It is a question that will not go away: will artificial intelligence take all or many of our jobs — and, if so, will we need some form of government-provided income to avoid mass unemployment and mass destitution?
Every day brings a new wave of AI stories and market reactions to AI developments, and none of them are particularly reassuring. Lord (Jason) Stockwood, an investment minister in the government, created more headlines than he probably expected when musing to the Financial Times recently that “some sort of UBI [universal basic income]” could be needed to “soft land those industries that go away”. Also required, he said, might be lifelong-learning mechanisms to retrain people whose jobs are replaced by machines.
Everybody will have shuddered when reading a long essay, “The Adolescence of Technology” by Dario Amodei, chief executive of the leading AI firm Anthropic, last month. It is a thoughtful and wide-ranging piece, taking as its basis that, before very long, machines will be able to do everything better than humans, that 50 per cent of entry-level positions will disappear within the next five years, and that job displacement will occur on an epic scale.
• Workers ousted by AI could receive universal basic income
This is where the idea of a universal basic income comes in. People who have no jobs will still need an income and the existing welfare system will not be sufficient to provide it.
Amodei also urges us all, however, to avoid what he describes as “doomerism”. Among the consequences of AI adoption, he suggests, will be that economies should be able to sustain growth rates of between 10 and 20 per cent a year — something that we have never before seen. If the machines are creating all that income and wealth, the task will be how to fairly distribute it.
He has thrown down quite a gauntlet, and addresses head-on the criticism thrown at him by economists over his previously expressed views in this area. This is that he has fallen for what is known, and has been long known, as “the lump of labour fallacy” — the idea that there is a fixed amount of work to be done in the economy and, if automation does more of it, there will be fewer jobs for humans.
The fallacy arises because the wealth and productivity gains generated by technology in the past have themselves generated jobs and new occupations that we would not have previously imagined. Amodei argues that AI is different because, among other things, of the speed of its adoption and its cognitive ability. In other words, we have never had a technology quite like this.
What should we make of this? Are most of us facing a life on government support, and would that support come from taxing the firms and the minority of individuals scooping up the gains from AI? Although some of Silicon Valley’s biggest names are fond of talking up the possibilities of a universal basic income, they are not in general stepping forward to offer up the potentially eye-watering taxes needed to pay for it. For, make no mistake, a UBI would be very expensive.
The Institute for Fiscal Studies recently pointed out that a UBI for every working-age adult equivalent to the £400-a-month universal credit (which of course is not universally paid) would cost more than £200 billion a year.
When you play around with this, the numbers soon become huge. A UBI equivalent to average earnings of about £3,000 a month would cost £1.5 trillion a year, which is more than the government currently spends on everything (£1.37 trillion).
• Tech barons are warning of AI doom, so why don’t they slow down?
But we may be getting ahead of ourselves. Is AI going to destroy not just some jobs but all but a tiny number of them, resulting in mass unemployment? A large degree of scepticism is in order, and not just because of the lump of labour fallacy.
Employment has levelled off in both America and Britain, after growing strongly in the 2010s. In America, though, this appears to be more of a tariff effect, undermining business confidence, while in the UK it is largely the result of the rising cost of labour because of increases in employer national insurance, a higher minimum wage and enhanced employment rights. An AI effect is hard to discern, except possibly in the technology sector itself.
UK unemployment has risen, and may do so further, but some of this is due to declining economic inactivity (people who were not available for work are joining or rejoining the job market) and those cost-of-labour effects. Historically, the unemployment rate remains quite low.
If slow adoption of artificial intelligence and the creation of new jobs to replace most or all of those destroyed argues against mass unemployment — though that may merely be a comfort blanket — it is undeniably the case that jobs will be lost as a result of AI.
If you wanted to protect people losing those jobs, however, a UBI would not seem a sensible way of doing it. Something rather different and more targeted, a guaranteed minimum income, would work better. In the current context, though, when we are looking to curb rather than add to a soaring welfare bill, it is clearly best for the government to keep its powder dry.
You might say at this point that a cheesy thing to do would be to ask AI what it thinks. So, never one to pass up on the obvious, I did — by consulting ChatGPT.
Its response, I am glad to say, was remarkably sensible. A universal basic income, it said, “could become necessary if AI displaces large numbers of workers faster than the economy creates new roles, and governments decide that traditional welfare isn’t enough to maintain social stability and demand”.
But: “It might not be needed if job creation keeps pace with AI deployment, other policies (education, reskilling, targeted support) successfully cushion transitions, and new economic models evolve that let people work alongside AI in meaningful ways.” And: “AI will change work — but experts disagree on how much net job loss it will cause.”
You might say ChatGPT is talking its own book — that the last thing the machines want to do is stoke fears of AI generating mass unemployment. It seemed to me, though, that it provided a fair summary of the debate. This is no time to be panicked into the unaffordable.
PS
The challenge I set last week was to find a higher marginal rate of tax than the 20,000 per cent identified by reader David Macdonald, in which £1 of extra savings income resulted in £200 of additional income tax.
Some people said that if the extra savings income was a fraction of a pound, and still resulted in £200 of additional tax, you could get above 20,000 per cent. I thought, however, that this was not in the spirit of the competition.
But there were “genuine” marginal tax rates above 20,000 per cent without resorting to fraction of a pound. Keith Butler, who describes himself as a semi-retired chartered accountant, goes one better with this example. Taxpayer X has income of £50,270, giving a tax liability of £7,540. But because he is a basic-rate taxpayer, his wife is allowed to transfer 10 per cent of her personal allowance to him, giving a reduction in his tax of 20 per cent of £1,260, or £252.
If, however, his income increases by £1, he becomes a higher-rate taxpayer and the ability to transfer is withdrawn — so his extra £1 costs £252 in additional tax. This, he points out, equates to a marginal tax rate of 25,200 per cent. Don’t retire completely just yet, Keith.
Ed Wood of the wealth manager Rathbones, whose email tells me he is a fellow of the Personal Finance Society, applies the same principle to a taxpayer earning £125,140, the higher-rate limit, whose savings income rises from £500 to £501. At that point, the £500 tax-free savings limit is lost and, as he points out, things then get pretty gruesome: £500 of savings income is taxed at the 60 per cent marginal rate, which applies between £100,000 and £125,140, while the extra £1 is taxed at 45 per cent, meaning total additional tax of £300.45, or a 30,045 per cent marginal rate. Excellent, or perhaps not.
The real point here, made by many people, is not just that it is possible to come up with crazy marginal rates in specific circumstances, but that the cliff edges and distortions in our over-complicated tax system throw up unavoidable tax rates of 60, 70 and even 80 per cent. It is not just income tax, either, that creates this problem.
Oh for a simpler and less disincentivising system.