Can you learn AI by programming racing cars? Amazon’s DeepRacer League makes it possible2. October 2019
Can you learn AI by programming racing cars? Amazon’s DeepRacer League makes it possible
New York, October 2, 2019.
Employees of companies such as Morningstar Inc. and Liberty Mutual Group Inc. learn advanced artificial intelligence techniques by programming and driving self-propelled miniature cars. With this goal in mind, Amazon Web Services Cloud Business founded the DeepRacer League. On the basis of the reinforcing learning, the members develop algorithms based on the trial-and-error principle and observations. The technology differs from a type of AI commonly used in business, called supervised learning, in which algorithms must be fed with labeled training data to learn to recognize images or make predictions. In deep racers, the cars provide their own data: Images taken with cameras.
Anyone with an Amazon Web Services account can join the league. Teams or individuals can participate online in “virtual” races or in person at events around the world.
The teams develop and train AI algorithms with the Amazon SageMaker software, use them in self-propelled cars of about 10 inches and then drive them on a distance of about 17 feet by 26 feet. The fastest car wins. Thanks to the training, Amazon expects dozens of projects based on machine learning and other techniques to be in use by the end of 2020.
In addition to autonomous vehicles, reinforcement learning can be used to help robots run faster or develop safety systems that can automatically adapt to different environments. “It’s a pretty complicated technology and there’s a pretty steep learning curve,” said Mike Miller, general manager for AI devices at Amazon Web Services.
At Morningstar, more than 450 software developers, stock analysts and quantitative researchers have enabled employees to use the technology since January, and they have formed 100 racing teams in 10 countries.
Insurer Liberty Mutual now employs approximately 270 people, including software engineers and data scientists who participate in DeepRacer.
“It’s a fun way for people to gain practical insight into the very important algorithms in a secure environment where they won’t confuse core applications,” said James McGlennon, the company’s chief information officer.
The company already uses other machine learning techniques to optimize car insurance pricing based on risk factors and search for anomalies in operations.