What will AI do to jobs?
There is a huge gap between what current models are capable of, and how they’re being currently used. AI is going to automate jobs away; AI is also going to increase productivity, and to create new jobs.
The good news. Take programming, for example: if one AI-assisted programmer can do the work of 10 StackOverflow-assisted programmers, then we will build much more software. This is the Jevons Paradox. So programming teams may shrink, but they will proliferate. And we’ll also need many more prompt engineers and product managers. So, perhaps fewer programmers per team, but more teams, and work created for other roles. We might imagine that the same will be true for many of the other fields where AI-assistance will turbocharge productivity. There will be orders of magnitude more writing, more lawsuits, more art, more architecture.
The bad news. Technological unemployment is going to be sweeping and faster than anything we’ve seen before. It won’t be obvious in the next year or two, but it will be stark by the 2030s. This is terrifying. Even if new jobs are being created, there will be a lag, and the people who lost their jobs may not be qualified for the new jobs. People will move towards jobs in lingering niches where AI can’t yet compete, perhaps involving physical interactions (like hairdressing and plumbing), and rich human interactions (like therapy and care work). And worse still, if the pace of AI continues to speed up, then even those new jobs will probably be automated (to a 'good enough' degree) before long.
Will there be a limit to how far AI improves? In the long run, no. In the short run, yes there will be limits to how far scaling data & compute will get us, but still exceeding most knowledge workers for most tasks. So on balance, the tsunami of technological unemployment will outpace Jevons Paradox.
What hope is left? The optimistic, post-scarcity vision is that we fund Universal Basic Income through taxation. But that feels like a global optimum, with many local pessimums in between - in other words, it's hard to see how we get there from here, without a few rapacious Silicon Valley winners taking all.