Yanis Varoufakis

Last week brought a rare good-news story: Artificial intelligence enabled researchers to develop an antibiotic capable of killing an exotic superbug that had defied all existing antimicrobial drugs. An AI-driven algorithm mapped out thousands of chemical compounds in key proteins of Acinetobacter baumannii, a bacterium that causes pneumonia and infects wounds so severely that the World Health Organisation had classified it as one of humanity’s three “critical threats”.

Once the mapping was done, the AI proceeded to invent an effective drug with novel features compared to existing antibiotics. Without AI’s help, the life-saving antibiotic would remain a pipe dream. It was a scientific triumph for the ages.

But there is a nasty flipside. Remember Chris Smalls, the Amazon warehouse worker who organised an employee walkout from the company’s Staten Island, New York, facility to protest working conditions during the pandemic?

Smalls shot to brief fame when it was revealed that, having fired him, Amazon’s rich and powerful directors spent a long teleconference planning to use character assassination to undermine his cause. Still, a couple of years later, Smalls successfully organised the first (and still only) formally recognised Amazon employees union in the United States. Today, such successes are imperiled by the same AI technology that produced the germ-busting antibiotic.

Smalls’ union was a bitter setback for Amazon managers, who had been trained for years to use any means, fair or foul, to prevent workers from unionising. In a training video leaked in 2018, managers were coached to watch for warning signs of organising activity. They were urged to use surveillance cameras outside Amazon’s warehouses to spot employees who linger after their shift, potentially seeking to persuade colleagues to join a union. They were also encouraged to eavesdrop on employees’ conversations, listening for phrases like “living wage” or “I feel drained”.

Soon after, software replaced, or at least aided, the bosses’ primitive surveillance methods. In 2020, Recode reported that Amazon had purchased the geoSPatial Operating Console (SPOC) to monitor workers prone to unionisation efforts. And Vice exposed how Amazon’s human resources department monitored employee listservs and Facebook groups to predict work slowdowns, strikes and other collective action.

The software categorised worker traits and behaviours according to whether they were correlated with pro-union tendencies. But the software’s predictive power disappointed Amazon, so the company continued to rely on regional managers keeping tabs on workers the old-fashioned way.

All that has now been eclipsed by AI. Why keep an eye or an ear trained on employees, or purchase software to read their posts and Facebook pages, when a centralised AI can detect union-friendly phrases and behaviors in every Amazon warehouse automatically in real time and at zero cost?

Disconcertingly, union-busting AI relies on exactly the same scientific breakthroughs that yielded the germ-busting AI. Before AI, researchers categorised molecules as vectors that either contained or did not contain certain groups of chemicals. This was no different, and no more efficient, than Amazon’s SPOC software categorizing employees on the basis of their perceived temptation to form a union.

AI germ-busting programs, in contrast, rely on neural networks and machine-learning models capable of exploring chemical spaces that human researchers would need decades to survey. They are then trained to analyse the molecular structure of a germ’s proteins and to identify compounds with a high probability of killing it.

The AI union-busting programs rely on the same process. The only difference is that, instead of chemical spaces and molecules, AI explores warehouse spaces to focus on employees, whose real-time data is constantly uploaded to the program by the electronic devices they must carry everywhere they go in the workplace, including the toilet.

These AI-driven systems learn how to devise strategies to neutralise their programmed target, whether it is a bunch of proteins at the heart of a germ or a band of workers in the break room. In both cases, AI categorises its targets into vectors which are subsequently used to maximise the probability of eliminating them.

It was inevitable. Humanity proved brilliant enough to develop AI algorithms capable of fully decoding a killer bug’s proteins, without any human input, and creating an effective antibiotic. Was there ever any doubt that conglomerates like Amazon would seize upon this opportunity to identify, and shrink, workplaces along their supply chain where AI predicts a higher probability of unionisation?

Economists earnestly profess that the forces of demand and supply work reliably to ensure that technological change benefits us. This fiction allows them to avert their gaze from the vicious class struggle going on under their noses, wrecking the lives of millions while rendering the macroeconomy unable to generate (at least without untenable levels of debt) enough demand for the goods that the technology can produce.

Warren Buffett, who owes his success largely to ignoring economists’ illusions, famously quipped that the class war is real and that his class is winning it hands down. That was before algorithm-driven digital devices replaced foremen on the shop floor, dictating a pace of work and a total surveillance regime that made the factories in Charlie Chaplin’s Modern Times look like a workers’ paradise. As if that were not enough, AI is now empowering conglomerates to snuff out the only institution able to give workers a modicum of power in a world where they have next to none: labor unions.

The class war Buffett acknowledged will soon pit AI-clad cloud-based capital in every sector against a worldwide precariat free only to lose and lose again. Whatever one’s politics or aspirations, it should be clear that this economy is both unspeakable and unsustainable.

Yanis Varoufakis, a former finance minister of Greece, is leader of the MeRA25 Party and professor of Economics at the University of Athens. Copyright: Project Syndicate, 2023.

www.project-syndicate.org