What cars did to your city, AI will do to your brain

Once again, I rage against the dying of the light. Also this week: some maps to show that England is not concreting the countryside; and the longest conveyor belt in the world.

There’s a concept in transport planning called “induced traffic”, which is essentially the evil twin of “if you build it, they will come”, and which is best explained with an example. Imagine two points that lots of people travel between but which, at current congestion levels, are an hour apart. Wouldn’t it be great, some agency somewhere concludes, if they were only half an hour apart? So, the authorities spend a lot of money on widening the road, to increase its capacity and cut journey times. That’d decrease congestion, right?

Wrong – because the number of people who think that journey worth spending half hour on is a lot higher than the number who thought it worth spending an hour on. So the traffic increases, the new lanes fill up, and in all probability the journey time will soon enough end up closer to an hour than 30 minutes again. And the local authorities just spent a lot of money and gave up a lot of land to do it.

This is an over-simplification: it doesn’t mean roads are never worth building or widening; providing space for extra journeys is often a good thing. But the general principle holds: journey time and demand are not independent variables. This is why certain American cities constantly seem to add road lanes yet traffic jams never seem to get any better.¹

Anyway. I keep getting ads from AI companies which claim they can give office workers hours of their day back by automating their emails. Based on everything I know about the phenomenon of induced traffic, I absolutely do not believe this to be true.

To be fair, I don’t have a job, and I don’t have a boss: relatively little of my inbox is given over to largely pointless busywork. There are some types of emails – requests for documents; factual questions from people who could have just googled it – where I can imagine AI could be helpful.

Nonetheless, it feels to me that the most time-consuming emails are the sort you can’t automate. There are the ones where you actually have to make a decision (AI can check your calendar to see if you’re free; it can’t decide whether it’s worth taking the meeting in the first place). There are plenty of others where, if you’re going to send someone an answer, you’ll really want to have read the bloody thing before signing your name to it. There are things you can’t outsource to a robot.

Worse, the existence of induced traffic tells us that, if you make it easier to send emails, more emails will be sent. More people will request things they already have, because it’s easier than looking for them. (They may even set their own AI assistants to do it automatically.) An AI will respond to an email, because that’s what it’s been told to do, then the AIs that receive it will respond in return.²

The result, I fear, is going to be a lot more emails: robots chatting quite happily to one another, unobserved by human eyes. Inboxes will become even more unmanageable, as it becomes harder and harder to find the stuff you actually want in the ever multiplying pile of pointless emails. Meanwhile, the real work – where decisions are made, action points issued, a human brain is required – will surely move to the places where the AIs have yet to take over. Maybe it’s to Slack or a project management platform (I’ve no idea, I’ve not set foot in an office in half a decade). Or perhaps meetings, which so often feel less like work than the thing that gets in the way of doing it, will become work again after all.

That’s not the only lesson urban planning has for AI. One of the biggest differences between cities built for pedestrians and those built for cars that I’ve spotted in my years of wandering round them trying to figure out how they work is the role of serendipity. In a human scale city, a place built around public transport or a walkable centre, you can wander round and bump into things, find the little shop or bar or pretty street you didn’t know to look for.

You can’t do that when driving: you pass things so fast you’ll miss them. More to the point, because there’s less footfall, fewer of the things spring up anyway. (Compare the street life of the main routes through inner London with those of equivalent bits of the more car-based cities I’m too polite to name³.)

It feels to me like AI might do something similar for the human brain. The sort of stuff I write – factual but intended to entertain, rather than just inform – requires reading lots of stuff and seeing what grabs me. Sure, I can ask an AI to give me a bunch of facts about [insert topic here] – but it won’t spot the striking coincidence, the funny detail, the historical anecdote I never knew was there but which I suddenly realise is the framing device for my entire story.

As with emails, the work isn’t really in the writing: it’s in the decision making and selection process, all the bits you can’t outsource. And just as cars remove serendipity and thus life from cities, AI will do it from research and writing.

To put that another way, what cars did to our cities, AI has the potential to do to the human brain.

I’m really not sure that that’s a good thing.

Self/non-self promo

Big week for fans of the late lamented Paper Cuts. Firstly, Miranda Sawyer has a new podcast out: it’s called Talk ‘90s To Me and episode one is about Oasis.

Secondly, this week you can apparently get 25% off preorders of the new version of The Compendium of (Not Quite) Everything at Waterstones, if you so wish, and why would anyone not? Code is “SUMMER25”. Runs til Thursday.

Your Regular Reminder That No, Britain Is Not Concreting Over The Countryside

One of the arguments you hear most from pro-homelessness campaigners, as I sometimes think we should start referring to NIMBYs, is unnervingly similar to one we hear from anti-immigration campaigners: to whit, that the country is full. It was all very well building homes in the past, the logic seems to run. Back then we had room to spare. Today, alas, there is not a single blade of grass left anywhere in England.

This is, of course, nonsense, as a glance at either a satellite photo or out of a plane window will tell you. If it sometimes feels like the opposite that is generally because many of us spend most of our time in cities or towns, or on the roads or rails that link them: the vast tracts of land without people in them are, by definition, the bits that people are least likely to see. All that’s a bit anecdotal, though, so it’s quite nice that someone with vastly more technical skill than me has done the hard yards to prove it.

Russell Curtis is an architect and housing campaigner, whose work I’ve featured before. (He’s the guy who periodically likes to point out quite how absurd it is that 3% of London is golf course.) Late last year he made this map of how many homes England’s councils need to build per hectare of land to meet the government’s housing targets: