No, Elon, a Tesla won’t fix Justin Timberlake problems
By Liam Denning
The big challenge for robotaxis is the sheer number of unforeseen situations they are expected to deftly handle. How would they react, for example, to an allegedly intoxicated ex-boyband singer veering across the genteel byways of the Hamptons? Impossible to know? Not to Elon Musk. Replying to a tweet from a fan, he concurred that had Grammy-winning entertainer Justin Timberlake been driving a Tesla, it would never have allowed him to run a stop-sign near Sag Harbor, which led to his arrest last week.
That Timberlake should figure, however obliquely, in Tesla Inc.’s robotaxi hype is itself an unforeseen situation. But these edge cases have a way of cropping up — and this one comes at an interesting time.
The National Highway Traffic Safety Administration is due in November to issue a regulation stipulating that, beginning later this decade, new vehicles come standard with technology to passively detect and prevent intoxicated or otherwise impaired driving. The NHTSA was tasked with this via a provision buried in the Infrastructure Investment and Jobs Act of 2021, and has issued an advance notice of the proposed rule to solicit feedback. The agency can delay issuing the regulation if it sees fit, but would need to explain its decision to Congress. Almost a third of vehicle-related fatalities in 2022 involved at least one intoxicated driver, equating to one death every 39 minutes.
If technology can mitigate those depressing numbers, great. But actually doing so, and in a passive manner, is a complex undertaking. (So-called ignition interlock devices, whereby drivers use a breathalyzer connected to the car to allow it to start, aren’t passive and are generally reserved for drunk-driving offenders). Any such solution rests on Driver Monitoring Systems, or DMS, which consist primarily of sophisticated cameras tracking the driver’s eye movements and positioning to detect drowsiness or distraction. These are becoming more common in the US as a means to ensure drivers aren’t abusing partial automation features.
Accurately detecting intoxication, which can manifest in different ways, likely requires pairing such cameras with other technologies, according to Colin Barnden, a DMS expert and principal analyst at Semicast Research. These include products such as NeuroMonitor, a software package from Israeli startup CorrActions that uses existing sensors in equipment such as the steering wheel to analyze ”micro muscle movements” in order to determine drivers’ level of cognition.
Although artificial intelligence raises fears of robots one day killing us off, Americans would perhaps be more panic-stricken at the notion of robots taking their car keys. Even as the sophistication of DMS technologies improve, how they ultimately get deployed against drunk-driving remains as much a social and political matter. It is easy to imagine future edge cases. For example, a vehicle detects someone being over the alcohol limit and shuts off the ignition — which may then raise other risks if they are alone and far from home late at night.
There are multiple ways to handle such situations short of shutting down a car, including limiting speed and/or heightening warnings about lane position or distance to other vehicles. And we should consider the benefits of DMS more laterally. More than 3,300 road fatalities in 2022 were linked to distracted driving, lives cut short for the sake of a text message, essentially. Warning smartphone-addled drivers to keep their eyes on the road and their thumbs on the wheel is a no-brainer. Moreover, as advanced driver assistance systems proliferate, DMS offer a way to fine tune what those systems do in a given situation by factoring in the driver’s level of alertness.
Whether detecting a drunk driver or a sober distracted one, the point of these systems is to, as Barnden puts it, make human drivers safer rather than replace them altogether. Yet the public debate about automated driving focuses overwhelmingly on the latter. That is in no small part because of Musk’s repeated assertions that robotaxis are just around the corner and the branding of Tesla’s driver-assistance packages as Autopilot and Full Self Driving. Tesla’s own DMS capabilities scored poorly in the Insurance Institute for Highway Safety’s recent ranking of partial automation safeguards. Tesla was forced to recall more than two million vehicles in December after the NHTSA determined Autopilot wasn’t doing enough to ensure drivers remain engaged.
As the IIHS has said, “vehicle technology alone is not going to be the single solution to prevent all fatalities on our roads.” Philip Koopman, an associate professor and automated driving specialist at Carnegie Mellon University, concurs. He notes that while the US and the UK had similar fatality rates 40 years ago, the UK’s is now less than half that of the US, indicating other factors are at play. For my part, living in the New York suburbs, the paucity of public transportation options is one potential factor that has nothing to do with robots.
The notion that a Tesla wouldn’t have “let” Timberlake run a stop-sign fits within this context. The robotaxi vision, which now underpins the vast majority of Tesla’s massive market capitalization, centers on negating the driver’s role. Yet it remains just that, a vision, and the hyping of vehicle automation has meanwhile created confusion about such systems. Despite the name Full Self Driving, Tesla advertises that system as being able to identify stop signs and slow the vehicle to a halt, albeit “with your active supervision.” Even the capabilities of automatic emergency braking, a useful and relatively straightforward feature to help prevent collisions on models produced by several manufacturers, are overestimated by a majority of US drivers, according to a recent survey conducted by the American Automobile Association, or AAA.
If sober drivers are prone to such overconfidence already, imagine adding a few drinks to the mix.