Welcome to the second post regarding the future of automation by looking at the past of automation. I originally wanted to cover the jobs and economic aspects in my post debunking AI doomerism. But as I’m also pretty wary of automation being a problem economically. Regardless, the idea of job automation leading to massive job replacement is a pretty constant theme of discourse and coverage so I’m going to continue with the theme of why us fleshy humans aren’t doomed after all.
As far as the economic argument goes, it’s pretty simple. Basically that that people can’t do a job as good as as a machine or an algorithm, or in this case an AI1. And that may be true for certain jobs. It seems to be getting a lot of coverage by journalists because it’s probably true for some parts of journalism. But even that could be unnecessary chicken little-ism when it could actually be a harbingers for better things to come.
For that we will go to the story of bank tellers. Prior to the invention of the ATM, you obviously had to go to a central bank branch and wait in line to deal with a person who could get cash for you. It was an objectively worse system. If you needed money on a weekend, you had better hope that you had a generous friend to float you, and remember this was before card payments were accepted so if you were out of cash, you didn’t buy anything at all.
So the ATM comes along and does that job far better and more convenient than a human can and everyone is happy, except the bank tellers whose services are no longer required, of course. But a curious thing happened that nobody really expected. Because running a branch full of human tellers was expensive, there were relatively few bank branches. But now that the ATM is able to handle all the easy stuff and the complicates stuff is left to humans, you can run a branch with one or two tellers manning the counters. The result was an explosion of bank branches to the environment we’re familiar with today where you can’t go two blocks without a branch of some bank or another and the fact that there are now so many more branches means there are now more bank tellers than before their job was “automated away”
I use that example to just show we have no idea what will happen. If an AI is able to take care of the very basics of journalism and say get police blotter, sports scores, etc… It could easily mean that local papers that have become unsustainable are viable again. Rather than fewer people, you could have an entire paper staffed with one or two humans to cover the interesting school board issues while the computer is able to do the work that would have required a team beforehand. That paper is now able to function for many small towns and create tens of thousands of new jobs. I could be wrong, but the point is, the certainty people use to talk about the problem is very misplaced.
There are plenty of examples of this sort of thing. I spoke with the head of McDonald’s for Spain and Portugal and he told me that in stores with automated ordering kiosks, employment went up. Because the previous bottleneck in production was ordering so now there were more people required in the back to increase the output. In the Iberian peninsula, McDonald’s doesn’t tend to operate on the margins of viability, but in the US in particular, it could also make restaurants possible where there wouldn’t have been any before by lowering the operating costs as well.
But let’s talk about a situation where all the jobs actually were automated away and yet nobody really thinks about it. In 1870 around 50% (opens to PDF report) of the US population worked in agriculture. In 2019, that number is around 1%. I don’t know anyone that wouldn’t argue that machines have fundamentally transformed just how many people are needed for agriculture. So here’s a case where automation really did take all the jobs away, so why didn’t it mean economic doom?
Well…for some people it did. We have to remember that while we have now come to use luddite as something of an insult for people resisting change, the real Luddites were completely correct in their narrow context and most died having been ruined by the automation that shut down their workshops. So we really do need to have empathy for people displaced and the friction that comes with that and figuring out how to deal with it. But regardless, these changes are mostly inevitable so it’s far better to be reactive than try to proscribe the changes all together.
But to the larger point, obviously society has done much better because of those jobs being automated away, so…why? It’s really down to the fact that people will still want to spend money on things and experiences, those things just change. Now that food is mostly automated and cheap2, people have decided to spend their money on nicer things3 like going to movies or even within the realm of food, by going out to eat rather than eating in. This is the general trend of goods getting cheaper and services getting more expensive because the time of the people involved becomes more and more expensive.
So the idea that all the jobs will disappear is entirely dependent on the idea that people will want to stop spending money. If there is more left over after buying ‘things’ then people will just buy more services. Why do I like drinking a beer at a bar better than at home? Because of the services of the bartender and the social interaction. Say that beer suddenly became insanely cheap, I’d still go to the bar and it would still cost more because you’re paying for the land and building and the time of the staff, both of which cannot be automated away.
But even more generally the lesson to be learned from all of this is pretty simple. The world exists as it does today because of the cost structures surrounding it. People choose X over Y because X is more value for the money for what they need it for. And jobs currently exist to supply things based on those costs. But you can’t take the assumption that AI will be a technology that fundamentally changes things and then at the same time assume the only change will be job loss. If you assume a structural change, the structure changes and the entire value chain from land and materials to consumption changes.
And just to tie that into my previous point, this is what I was really getting at with the Hayekian knowledge problem being the fundamental reason why AI won’t doom us. They just can’t know everything for all the inputs of when it’s beneficial for anyone to change their behavior.
So yes, things will change, but overall it will mostly be fine, though for sure it will bring new and different problems.
It really is important to distinguish an AI from an algorithm. The latter has clearly defined inputs and outputs and gives very predictable results.
when you complain about food costs remember that the average household spent 46%(opens in PDF) of their income on food in 1900
“Things” in this context is referring to both physical stuff and services.