The Daily Telegraph chose to run a feature on self-driving tech, for which I provided some quotes.
This is the relevant section from the story, which is behind a paywall (although you can see it in the tweet thread).
For years, we have been told that such AVs are just around the corner, ready to take us wherever we want. They never are. In some confined, well-regimented places like ports, or even the suburbs of Phoenix, Arizona, driverless vehicles already shuttle about. “The problem is,” says Jack Stilgoe, professor of science policy at UCL, “most of the world does not look like Phoenix.” It’s much more complicated. And the learning-by-doing methods that often inform AI are simply impossible on the roads.
It was trial and error, for example, that helped a Google computer learn, and then master, the complexities of first Chess and then the boardgame Go. Other researchers have even let a tiny aircraft crash almost 12,000 times as its computer learnt to fly it. But the skies are comparatively empty. And letting computers learn by getting it wrong as children jump out from behind parked cars is unthinkable.
The result is that forecasts have got more pessimistic. Chris Urmson, former head of Google’s AV team, thinks they will only be phased in over the next 50 years.
On inspection, there is good reason for that long wait. For it is beginning to dawn that fully autonomous vehicles will never be able to manage on our messy streets. “The story we are told about self-driving cars is that the technology will be able to adapt to the world and that the world needs to do nothing,” says Stilgoe. “I think we know that’s a lie. The world always adapts to meet the technology.”
In other words, AVs will never drive in the current environment. Rather, we will change our environment to enable AVs to drive. To the human eye, a traffic light, even with the sun behind it, is a straightforward signal. To a computer, judging its shade and illumination in such circumstances can be extremely hard. Far easier would be to have a “smart traffic light” that communicates stop and go signals digitally and wirelessly to the car’s computer. Piece by piece, over the next decades, we will “upgrade” and adapt our world in this way to suit AI. “The world has been designed to be human-readable,” says Stilgoe. “AI needs the world to be AI-readable.”
That could well involve other changes, aimed at simplifying our complex roads for AI’s benefit. Reserved lanes for AVs, perhaps; more strict enforcement of speed limits; greater surveillance, where ubiquitous cameras ping real- time data back and forth. “These are relay consequential discussions we need to have,” says Stilgoe. “It’s not just about plucking out the human driver and switching in a computer and nothing else changes.”
Of course, one of those critical discussions will be about “how safe is safe enough” when it comes to computer drivers. The Trolley Problem will not go away. But that is only part of the greater truth we must grapple with – that ushering in the age of AVs will not only mean a new code of ethics, it will mean re-engineering the world.
“The sooner we acknowledge that,” says Stilgoe, “the better.”