill – regularly celebrates the operationalisation of call centres and the creation of customer care jobs. Mayor Geordin Hill-Lewis recently noted again that “over 100,000 Capetonians now [work] in the sector”. What do they say the future holds? Cape BPO CEO Clayton Williams explained that they “have a…vision of locating [an additional] 20,000 seats across the Cape Flats over the next 10 years”.

That’s not happening. DM

Wade Seale holds a PhD in philosophy and is Africa AI Ethics Carnegie Corporation New York Postdoctoral Fellow at the Institute for Humanities in Africa (Huma) at UCT.