The UK Treasury’s Autumn Statement is usually all about big, macho numbers. Billions and tens of billions are the order of the day. Millions are chump change. But hidden in the £100 billion National Infrastructure Plan is a small but very significant commitment – one that could make a vast contribution to the UK’s economic growth.

The announcement is to spend £10 million on a test-bed for self-driving cars. Here’s what this could mean and why this is a big deal.

What are self-driving cars?

Self-driving cars are one of the technology marvels of the age. Less than a decade ago, DARPA, the US government’s department of military wizardry, offered a prize for the best vehicle that could drive itself. The winner was Stanley, developed at Stanford University by a team led by German computer scientist Sebastian Thrun. Thrun was later signed up by Google, who developed the technology to such an extent that half a dozen driverless Toyota Priuses have been quietly and safely driving around Silicon Valley for over two years.

Why are they a big deal?

Driverless cars are great. First of all, they’re safer. They don’t get drunk, they don’t for the most part have accidents (the only accident to have befallen Google’s test fleet took place when a human was driving it). Safer cars may seem like a footling benefit, until you reflect that nearly 2,000 people die in car crashes in the UK each year and around 80,000 are seriously injured.

But the economic benefits go further than that. By sensing where other driverless cars are, they can drive faster without compromising safety. They can drop passengers off and park themselves. And of course they free people from the task of driving, which for all Jeremy Clarkson would have us believe, is not only tedious but actually bad for you.

This means they have the long-term potential to reshape a lot of things we take for granted, as this great Economist article points out. If cars will park themselves and then pick you up, we can dispense with a lot of costly inner-city parking. If cars can interact to ease stop-start traffic, there’ll be less congestion. It’ll transform the logistics industry (worth noting the short-term human impact of this will be significant). And if people can travel faster and have more time to do other things, the economy will benefit. (The New Yorker recently ran a great piece looking at these benefits – read the whole thing, as they say.)

The important thing to realise is that the productivity benefits of a technology like self-driving cars come not from manufacturing them but from deploying them. It’s a bit like computers. A few American companies got rich in the 1990s from manufacturing PCs and writing software for them, but the really big economic gain came from non-tech companies like big-box retailers adopting them and using them to transform how they did business. Chances are it’ll be the same with self-driving cars: manufacturing them is a nice economic fillip; using them is economically transformative.

What’s a test-bed and why is it a good thing?

This is where the test-bed comes in. The government’s investment is not about doing more R&D to develop self-driving car technologies. That work has to a great extent been done, and needs limited government intervention now that Google and major automotive manufacturers are on the case.

The real barriers to self-driving cars are social, not technological. How will our legal system handle a self-driving car accident (which, safety record notwithstanding, will happen eventually)? How will we decide the rules of driving self-driving cars? At the moment, Google cars always have a driver ready to take control – but need this always be the case? How do you tell the world a car is self-driving? (In Nevada, which has legalised them, they have an infinity sign on the door, which is cute.) And if the benefits of self-driving cars lie in how they allow to re-imagine commuting, work and urban planning, what does this actually mean in practice?

So we need some real-world experimentation. (I blogged about this a few months ago.) You can demo a robotic car in a lab, and test it in isolation on a track or with special permission on a street. But you can only understand the social implications of a technology when you release it into the real world. (This is the principle behind the Milton Keynes trial of the ULTra PRT, a personal transport system for which we’re proud to say Nesta provided early funding.)

A test-bed can do this. A test-bed is, at its core, a place where you can trial a new technology to work out how it interacts with the wider world. Here’s one way it could work:

• Find a local area that wanted to be the test-bed. (I’d imagine this might be attractive to a number of places with significant car industries – if many places were keen you could run a national challenge to pick one.)
• Gather a group of people who care about self-driving cars and making them work: one or more car companies, the local council or LEP, community groups, local employers and retailers. Leave it to them to identify some local opportunities that could be tested out – could a particular road or road-lane be adapted for self-driving cars? A cluster of offices or shops? A small town?
•  Provide some grant funding, on the condition of matched funding from partners. Leave it to the locality to decide how to spend it, but the sorts of things it could fund could be the purchase of self-driving cars for the trial, funds for re-designing traffic flows or parking systems, and public engagement. (Don’t forget that the EU is about to hand out billions of Euros in regional innovation funding – any area doing this should definitely consider submitting a big ERDF bid.)
• Run the programme for an agreed period of time – perhaps two years – gathering data and views on what is working and what is not. Unforeseen events will certainly happen: that is the whole point. The important thing is to learn from them and think about how the lessons can be applied more widely.

The end result should be tried-and-tested set of norms and ideas about how best to deploy self-driving cars, and practical experience of working with them, that can then be applied across the UK.

If you wanted to be grand, you could call this experimental government – it’s the state enabling experiments and learning from the results. Either way, amid the billions that’ll be doled out in tomorrow’s Autumn Statement, this looks to be £10 million well spent.