Starsky Robotics, the self-driving trucking company responsible for a number of autonomous testing milestones, has shut down.
Stefan Seltz-Axmacher, the company's CEO and co-founder, said Thursday that fundraising conditions had become more arduous over the past year. A funding round expected to close in November instead fell apart, setting in motion Starsky's sudden demise.
But the reasons for a once-promising company's fall extend beyond a single circumstance. Collectively, they might send a chill through the rest of the self-driving industry.
In addition to the financial squeeze, Seltz-Axmacher said unrealistic promises surrounding the capability of artificial intelligence, investors who place profits ahead of safety and a trucking sector already enduring economic weakness all contributed to the company's end.
Those challenges, he says, aren't necessarily specific to Starsky. Across the industry, they're contributing to delayed deployment timelines that might soon bring forth a widespread reckoning.
In a blog post published Thursday, Seltz-Axmacher lays bare the substantial potholes on the road to autonomy. He lists the "professorial" pace at which many companies develop their products, the lack of tangible deployment milestones and the "open secret that there isn't a robotaxi business model" as key hurdles.
"The biggest, however, is that supervised machine learning doesn't live up to the hype," he wrote. "It isn't actual artificial intelligence akin to C-3PO. It's a sophisticated pattern-matching tool."
Self-driving systems developers have relied on machine learning to teach self-driving smarts. Early on, swift advances led to the belief there'd be exponential improvements in the competence of these systems.
But they require vast amounts of data and have not yet reached a level of sophistication that would enable the widespread deployment of self-driving technology. As they improve, they require even more data to identify rare traffic occurrences known as "edge-case'' scenarios, which in turn get more expensive and harder to find.
Solving remaining edge cases in a way that enables commercialization of the technology might cost billions more, Seltz-Axmacher said. At the same time, investor enthusiasm for self-driving tech has cooled.
"The VCs started to realize early last year that something was amiss in the autonomous industry," he tells Automotive News. "They're putting a lot of money in, and they're not getting anything out."