Apple’s Drive.ai Acquisition Shows Self-Driving Cars in the Doldrums

Apple’s recent acquisition of autonomous vehicle startup Drive.ai throws a monkey wrench into the theory that it had backed away from an existing self-driving car initiative, known as Project Titan.

Early this year, Apple laid off 190 Project Titan staffers, including more than 100 engineers. But now Apple is bring on more engineers from Drive.ai, along with the startup’s fleet of self-driving orange vans and, presumably, its patents and other intellectual property.

On that basis, the buyout has been framed as a sign of Apple’s renewed interest in autonomous driving. But in fact, the deal highlights just how rough things currently are for autonomous vehicles, both as a technology and as a business proposition.

For a start, this was the most fiery of fire sales: Drive.ai had already filed notice with California that it would shut down, and had in fact, according to news site Axios, already ceased operations. Apple will reportedly pay less for Drive.ai than the $77 million in venture capital already sunk into the operation, and far less than the $200 million peak valuation that the startup reached in 2017.

That decline coincides with the growing realization that autonomous vehicles won’t be ready for widespread use anytime soon. For example, Tesla CEO Elon Musk had declared in 2015 that his company’s vehicles would be fully autonomous as of last year. But Tesla’s cars are still nowhere near fully autonomous.

Drive.ai’s planned shutdown came amid an apparent retrenchment of that company’s ambitions. The startup had previously partnered with Lyft to explore self-driving taxis, and worked on technology to retrofit large fleets of conventional cars for autonomy.

But Drive.ai’s most prominent recent initiative was automated mass transit – those orange mini-buses, which operate under tightly controlled circumstances that reduce the risk of accidents. In Frisco, Tex., those vans shuttled riders between two fixed locations (the service was scuttled in March). In Arlington, Tex., the company’s buses are currently used for making flexible loops through a roughly one-square-mile cluster of stadiums, expo centers, and hotels.

It seems increasingly clear that such limited routes, rather than freewheeling cross-country robo-journeys, epitomize the real, unsexy near-term potential for driverless vehicles. Other startups like Optimus Ride and Voyage have operated services either on fixed routes, or in sedate and predictable settings like retirement communities. Even Alphabet spinoff Waymo limited the ‘launch’ of its commercial robotaxi service to a few handpicked customers on the tidy suburban streets of Chandler, Az.

Apple, which has never publicly discussed its work on self-driving cars, did not respond to inquiries by Fortune about the Drive.ai acquisition or the fate of Drive.ai’s Arlington service.

Limiting the area in which self-driving vehicles operate significantly reduces the number of so-called “edge cases” that provide the biggest challenges for the computing systems behind autonomous driving. Edge cases, as described in a seminal 2017 essay by MIT roboticist Rodney Brooks, include stopped trucks or other obstructions, road changes that aren’t reflected on maps, or police presence.

These unexpected disruptions are like Kryptonite to autonomous driving systems because, like all artificial intelligence, they lack the complex ‘general intelligence’ that makes humans so adaptable (if unreliable) behind the wheel. When (or whether) artificial general intelligence will be realized remains hotly debated, but most informed projections put its arrival decades into the future.

Despite the grandiosity of the Project Titan moniker, Apple has been operating along similar, cautious lines for years. Recent reports have posited that Apple is building an electric van, the same basic design as Drive.ai’s vehicles. More substantively, Apple was reported last year to have partnered with Volkswagen to convert some of the automaker’s electric vans into self-driving employee shuttles—a modest proposal which the New York Times at the time said was already “behind schedule and consuming nearly all of the Apple car team’s attention.”

The Drive.ai acquisition, then, is hardly a rekindling of Apple’s autonomous ambitions. From its reported discount price tag to the glorified shuttle buses at its heart, the deal instead points to dashed hopes, hedged bets, and the continued shrinking scope of the erstwhile Project Titan. Apple may have added Drive.ai’s engineers to its roster primarily to get their specific experience with automated, fixed-route transit.

Apple, then, may no longer be trying to build the iPhone of cars. That would be a disappointment for investors because selling private cars to households has much more potential upside than selling vans to cities and retirement communities. But for society as a whole, low-speed, low-risk autonomous mass transit may in fact be a rosier mid-term scenario.

Though not conclusive, studies have projected that individual self-driving cars could make traffic congestion worse. Another found they’ll reduce carbon emissions and climate change—but only if the vehicles are shared by several riders.

With that in mind, autonomous shuttles for senior citizens may be more exciting than they seem.