Amazon has introduced a toy car designed to help developers safely test the self-driving code that could be used in full-size vehicles.
DeepRacer wants to help democratize the development of reinforcement learning. The technique was used by Alphabet researchers to create AlphaGo. An AI that defeated the human world champion of the board game Go, which is notoriously difficult.
Created by Amazon’s web services division, the car has an Intel dual-core processor, HD camera, monster truck tires and more. Amazon application developers can start learning in just a few lines of code.
The company will organize a contest called DeepRacer League. Where developers can compare their cars and models with others to win prizes in what Amazon calls the world’s first autonomous racing league in the world.
The Amazon league was created after the company’s own internal careers
“It started to be quite competitive,” said Andy Jassy, CEO of AWS, at the company’s reinvent conference in Las Vegas. “We had to remind people that we were actually trying to build this and launch it for customers. But in reality it was something interesting, educational for us.”
Reinforcement learning uses rewards, such as skill points to train AI. Which is why models are often compared to video games. The AI repeats the tasks to find the best result.
Also published on Medium.