Why (most) European workers are right to love AI

Fears that automation will destroy jobs and radically transform the global economy are as old as capitalism itself. Concerns that robots could soon surpass human intelligence – and potentially enslave humanity – are also several decades old.

Why should we be afraid now? The reason, according to many “experts” and reporters, is the rise in everyday use of artificial intelligence, in particular large language models (LLMs) such as ChatGPT.

Indeed, barely a day goes by without some Terminator-loving hack or self-proclaimed “futurist” arguing in some influential media outlet that AI will quickly render the West’s workforce permanently unemployed.

As a recent Forbes headline bluntly put it: “AI will take your job. Get over it.” (Full disclosure: I also love Terminator.)

Surprisingly, however, European citizens seem largely immune to the doom-mongering. A recent survey conducted by the European Commission found that 60% of EU citizens believe that robots and AI actually have a positive impact at work; 70% said AI boosts their productivity.

Whence the ambivalence? Have EU citizens not gotten the memo? Are they secretly working for Skynet?

As it turns out, Europeans’ scepticism about AI’s work-related dangers is likely well-founded. Several recent studies have shown that, although some professions are indeed at risk, the overall impact on employment will probably be negligible, at least for the foreseeable future.

One study released this week by the Economist, for instance, found that the number of US interpreters and translators – jobs which LLMs are typically thought to put in serious jeopardy – has actually increased by 7% over the past year.

The study also found no negative impact on other professions thought to be at risk, such as sales or financial services. It also noted that white-collar work’s share of total US employment has risen slightly over the past year.

Other research has reached broadly similar conclusions. The Financial Times’ chief data reporter, John Burn-Murdoch, recently noted that insurers, secretaries, accountants, and travel agents have not shown any sign of AI-triggered disruption in the US.

Indeed, even those who claim that AI will radically transform our lives often admit that it is unlikely to do so in the near future.

In an interview this month with the New York Times, Daniel Kokotajlo, the co-author of the recent much-cited “AI 2027” paper which argues that “artificial superintelligence” will be developed by the end of this decade, admitted that “most jobs will be safe” for the next year and a half. (Call centre workers and coders are the exceptions, he said.)

Where’s the worry coming from?

This raises two important questions. First, why aren’t more jobs currently being eliminated by AI? Second, given AI’s relatively limited negative impact on employment, why is there so much discussion of it?

The first question likely admits multiple explanations. The Economist's study suggests two. First, many companies still don’t use AI for “serious” work. Second, many firms appear to be using AI to boost workers’ productivity (and, hence, companies’ profits) rather than to lay them off.

Arguably, there is a third, more obvious reason. Many of the jobs that AI can do “well” remain far below the proficiency of a human being – as anyone who has experienced simultaneous AI-interpreting (as I recently did) can testify. (Translations of poetry and metaphors are especially bad – and often hilarious.)

Of course, many people, including Kokotajlo, argue that AI is likely to become significantly more proficient at such tasks over the next couple of years. But is he right?

Many cognitive scientists remain sceptical: some even argue that LLMs, which are essentially data-crunching predictive machines, are structurally incapable of developing anything close to human – let alone superhuman – intelligence.

Moreover, Kokotajlo’s quasi-apocalyptic predictions are difficult to square with his own repeated admissions that “the future is very hard to predict”.

The claim that a superintelligence is imminent is simply “a best guess, but we have a lot of uncertainty,” he told the NYT. “It could go faster, it could go slower."

In other words: Kokotajlo, like pretty much everyone else, has absolutely no idea what the future will bring. (In the interview, Kokotajlo also pushed back his prediction of the dawn of superintelligence from 2027 to 2028 – thus granting us a one-year reprieve from robots’ total domination.)

Am (A)I to blame?

In fact, the reason for the prevalence of the discussion of AI's impact on work is likely more mundane – and slightly sordid.

As Burn-Murdoch points out, journalists, along with coders, are one of the few professions that are indeed likely to be affected by the large-scale rollout of LLMs.

The reasons for this are twofold. First, journalism – at least of the written form – largely consists of “nice, clean, linear and sequential tasks, exam-style questions and essay assignments” that ChatGPT and other LLMs excel at.

Second, much journalistic work is done on a freelance basis, which allows an LLM to be easily “swapped in for a non-staff copywriter without HR getting involved”.

Whatever the reasons, the conclusion appears inescapable: by reporting so much about the potentially harmful impact of AI on workers, we journalists may not actually be reporting the objective truth.

We may, instead, simply be projecting our own fears.

Economy News Roundup

A federal appeals court on Thursday reinstated Donald Trump’s most sweeping tariffs, a day after a trade court had ruled the US president had exceeded his authority in imposing the duties. Read more.

The judicial drama came just days after Trump agreed to delay the imposition of 50% tariffs on the EU until from June 1 to July 9. Read more.

Trump’s threat also came amid fears that he would seek to force the EU to adopt a more hawkish line on China in order to avoid the tariffs. Read more.