Amazon has unveiled a toy car designed to help developers safely test self-driving code which could be used in full-sized vehicles.
The 1/18th-scale vehicle is called DeepRacer and costs $249 if pre-ordered, or $399 post-release.
DeepRacer wants to help democratise reinforcement learning development. The technique was used by Alphabet’s researchers to create AlphaGo, an AI which defeated the human world champion of the notoriously difficult Go board game.
Created by the Amazon Web Services division, the car features a dual-core Intel processor, HD camera, ‘monster truck’ tires, and more. Amazon claims developers can start learning in just a few lines of code.
The company will be hosting a competition dubbed the ‘DeepRacer League’ where developers can pit their cars and models against others for prizes in what Amazon calls the world’s first global autonomous racing league.
Amazon’s league was created after the company’s own internal races.
“It started getting pretty competitive,” Andy Jassy, Chief Executive of AWS, said at the company’s re:Invent conference in Las Vegas. “We had to remind people that we were actually trying to build this and launch this for customers. But it was actually kind of interesting, educational for us.”
Reinforcement learning uses ‘rewards’ such as score increases to train AI, which is why models are often tested against video games. The AI repeats tasks in order to find the best outcome.
Yesterday, our sister publication AI News reported on Uber’s achievement in developing a new type of reinforcement learning which enabled its AI to succeed where others have failed in tackling games with little in the form of reward.
Interested in hearing industry leaders discuss subjects like this and their use cases? Attend the co-located AI & Big Data Expo events with upcoming shows in Silicon Valley, London, and Amsterdam to learn more. Co-located with the IoT Tech Expo, Blockchain Expo, and Cyber Security & Cloud Expo.