The Amazon-owned mobility company Zoox has revealed its autonomous vehicle—a boxy-looking affair that bears few reference points to the human-operated cars we’re accustomed to.
The vehicle has a compact, squarish design, with attractive rounded corners and large glass doors on both sides. There’s no front or back per se–the vehicle can roll in either direction. Inside, you find two seats on each side, facing each other. The vehicle’s main resemblance to conventional automobiles is its four rubber tires.
And despite the big unveiling this week, the vehicle isn’t actually Zoox’s main product. “At the core we’re in the autonomous vehicle rides business,” CEO Aicha Evans says. In other words, Zoox plans to eventually offer a driverless taxi service in urban areas. “We’re reinventing personal transportation to be safer, cleaner, and more enjoyable.”
“We wanted a safe and compact design built for urban autonomous driving,” Zoox co-founder and CTO Jesse Levinson tells me. The Zoox vehicle has a powerful electric motor that can propel the car up to 75 miles an hour in either direction. Levinson says the vehicle, while optimized for short, local trips, also had to have enough power for special situations, such as when the best route to a local destination involves hopping on the freeway.
The vehicle finds its way around using cameras, lidar sensors, and radars, most of which are mounted at the four corners of the roof. The views of those primary sensors are overlapping, and together create a 360-degree field of vision, Levinson says.
The unveiling comes about six months after Amazon acquired the six year-old Fremont, California-based startup for $1.2 billion—a price tag roughly equal to the e-commerce giant’s largest acquisitions (Zappos in 2009 and Ring in 2018).
Some speculated that Amazon’s Zoox acquisition may point to the eventual development of a fleet of driverless delivery vehicles. Amazon has been keenly interested in finding a way into autonomous vehicles for a long time. It earlier invested in the electric truck maker Rivian and the self-driving car startup Aurora, which recently acquired Uber’s autonomous car division. Zoox was a chance to own, not just invest.
Autonomous vehicles are “driven” by large artificial intelligence models that control the car using data gathered from various sensors and cameras. Training these models has been a hard challenge for autonomous car companies because they must be ready to react to a virtually limitless number of driving situations on the road, some of them dangerous, and some of them involving pedestrians. The intensity of this challenge has led companies like Uber to move away from investing in self-driving technologies, which have developed slower than original projections. But Levinson says the industry has made great progress.
“One of the most exciting trends over the past decade has been the use of larger amounts of data to train AI models to predict things in the environment,” he says. According to Levinson, the models have become advanced enough to, for example, detect whether a pedestrian crossing the street is holding a phone and if they’re distracted. He says AI has now advanced to the point where it can comprehend and predict the street-level environment better than humans can.
“This has also accelerated with advances in computer hardware, which has enabled massive increases in compute power,” Evans adds.
While Evans and Levinson say they have an idea of how long all this will take and when the Zoox’s ride service can launch, they’re not sharing publicly yet.
Despite revealing the look of its cars, Zoox is still in testing mode. The company is already testing its vehicles in Las Vegas, San Francisco, and around the Bay Area. In the future Zoox will further evaluate the vehicles on test tracks and then private roads.