Will AI widen inequality between workers?

{"text":[[{"start":null,"text":""}],[{"start":7,"text":"Welcome back to the AI Shift, our weekly dive into how AI is shaking up the world of work. For today’s edition we are digging into some of the insights from our new poll of how thousands of workers in the US and UK are using AI in their jobs, which we carried out with research firm Focaldata."}],[{"start":24.6,"text":"The headline finding is that the highest-paid and most well-educated workers are using AI in their jobs much more than lower-paid workers, risking an exacerbation of existing divides in productivity and earnings. But there were a lot more nuggets buried in the data which we will get into below, as well as some important nuances to keep in mind when thinking about the inequality narrative."}],[{"start":46.900000000000006,"text":"How can we reconcile the finding that AI usage may increase inequality with studies suggesting it levels the playing field? (John)"}],[{"start":55.35000000000001,"text":"One notable thing about our results is that the inequality story contrasts with the consistent finding from experimental studies of call centre workers, writers and management consultants that AI usage tends to reduce inequality between more- and less-skilled workers, tending to provide a bigger boost to speed and quality for lower-ability workers than their more competent co-workers."}],[{"start":81.10000000000001,"text":"I think the two findings are consistent, it’s just that they are capturing different things. First, our survey is measuring the impact across the economy while the experiments are measuring the impact within individual firms and occupations. And second, the experiments implicitly assume universal adoption of AI at work whereas our survey finds the reality is unequal adoption. If knowledge workers use AI more than workers with routine white- or blue-collar jobs, and if AI augments many knowledge work jobs while eroding more routine roles, then AI can narrow inequality among call centre workers and among management consultants, respectively, while simultaneously widening the gap between the two professions."}],[{"start":122.80000000000001,"text":"People are more worried about job losses for everyone else than for themselves (Sarah)"}],[{"start":128.25,"text":"This is something I’ve wondered about before, and it’s interesting to see it confirmed in the polling data. When people were asked what they thought AI would do to the number of jobs in their own industries over the next five to ten years, they were somewhat pessimistic: 38 per cent thought AI would lead to more jobs lost than created, while 17 per cent thought the opposite. (The most popular answer was that it would not make much difference). But when people were asked the same question about the economy as a whole, the responses were much more gloomy: 51 per cent thought more jobs would be lost than created, compared with 18 per cent who thought the opposite."}],[{"start":165.15,"text":"What to make of this? My suspicion is that when you’re inside a job or industry and know it intimately, you know how hard it would be to automate it away completely. In contrast, it’s very easy to make assumptions about the automatability of other people’s jobs. I don’t know any lawyers who think their job is entirely replaceable by technology, for example, but I know plenty of non-lawyers who reckon that lawyers are surely going to be wiped out. For what it’s worth, I think people’s predictions about their own industries are probably more interesting and useful, given how much more information they have about them."}],[{"start":198,"text":"There’s been a surprising lack of AI training in the workplace so far (Sarah)"}],[{"start":202.6,"text":"I was struck by the fact that only 14 per cent of workers in our survey said they had been given formal training on AI by their employers so far (while a further 21 per cent have been issued with informal guidance). Of course, there are plenty of companies or roles for which AI is not particularly relevant. But even so, that feels like a low number, given that — based on our survey responses — about a quarter of UK workers and almost a third of US workers are using AI tools on any given workday."}],[{"start":236.35,"text":"I’m not really sure what to make of this, given that it’s more than three years now since ChatGPT was released, and providing workplace training is not a particularly expensive intervention. Perhaps it’s an indicator that many companies are still trying to figure out how and whether to integrate AI into their workflows at all. If any readers have more insight, drop us a line."}],[{"start":257.65,"text":"Coders and writers seem to be using and thinking about AI quite differently (John)"}],[{"start":263.45,"text":"The great thing about having detailed data on how thousands of workers are using and thinking about AI is that we can drill down to jobs involving different types of tasks. One fascinating finding here was that workers whose jobs involve writing code (primarily developers and analysts) report some of the most positive feelings about AI in the workplace, while professional writers (a group including copywriters, content creators, translators, writers and editors) are more negative about AI than any other professional group, including blue-collar workers and those who don’t use AI at all."}],[{"start":298.09999999999997,"text":"This matches up with anecdotes I’ve heard over recent months, and it struck me that perhaps one thing that’s going on here is that coding and writing sit at very different places in the bundles of tasks that make up each type of role. In software and analyst roles, writing code has generally been a routine and laborious part of the job, and a means of getting from idea to output. As such, using AI to automate coding takes away a dull part of the job and frees up time for coming up with and executing on ideas. Whereas for writers, the writing is the output — the thing the worker takes pride in — as well as often being the place where the cognitive and creative work happens. Automate that away and there’s not much left."}],[{"start":341.9,"text":"A word of caution on the high-pay/low-pay divide (Sarah)"}],[{"start":346.5,"text":"Our news story on FT.com today leads on the fact that the highest-paid people are using AI in their work much more heavily than people lower down the pay and experience ladder. One obvious possibility is that — if these workers are using AI successfully to supercharge the volume or quality of their work — this will lead to a growing divide in pay and productivity. But we should also remember that self-reported productivity gains need to be treated with caution. In the world of software development, we are pretty confident that senior developers are indeed now seeing boosts to their output thanks to AI coding agents. But outside that realm, I don’t think we have solid evidence just yet."}],[{"start":388.45,"text":"We have also seen examples of very highly paid workers relying too heavily on AI to the detriment of their reputations and the quality of their work. (Just this week, the elite law firm Sullivan & Cromwell told a US federal bankruptcy court that a major filing it made in a high-profile case contained multiple “hallucinations” made by AI software.)"}],[{"start":410.5,"text":"Ioana Marinescu, associate professor at the University of Pennsylvania, makes another point worth bearing in mind: these well-paid AI users might be “winners” for now, but “losers” in the long run. “The greater use of the technology might imply greater vulnerability to eventual replacement by AI,” she noted in an email to our colleague Madhumita Murgia. "}],[{"start":432.2,"text":"One possibility her research has explored is that “intelligence work (of the sort that can be done fully remotely) will plausibly be largely replaced by AI, while workers will move to physical (i.e. in person) jobs.”"}],[{"start":446,"text":"Recommendations"}],[{"start":null,"text":"

1. Together with our brilliant colleague Madhumita Murgia, the FT’s AI editor, John and I will be doing a live “Ask the Expert” Q&A session today (Thursday). If you’ve got a question for us, you can submit it here, and check out the discussion live at 1pm BST. (Sarah)

2. Former Biden economic adviser Ernie Tedeschi has a great article for Stripe Economics on what the decline of the travel agent tells us about how technology-driven job displacement can play out (John)

"}],[{"start":null,"text":""}],[{"start":454,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1777186411_5567.mp3"}