Look out, the robots are coming!

One of the first signs to greet you when you arrive on the Isle of Wight by car is a large warning that Island roads are different. Apart from the obvious incompletion, the statement is misleading: the roads over here are fairly similar to many rural parts of the south and west of England, indeed many country areas throughout Europe. It is the drivers that are different.

The Isle of Wight has that lethal cocktail of young petrolheads who know every inch of the roads and like to drive their souped-up Subarus at the limit wherever they go, and the over-75s who max out at 35 mph and can hardly see the person sat next to them, although their head is turned towards them most of the time.

Add horses with riders who sedately turn out from a hidden bridleway, rickety pedestrians, cyclists who can remember pre-war editions of the Tour de France, and the surprise ploughing contest for vintage tractors, and it is constant mayhem.

I challenge any manufacturer of a ‘driverless’ car to demonstrate their product over here. I fear that none would last an hour before being involved in a pile-up in one of our narrow lanes which pass as country ‘A’ class roads.

Yet at times that the technology sections of our major media are short of news, their editors feel it necessary to tell us how close we are to the advent of driverless cars, and all manner of ‘intelligent’ robots and other systems, thanks to the latest advances in Artificial Intelligence (AI).

Unfortunately those with experience of the several previous AI ‘revolutions’ seem to have retired to remote sunny beaches, and those who really do understand AI’s limitations are being kept mum by those who are trying once again to cash in on it.

AI is capable of some amazing things, but it is also incredibly limited in certain respects. Fundamentally, AI-run systems can be based on pre-determined rules, such as basic physical laws, and those which it learns. Pose them problems which are outside their existing rule-set, particularly those at the edge of their capabilities, and they tend to perform very poorly – as do many humans. However when a human crashes a refuse truck into a crowd of people, the mob calls for the driver to be prosecuted. When the first driverless refuse truck hits a crowd, the whole idea of driverless vehicles will be hung out to dry.

Neither should you be under any illusion that any of this is radically new. About twenty-five years ago I had the privilege of visiting the bridge on one of the large ferries which cruises out of Stockholm, Sweden, towards the coast of Finland. Its captain proudly showed us its mainframe computer, which was used to navigate the ship through the intricate maze of rocky islands during its journey, and to dock the ship in Stockholm. We have managed to shrink the sensors and computers necessary for that sort of system and squeeze them into a car, at a price which many of us should be able to afford, but that is the most significant breakthrough in that period.

Similarly, aircraft autopilots are over a century old (dating back to 1912), and have been successfully landing commercial aircraft since 1965. Today’s autopilots are far more sophisticated, of course, and the addition of GPS location facilities has made a great difference. Cruise missile systems such as the Tomahawk have been capable of long-range precision guided flight since 1993.

Ironically, the biggest challenges to driverless cars and robot systems more generally are human, and animal. Designing a driverless car which does not collide with other driverless cars is relatively straighforward; catering for the erratic actions of other human drivers, cyclists, pedestrians, and animals such as horses, cattle, and sheep remains far tougher. We are thus likely to see driverless cars operating on US roads and highways long before they venture onto regular European roads, and may never see them at all on rural lanes and byways.

The other human factor which has to be considered is security. As with the Internet of Things, manufacturers aspire to great products which often turn out to be great hacks too. The security track record of cars is already pretty grim, although there are few models which offer the intruder much return for their effort. Once there is the prospect of taking complete control of a vehicle, we can expect to see much more worrying effort being put into such hacks.

So for the moment, at least, every time you see a bored tech journalist trying to push our panic buttons over driverless vehicles, robotics, or AI, remember that real life is different.