Midshipmen are creating artificial intelligence programs all set on solving problems from parking to cancer prediction.
The Naval Academy’s artificial intelligence conference Tuesday was an opportunity for 15 groups of midshipmen to show off their work to conference attendees, academy staff and community members.
The poster session was something of a military science fair. But instead of burbling dry ice volcanoes and construction paper-covered displays, midshipmen presented artificial intelligence programs they expect to get off of their posters and into the real world soon.
A chatbot for the academy’s MyNavyPortal website would fix its “practically unusable” FAQ section. The site’s search function doesn’t work, meaning students and faculty have to scroll through 87 pages of 12,000 questions to get the answers they need.
To do fix this, midshipmen Michael Garcia, Ryan Eilers, Lani Davis use a system called natural language processing.
To put it simply, “It’s magic math that turns words into numbers,” Eilers said.
The chatbot can answer common questions with more of an organic conversation. Instead of scrolling endlessly, users can message the bot, nicknamed “Chatty Cathy,” and get quick answers for their questions.
Once the program is up and running, Garcia said he wants to make the program a little more friendly with upgrades, such as the ability to make small talk.
“It can be used by the time we hit the fleet, which is a whole other level of cool,” Eilers said.
Midshipmen Timothy Forman and Kevin Lu hope to use their AI to crack down on sex trafficking.
One of the ways law enforcement officers find sex traffickers is through pulling phone numbers off of personal ads on the internet, they said. But traffickers have started to hide their numbers from the programs used to sniff them out with emojis, symbols, and other tricks. While these numbers are easy to detect by the human eye, there are too many ads to have a person look at each one to extract phone numbers.
That’s where Forman and Lu’s program comes in. The program learns how to read the coded language used in the ads to extract numbers fast so the police can do their jobs faster.
Another project could change the way doctors plan cancer treatments.
Midshipmen Jason Henry, Deon Odom, Toshi Oue and Tanner Strawbridge are working on a program to more accurately predict cancer survival rates. Using machine learning and a database of public cancer data, the group is working to create an algorithm that can be used as a tool for doctors in determining the best forms of treatment for a patient’s chance of survival.
The group is working with a doctor from Johns Hopkins medical center and hope to have the program ready by April.
Other projects tackle malware detection, use a Twitter-based machine learner to combat cyber attacks and use algorithms and robotics to defeat Army in a capture the flag drone competition.
Programs like these are meant to be practical but are also motivating for conference attendees.
Grace Knipe, a freshman at Severn School who came with a group of classmates, said she was excited by the project on sex trafficking.
“To see how someone’s working to prevent it with a tangible idea is inspiring,” Knipe said.
© 2019 The Capital
Distributed by Tribune Content Agency, LLC.