Has a job ever forced you to train your replacement? That’s the situation some Uber drivers may find themselves in. TechCrunch reports that Uber wants to equip its drivers’ cars with sensor suites that would collect data to train AI models that would presumably replace human drivers with better-trained robotaxis.
Robotaxis are all the rage these days, with companies like Waymo vastly expanding their coverage from one city to the next and Tesla giving its best shot. Uber bailed out of the robotaxi game in 2020 after a high-profile crash where one of its test cars struck and killed a cyclist in Tempe, Arizona. Robotaxis still have their problems, but have improved significantly since the 2018 Uber crash. Uber co-founder and former CEO Travis Kalanick has publicly called giving up on robotaxis a mistake, and current management seems to have realized it missed the boat on replacing its human drivers with machines.
Rather than compete with Waymo, Tesla, and others as it has in the past, Uber wants to gather data through its human drivers and sensor suites attached to their cars as they make their normal rounds. This data can then be used to train AI models to become better autonomous drivers. Uber’s chief technology officer, Praveen Neppalli Naga, explained this move to TechCrunch:
Naga said that the limiting factor for AV development is no longer the underlying technology. “The bottleneck is data,” he said. “[Companies like Waymo] need to go around and collect the data, collect different scenarios. You may be able to say: In San Francisco, ‘At this school intersection, I want some data at this time of day so I can train my models.’ The problem for all these companies is access to that data, because they don’t have the capital to deploy the cars and go collect all this information.”
From rideshare to data broker
This may be a rather shrewd move on Uber’s part. Before any robotaxi service can begin autonomous operation in a new city, it needs to collect detailed data on the area. Humans drive the robotaxis manually, without passengers, to gather this data through the same sensors that will later help them navigate the urban maze on their own. Uber’s idea to turn existing rideshare cars into data collection devices may let it give companies a head start on this process. When Waymo wants to move into a new city, it could buy Uber’s data to accelerate and enhance the process of learning it.
Another area where Uber wants to help autonomous vehicles is expecting the unexpected. “Uber told CBS News that one of its goals is to track unpredictable events, like trash cans blowing into a roadway or a pedestrian suddenly appearing in the dark, that synthetic models are worse at predicting.” It’s interesting that Uber used this example. A person suddenly appearing in the dark is exactly what led to the fatal Uber crash back in 2018. Yet this is still one area that autonomous vehicles are in desperate need of improving. A three-way standoff on a dead-end street is comical, but blowing past school buses and blocking emergency vehicles could be downright deadly if things don’t improve.
Unfortunately, the people being asked to provide this data are the same people who may find themselves out of a job once robotaxis are common and reliable enough to replace them. Rideshare drivers have protested in Los Angeles and run Waymo out of Boston. Picking up a few hours here and there driving for Uber is one of the few ways people can reliably make a few extra bucks at a time when jobs are scarce, wages are stagnant, and everything’s too expensive. Replacing human drivers with clankers deprives us of even that small bit of relief.
