Uber wants to turn its millions of drivers into a network of sensors for self-driving companies

Uber has long-term ambitions that go far beyond transporting passengers: the company eventually wants to equip human drivers’ cars with sensors to ingest real-world data for autonomous vehicle (AV) companies — and potentially other companies to train AI models on physical-world scenarios.

Praveen Neppalli Naga, Uber’s chief technology officer, revealed the plan in an interview with TechCrunch. The StrictlyVC event In San Francisco on Thursday night, he described it as a natural extension of the fledgling program the company announced in late January. AV laboratories.

“That’s the direction we want to go in the end,” Naga said of equipping vehicles with human drivers. “But first we need to understand the sensor sets and how they all work. There are some rules — we need to make sure that we’re clear in each state what sensors mean and what sharing means.”

For now, AV Labs relies on a fleet of small, custom sensor-equipped vehicles that drive themselves separately from Uber’s driver network. But the ambition is bigger. There is Uber millions of drivers on a global scale, and if even a fraction of these cars could be turned into rolling data collection platforms, the scale of what Uber could offer the AV industry would dwarf what any individual AV company could assemble on its own.

The idea that drives the program, Naga said, is that the limiting factor for AV development is no longer the underlying technology. “The big gap is information,” he said. “(Companies like Waymo) have to go around and collect data, collect different scenarios. You can say, in San Francisco, ‘At this school crossing, I want some data at this time of day so I can train my models.’

Becoming the data layer for the entire AV ecosystem is a pretty smart play, especially considering Uber abandoned its ambitions to build self-driving cars years ago (founder Travis Kalanick made this move public). big mistake). Indeed, many industry observers have wondered whether, without its self-driving cars, Uber might one day become irrelevant as AVs become more and more common around the world.

The company currently has partnerships with 25 AV companies, including London-based Wayve, and is building what Naga describes as an “AV cloud”: a library of labeled sensor data that partner companies can query and use to train their models. Partners that Uber plans to make more aggressive invest directlycan also use the system to run its trained models against real Uber rides in “shadow mode,” simulating how the AV would perform without actually driving.

“Our goal is not to monetize this data,” Naga said. “We want to democratize it”

Given the obvious commercial value of what Uber is building, that deployment may not last long. The company has already prepared stock investments its ability to offer custom training data across multiple AV players and at scale could give it significant leverage over a sector that currently depends on Uber’s ride-hailing market to reach customers.