The future of roads being canvassed with autonomous cars draws near every day, and one thing’s for sure, they know how to stay in their lanes.
The researchers at Wayve, a start-up founded by scholars of the Cambridge University engineering department, have rolled their way into teaching a car working on Artificial Intelligence (AI). With a human driver in place, the car moves around and wherever dwindles off track, the driver corrects its movement. This steering in the right lane is a part of the company’s “reinforcement learning” algorithm, which is in the AI’s application. Infusing the car with synthetic consciousness through sensors, the reinforcement process demonstrates the learning factor being programmed in the car. As it corrects itself to stay on a singular path, the AI adapts and takes on a new skill, all in just 20 minutes. The company posted a video on YouTube of the learning process at work.
Amar Shah, Wayve’s Co-founder, in an interview explained the driving force behind the reinforcement learning process, “We want to give our vehicles better brains, not more hardware.” Reinforcement learning is being increasingly used as a tool to make AI more adaptive. Companies like DeepMind Technologies and Open AI have previously worked with AI, where it was taught how to play board games and beat human players in video games, respectively. Autonomous cars are a work in progress, and Wayve just broadened a car’s learning curve without taking a turn.