Video of my conversation with Lichelle Wolmarans about Cyber Republic, organized by the LSE Alumni on October 27, 2020.
The new driverless car from Google has no break pedal or steering wheel. And that’s because the consensus is that driverless cars should completely replace the human driver. Even blind people would be able to drive them. They will be like elevators: push a button and wait till the thing gets you there safely.
It is sensible to have driverless cars for highway, long-distance driving. Imagine a driverless lane, where you enter and automatically release control of your vehicle to the collective intelligence of the highway traffic system. Such a facility will decrease accidents dramatically, and make long-distance driving more like riding a train. But when it comes to driving in a city or in small roads in the countryside, completely driverless cars make much less sense that it seems.
First there is a number of ethical and legal dilemmas to resolve: for instance who should be held liable when a driverless car kills a human. And there are huge technical challenges to resolve too: pattern recognition is not so advance to deal with unexpected obstacles or very low visibility. To solve these problems will take years. Indeed the computing power to solve these problems requires supercomputing level of power, or a totally reliable broadband connection to cloud supercomputing. If you start doing the maths of millions of driverless cars computing in parallel and solving highly complex pattern recognition problems you get the idea.
But perhaps, the biggest problem all is the safety culture of the automobile industry. Given the wide margins of error when driving a car, safety is often compromised – and rightly so – within accepted limits so that cars are economical. Completely driverless cars change the nature of driving. The automobile industry will have to mimic the aeronautical industry, where new designs are tested to exhaustion and multiple systems run in parallel to safeguard against catastrophic inflight failures. But if the same standards were to be applied to cars, new models would need many years of testing before being released to consumers, and their price tags will be stratospherically (sic) higher!
Perhaps, better to think of future cars like horses. Animal intelligence offers a much better area for the application of machine intelligence for cars. A horse analogy car will have highly advanced sensing, and will compensate for driver ability. It will be able to take over if the driver behaves erratically and bring the car to a safe stop. It will collaborate synergistically with the human driver in a horse-rider analogy of a second order cybernetic system. Under normal driving circumstances it not take over the human driver completely but become an extension of the driver’s sensing and acting ability. The “animal car” will thus be the logical evolution of the current “plant-like” cars.
…And Toyota seems to agree. Gill Pratt, who stepped down recently from the Defense Advanced Research Projects Agency (Darpa), will move to Silicon Valley to head Toyota’s robotics efforts, the company said, according to a recent article in the Financial Times. According to the article, Toyota considered human drivers an integral part of driving, with a totally driverless mode becoming an option.