Lessons from the “Confused” Waymo Taxis of San Francisco
For weeks, Randol White dealt with a persistent and annoying problem right outside his San Francisco condo. His building overlooks an open-air parking lot that Waymo recently leased to store its vehicles. He and other residents were repeatedly roused at 4 a.m. by honking car horns. Looking out his window, he realized the Waymo cars were honking.
Waymo, a company owned by Google, manages a fleet of self-driving taxis. In the evening, dozens of the white taxis with their spinning headgear enter the lot, park, and power down until activated. But why were they honking?
It turned out that the root cause was a safety feature built into the Waymo taxis’ software colliding with the unanticipated consequences of emergent behavior. When too many taxis awake at the same time and head toward an exit, the parking lot lanes become clogged, just as might happen with human drivers. If one taxi gets too close to another, its programming instructs it to give a short honk and back up. If a third taxi is behind that one, more honking and backing up occurs. With each taxi honking and backing up before attempting to leave the lot, they could continue for ten minutes or more before one of them found an opening. If too many taxis joined the scrum, they might gridlock and keep honking as they attempted to carry out their programming. That’s a lot of honking at 4 a.m.!
Unintended consequences often follow when a reasonable, yet not-fully-tested idea encounters real-world chaos and, in this case, when robots meet reality.
Sophia Tung, an engineer with a view of the lot, caught the unusual show on video. The constant beeps and movements of the Waymos even merged into her dreams. Their nocturnal “After You” dance captivated Tung. “I found myself just staring at it for 10 minutes at a time, watching these machines figure each other out,” she said. “It was like watching a fish tank.”
Intrigued, Tung turned her observation into a project. She set up a webcam to livestream the view, adding relaxing music. The stream, cleverly named “LoFi Waymo Hip Hop Radio Self Driving Taxi Depot Shenanigans to Relax/Study To,” was initially for her own amusement while working. However, it quickly gained popularity online.
Randol White sees a silver lining. “San Francisco and the Bay Area are a hotbed of innovation. We try new things, and that’s why we lead the world and the country when it comes to tech.” Philosophically he adds, “This is just a hiccup, an annoyance.”
While White’s perspective is optimistic, this incident highlights a broader challenge in technological innovation: the struggle to seamlessly integrate new technologies into complex human environments.
The Waymo taxis, for all their technological sophistication, struggled to navigate the nuances and complexities of urban environments shaped by human behavior. Their developers had clearly mastered the technical capabilities required for autonomous driving. But they had yet to fully account for the messy realities of integrating this new technology into the real world.
This challenge isn’t unique to Waymo. We’ve seen similar issues with social media algorithms, self-checkout systems, and other novel technologies. Solutions that work flawlessly in the lab or in simplified testing can falter when exposed to the unpredictability of human society.
Designing new technologies to blend into human environments requires a deep understanding of how people actually think and behave. It demands an iterative, user-centric approach, with constant testing, feedback, and flexibility to adapt to unintended outcomes.
The Waymo incident underscores the essential role of design thinking and creative problem solving in technology development. If the developers had employed a more robust design thinking approach, they might have anticipated the “honking stalemate” by deeply understanding the nuances of the urban environment where their technology would operate.
By engaging in thorough field research, observing and mapping the daily rhythms of the parking lot, and testing in more complex, real-world environments, they could have identified and addressed potential issues before large-scale deployment.
Waymo acknowledged the problem and moved swiftly to resolve the software issues causing the erratic taxi behavior. But the broader lesson is one of humility and adaptation. As we continue to develop transformative new technologies, we must learn to make them not just functional, but truly suited to the human world.
Only then can we unlock the full potential of innovation to improve our lives, without the unintended chaos that can ensue when robots meet reality. As for Randol White and his neighbors, they can now sleep peacefully, their dreams no longer interrupted by the honking of confused robotic taxis – at least until the next technological hiccup comes along.