Imagine cruising down the highway in a self-driving car when, without warning, it swerves hard left, crossing lanes and careening off the road. This isn't a scene from a sci-fi thriller—it's a real phenomenon that parallels the 'hallucinations' seen in AI chatbots like ChatGPT. In a recent episode of 'Interesting Times,' author Andrew Miller joined columnist Ross Douthat to explore the unsettling errors autonomous vehicles can make and the security risks they pose.
Miller confirmed that such erratic behavior occurs, citing videos of Teslas suddenly veering into opposing traffic for no apparent reason. 'You struggle in vain to know what possibly encouraged it to do that,' he said, likening these incidents to AI hallucinations. While systems improve over time, perfection remains elusive.
Turning to security, Douthat raised fears of terrorists hacking into Waymo's network to control hundreds of cars simultaneously. Miller acknowledged the theoretical possibility but downplayed the ease: 'You'd have to hack into every car individually, and Waymo is a hard target with robust cybersecurity.' He argued that America's power grids and utilities present far softer targets for causing mass disruption.
Yet, the psychological impact of a hacked car is uniquely terrifying. 'It's tied to personal privacy and personal control,' Miller noted. The idea of losing command of your own vehicle, even if rare, taps into deep-seated fears, as reflected in popular culture—like the upcoming 'Naked Gun' movie featuring a murder by self-driving car. These anxieties, while disproportionate to the actual risk, are a natural response to unfamiliar technology that challenges our sense of autonomy.