At this year’s AWS re:Invent in Las Vegas, Amazon.com Inc. launched its AWS DeepRacer. While it is a toy car on the outside, it is also a highly sophisticated reinforcement learning platform designed to teach software developers build, train and optimize models in the cloud, leveraging Amazon SageMaker and AWS RoboMaker.

DeepRacer was demonstrated during AWS re:Invent in conjunction with a workshop teaching developers how to use DeepRacer to build a reinforcement learning virtual model in the cloud and then deploy it to the “real-world” car.

“We believe that AWS DeepRacer is … [a] tool for us to help get this kind of innovative technology into the hands of everyday developers and data scientists,” said Mike Miller (pictured), senior manager of product management for AWS AI at Amazon Web Services Inc. He explained that this type of machine learning can have a steep learning curve and can be cost-prohibitive. DeepRacer is a way to make machine learning training more accessible to more people.

Miller spoke with John Walls (@JohnWalls21), co-host of theCUBE, SiliconANGLE Media’s mobile livestreaming studio, and guest host Justin Warren (@jpwarren), chief analyst at PivotNine Pty Ltd., during AWS re:Invent in Las Vegas. In addition to reinforcement learning, they also discussed how developers can compete with their cars in the “real world.” (* Disclosure below.)

Teaching the software to teach itself

The essence of reinforcement learning is that, rather than working through a large number of presubscribed models, the software learns and instructs itself as it goes. Therefore, if something works, the artificial intelligence keeps it; if it doesn’t work, the AI discards and ignores it, making it more efficient over time. As part of assisting in this task, reinforcement learning is set up with a “reward” function, as well as a set of hyper-parameters around the training episodes that it can be used later to refine its functions.

Reinforcement learning’s strength is how it makes a series of short-term decisions to optimize long-term goals. It is particularly useful in challenging, highly varied environments, such as manufacturing supply chain optimization, healthcare treatments, and autonomous driving.

“What differentiates [reinforcement learning] from other machine learning techniques is that it doesn’t require large data sets in order to train a model to make a prediction,” Miller said.

To add extra fun, Amazon has created an AWS DeepRacer League, where developers will be able to compete in time trials during upcoming Amazon events. The fastest lap times will be eligible to race in a final during next year’s 2019 AWS re:Invent, with a grand champions cup going to the fastest car.

Watch the complete video interview below, and be sure to check out more of SiliconANGLE’s and theCUBE’s coverage of AWS reInvent. (* Disclosure: Amazon Web Services Inc. sponsored this segment of theCUBE. Neither AWS nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.

Photo: SiliconANGLE