AI, job loss and tax: the mind boggles

I’m no economist. But I’ve recently read Follow the Money: how much does Britain cost?, researched some Treasury statistics, and listened to a few podcasts from the Institute for Fiscal Studies. So I find myself in a sweet spot - I know enough to form opinions, but not enough to know why I’m wrong.

The first thing I’ve learnt: nothing grounds your lofty ideals for the UK like a look at our public finances. I finally get why so many people (politicians) bang on about the importance of economic growth in this country. It’s striking to see how much goes into key services which are still creaking at the seams, and how little goes towards some of the most politically contentious areas.

We rightly look out for those who would be at risk of social exclusion without state intervention. We transfer money from taxpayers - those able to work (and hopefully contribute to economic growth) - to those who need it. These are people who are sick, old, disabled, homeless, seeking refuge, unemployed, and others. In 2022/23, we spent £319 billion on what the account books call Social Protection’. This is 28% of the £1.1 trillion that the government spent in that year. If divided equally between every person in the UK, it’s just shy of £5,000 each (caveat: back of the envelope sums, and bear in mind that some of this money will only be available to those in England; Scotland/Wales/Northern Ireland manage some of their own provisions here with separate budgets).

We hear debates about the extent, viability and effectiveness of a lot of this support. We also hear worries about the future, such as how the burden of an aging population will increase over time. £137 billion went to supporting the elderly in 2022/23, for pensions and personal social services alone (in other words, not even considering something like the increased use of NHS services). This was up from £128 billion in 2018/19. A relatively modest rise, perhaps, but it’s a number which will probably keep rising. Our retired population is set to swell, and we don’t have the working reinforcements coming up the population pyramid to cover the whole bill.

Paul Johnson, Director of the Institute for Fiscal Studies and author of the aforementioned Follow the Money’, worries about this in the book, along with climate change and our capacity do deal with big shocks. But there’s another worry which he doesn’t mention, but I think it could easily eclipse the impact of all of these in the next decade or two. It sounds almost clichéd to say - the incredible and growing capabilities of AI (artificial intelligence).

I’ll sketch out the argument for now. Each step is contestable, and merits deeper investigation than my few-hours-long writing session today. But here’s a starter:

  • AI capabilities have improved and accelerated in their improvement over the last few years. Expert predictions for when we are more likely than not to see human-level or superhuman AI seem to have trended earlier. Those at the forefront of the field seem to often talk now of dates before 2030. This is a next-5-years thing, not a next-50-years thing.
  • How does this affect the job market? Probably a few thousand pages in working this out. For the sake of simplicity, let’s say it seems likely the supply of jobs could shrink, dramatically. Companies who use or deploy AI well will be able to deliver better results than their competitors; human workers will probably cost more than AI workers, and be less effective. It’s not immediately clear to me that there will, by nature, be much inertia in organisations realising this, and adjusting their workforce accordingly (read: lay a lot of people off). I’ve already heard anecdotes of coding and creative jobs taking a hit. This transition will not be universal, but it doesn’t need to be for us to see masses of jobs disappear without replacement. I also think this transition is somewhat unavoidable - for it not to happen, we would probably need a global agreement now to stop further AI development. From what I understand this isn’t feasible, or something you could get everyone on board with.
  • So we see many more people unemployed within the timeframe of the next UK government, or maybe the one after. That’s fewer people paying taxes, and more people needing support. The viability of the state safety net suddenly looks in doubt. I guess the extent to which this is true depends on the extent of AI-induced job loss; and the extent to which those other factors mentioned earlier (aging population, climate change) unravel in our society.

Maybe Paul Johnson doesn’t give this a mention as we can assume the economy is still growing in this scenario, as these AI-powered organisations generate astronomical profits. Then the question becomes: can the lost taxes from millions of workers be replaced by a reasonable tax burden on companies, one which balances international competitiveness with paying the social protection bills? The concentration of taxpayers - away from millions of people towards far fewer organisations - would still appear to be a fundamental shift in our economy, our politics and the nature of our welfare state.

These arguments and best guesses are clearly uncertainty stacked upon more uncertainty. I’m more confused now than when I started writing. But amidst this uncertainty, this surely merits serious attention. Firstly, I should sure up my understanding of the bullet points above. But then, if they are anywhere near correct, more and more questions follow. How would we effectively tax organisations who use or deploy AI, and the individuals who earn a lot from AI? Can AI also force the cost of public services down, and when does that happen in relation to the job loss? What do people work towards, go to school and university for, if jobs just aren’t there in the same number any more?

Then again, if the bullet points above are anywhere near correct, this is not the only fundamental restructuring of society that we will need to consider upon the arrival of human-level AI. The scary thing is that, in that world, this - pensions, the welfare state, the support schemes which millions in the UK rely on - may be some distance down our list of problems.

June 20, 2024