Artificial Intelligence is everywhere – from the voice assistants on your phone to facial recognition and self-driving vehicles helmed by some of the biggest companies in the world. These are all useful implementations for AI, but what about the fun parts, like… racing, for instance? Amazon Web Services’ DeepRacer does just that.
AWS DeepRacer is a way for developers to get hands-on with reinforcement learning (RL). Participating developers work with a fully autonomous 1/18th scale car – training, evaluating, and tuning to improve their car and their AI. Simply, developers are training AI on the science of decision-making – optimizing performance for the greatest reward.
Developers start off in a 3D virtual racing circuit, competing in time trial races and racing against other DeepRacer devs. These races can be hosted locally or from anywhere in the world. If they so choose, however, developers also have the chance to deploy their models onto AWS DeepRacer cars and join the DeepRacer League.
The AWS DeepRacer League is the first global autonomous racing league for developers in the world. AWS customers can compete globally with their AWS DeepRacer RL models, vying for prizes, glory, and the chance to win the AWS DR Championship Cup. The competition goes through five stages, each one unfolding over the course of a month.
At the end of it all, the best of the best get a chance to advance and compete in the final Championship Cup at re:Invent – AWS’s annual week-long conference. In a way, it’s almost like an esports competition – with devs against devs and AI vs AI. The main difference here is players moving from virtual races to physical races – watching machine learning in action, with real-world results.
Overall, AWS DeepRacer is a great way to experience machine learning firsthand – competing and collaborating, with a physical way to measure progress and improvement. There’s also the added bonus of making it more fun – gamifying AI development and bringing machine learning enthusiasts together.