There was a fun piece by Allister Heath in yesterday's Telegraph arguing that driverless cars could generate big economic growth.

It builds on this terrific article in Economist, which shows how industries with no links to the car business, from pubs to hospitals, would be transformed if driverless cars hit the big time.

I for one can't wait to get a driverless car. I don't mind driving per se, but as a parent of young children I spend more time than I'd like driving in a state of sleep-deprivation that would probably be illegal if the police could only find a way to breathalyse it. Far safer a computer drives me, and leaves me to do more interesting things. I also like the idea of the economic growth that could come from this disruptive new technology.

So here's a question: what can the government do to help self-driving cars go mainstream more quickly? Allister's answer would be simple: legalise them, and then get out of the way - and cancel HS2 while you're at it. (I believe he is contractually obliged to say this in each of his columns.)

But in reality it's a bit more complicated than that. Self-driving cars are a nice example  of an innovation with big systemic implications. They require complex changes to liability law (who do you sue if a driverless car hits your fence?). They change the way we lay out cities (who needs parking places if your car will meekly go off and park itself in a multistory once it's dropped you off). They change how we manage our streets (do robots need roadsigns?).

Some of these issues will need to be tested out. (What is the ideal layout of parking for a city with self-driving cars?) Some require democratic consent, which people will be unwilling to give unless they have some idea of what the future will hold. (Will people be happy to allow cars to drive with no-one on board? Without this, a lot of the benefit of driverless cars is lost, but I'm not sure people would spring for it without more tangible proof that they're safe.) If we legalised driverless cars tomorrow and waited for the revolution to come, it's likely that not much would happen. A few enthusiasts would adopt, but the public would demand tight regulation and we wouldn't know what infrastructure would work.

At the same time, big government master-plans wouldn't work either. Asking a Department for Transport civil servant to design a self-driving car strategy and then implementing it immediately would also be dumb. No-one - not the DfT, not Toyota, not Google's Sebastian Thrun - knows enough about what makes self-driving cars tick to design this system. (Score one for Mr Hayek.) It wouldn't work, and as soon as it started to unravel, the public would demand it be scrapped.

Instead, what government should be doing is creating spaces to experiment. How about this for a plan: the Technology Strategy Board and the DfT should run a competition, in which they ask British towns to volunteer to be self-driving car test-beds.

The town that can show the greatest buy-in and involvement (from a council willing to alter road layouts to workplaces willing to experiment with new commuting patterns to citizens willing to participate) gets made a self-driving car zone. Self-driving cars are completely legal there, immediately. The TSB and DfT, working with Google, Toyota, whoever, spend a bit of money (of the sort they already spend on new technological development) to fund some cars, and - here's the clever bit - we see what happens. A relatively small investment, perhaps in the low single digits of millions of pounds, plus locally-applicable changes to the law, plus corporate co-funding and local political will could really test and demonstrate what self-driving cars could do, and see what sorts of other opportunities they enable.

Nesta Associate Mariana Mazzucato's thought-provoking The Entrepreneurial State made the point that businesses on the whole don't like taking really big leaps into the unknown - at most, they like to take sensible, balanced risks in fields they know once the broad concept has been proven. (Remember that Google only piled into self-driving cars after the mad scientists at DARPA used taxpayer's money to show it could be done.)

But at the same time, government isn't omniscient. It can't make grand strategies tailored to technologies that haven't been fully developed yet. What it can do is be experimental: creating the space for businesses and citizens to try things out, in the hope that someone hits on something wonderful that businesses can then elaborate and perfect and scale up and market. Perhaps what we need is not just an entrepreneurial state but an experimental one.