Too Risky to Leave Self-Driving Cars in AI ‘Hands,' Says IFA Keynoter
Self-driving cars once "seemed impossible," and now "seem inevitable,” but societal risks abound if autonomous driving isn't done right, said Jack Stilgoe, science and tech studies professor at the University College London, in an IFA 2020 keynote Thursday. He runs a program at the college called “Driverless Futures?” It studies society’s “excitement with self-driving cars and how policymakers can make better decisions about them,” he said.
Sign up for a free preview to unlock the rest of this article
Export Compliance Daily combines U.S. export control news, foreign border import regulation and policy developments into a single daily information service that reliably informs its trade professional readers about important current issues affecting their operations.
The “race” is on to see “which companies, which countries,'” autonomous-driving technology “work first,” said Stilgoe. The “dominant justification” for autonomous vehicles is “one of human error,” Stilgoe said. “Humans are terrible drivers. They get drunk, they get distracted, they get old, so the story goes. If we can take out the human, drop in an infallible computer, we can save hundreds of thousands of lives” and solve a myriad of other social problems, he said.
That’s “not the whole truth,” said Stilgoe. “If we swallow this story, we could end up making some bad decisions.” Much of what needs to happen with self-driving cars “is beyond the control of tech companies,” he said. “It requires more than just getting the artificial intelligence.” The science of “negotiating” a self-driving car on public roads populated with other vehicles, cyclists, pedestrians and joggers “is not chess,” he said. “Getting this sort of negotiation right requires really careful thought about rules of the road, who should be where, what should we expect of one another, and is it OK to just leave that in the hands of artificial intelligence.”
The world isn't “machine-readable in a way that we as humans with our eyes currently interpret road signs, environments and all of the complexities,” said Stilgoe. “It is absurd to assume that artificial intelligence will be able to do all of that. In places where self-driving cars have started to be tested, we’ve already seen extensive efforts at mapping, some efforts at installing smart infrastructures, and we can expect more of that.”
Expect there to be “more pressure to upgrade the world in order to meet the expectations” of self-driving technology, said Stilgoe. “We need to think about what a good transition will look like. If we don’t think carefully, we could fail to realize the opportunities of this new technology, and we could also risk exacerbating our dependence upon cars.”