Data sciences student showcases machine learning to U.S. legislators

By Peter Murphy

“At first we weren’t sure if the invitation e-mail from AWS was spam or legitimate since it wasn’t something he had applied for,” says Caitlin Hoekstra, director of career development & experiential learning in UB’s School of Engineering and Applied Sciences. “I confirmed with AWS that Bipin was selected for this unique opportunity and was thrilled to have a student from UB represented at the event.”

In June, Biddappa PK participated in the AWS student league deep racer. “It’s like an F1 race, but basically for machine learning models,” Biddappa PK says. “By the end of the month, I found out I finished the race in the top 10% of people in the country.”

Every month, AWS conducts a DeepRacer on different courses. Racers develop their machine learning models to finish the courses and deploy them on a small robocar. In the student league, there is a different track each month, and it is a timed circuit. Students create models and race against each other to determine who can complete the track in the fastest possible time. According to Biddappa PK, these models use the concept of reinforcement learning.

“We deploy a model to the car, and it learns its path around the track. It’s going to go wrong a couple of times initially during the training process, but it will get better. It’s the same concept that’s used in any autonomous cars.”

Biddappa PK and other racers demonstrated machine learning in real time to state legislators from across the U.S. at the NCSL in Indianapolis. One tool available to address some of the objectives of different state legislatures, net-zero carbon emissions, improved healthcare systems, etc., is artificial intelligence. AI models are based on data, and when that data is biased, the AI models are also biased. Biddappa PK talked with state legislators about this idea and some of the other ways machine learning can help realize their goals.

“I spoke to many legislators about how machine learning models are trained and how bias in the dataset can impact the results,” Biddappa PK says. “There could be such bias in our data, and our models are trained upon this data. When you’re using AI and machine learning in military or healthcare industries, it’s very important to know about these biases in your data.”

The legislators were also interested in how machine learning could affect traffic. Biddappa PK discussed autonomous vehicles and their ability to enhance traffic flow and safety, and also discussed how machine learning could optimize traffic flow in the future.  

An audience with policy-makers to discuss machine learning was not what Biddappa PK expected when he first decided to become involved in AWS DeepRacer. He discovered the leagues on YouTube, and initially had other goals in mind.

“When I first started out on this race, all I was after was a $50 Amazon gift card that would be given to anyone finishing in the top 10. Now, I want to get into the top five because the top five people from each region are called for a bigger race to decide number one, and I want to get into those,” Biddappa PK says. “Now, I really want policymakers to see how AI and machine learning can contribute to their policies and help their people.”