Tuesday, March 20, 2018

Death by self-driving Uber car puts all autonomous vehicle research in jeopardy

Uber has called a halt to its driverless car testing after a woman pedestrian was killed by an Uber car operating in autonomous mode in Tempe, Arizona.
Uber had already suspended its self-driving car tests once before, after a crash - coincidentally, also in Tempe, Arizona - about a year ago, in which, miraculously, no-one was hurt. In that case, the Uber car was not able to react to another car unexpectedly failing to yield, despite the presence of a human in the driver's seat who could theoretically take over the controls when needed.
In this latest case, it is not yet clear exactly what the circumstance were, although police have said that the victim had "not been using a pedestrian crossing", i.e. she was probably jay-walking. Nor is it clear why the "driver" was not able to intervene. But it is nevertheless tragic, and also embarrassing for the whole driverless car program, which is being aggressively pursued in several states by several companies including Ford, GM, Tesla and Waymo, among others, as well as Uber. All of these companies are now looking hard at their own programs (Toyota has already suspended its own tests), and the USA is in the process of drawing up national safety guidelines for such cars.
In the meantime, organizations like Consumer Watchdog, which has been warning for years that the technology is being deployed before it is ready, and that such an accident was just waiting to happen, are saying "told you so", and calling for a moratorium of the testing of all self-driving vehicles on public roads. Autonomous cars may seem like a relatively straightforward technology and already well-tested, but it seems it is no match (yet) for the vagaries of human behaviour and the sheer number of possible situations that might arise in the real world.
The Center for Automotive Research is counselling perspective, and saying that this single fatality should be taken in the context of the 37,000 vehicle deaths (including 6,000 pedestrians) that occur in the United States every year. But their voice is a lone one in the wilderness, and this could be enough to drive a final nail into the coffin of autonomous car research.
A death by self-driving car seems somehow qualitatively different from a regular vehicle death (even if it shouldn't). Part of the problem is the hype that has built up around the technology, and the excessive and unrealistic promises that autonomous vehicles will one day eliminate all road fatalities. Even if it is actually a safer mode of transportation, its road to acceptance has just become that bit harder.

No comments: