Uber is taking a major step by turning its massive network of drivers into subjects of data collection. The company plans to install specialized sensors on drivers’ cars to gather the kind of real-world driving information that is required for perfecting self-driving technology. This technology has been well-researched on, but cannot be trained like or better than a human if the data is not first-hand. By accessing the millions of miles of actual trips, Uber hopes to feed its AI models more diverse data. It is not simply a private project, but rather something they plan to share with dozens of autonomous vehicle partners through a new cloud platform.
What is Self-Driving AI?
Self-driving AI is basically the brain of an autonomous vehicle. It is much like a technical substitute for a human actually driving a vehicle. It uses a combination of sensors like cameras, radar, and LiDAR to see the world in 360 degrees. The AI processes all that visual data in real-time to recognize objects, predict the movement of pedestrians on the road, and also to interpret traffic signs, along with recognizing the vehicles and objects on the road. It’s constantly making quick and smart decisions to navigate better. It has the advantage of machine learning, where it has access to a huge amount of data. The system is trained on millions of miles of driving data, so it learns from experience. With the data about literally everything about driving, it manages to escape any hazardous road situation in most cases.
How will Uber achieve this?
Uber is going to pull this off by turning their massive global network of drivers into a giant data-gathering machine. You heard it! Instead of relying on some specialized test cars, they plan to install sensors directly onto the vehicles already out on the road. This allows them to capture an incredible variety of driving scenarios, and the driving moves are pulled off by the drivers themselves.
On the backend, they are launching something called AV Labs and building out an AV cloud. This infrastructure acts as a hub where all that incoming sensor data is processed and stored for assessment and future use. This data is not merely private, as they are partnering with many autonomous vehicle companies to train their AI models faster. This could lead to something big for the self-driving promoting companies and be a big win for AI.
