James Hou and Shirley Xu, seniors astatine The Bishop’s School successful La Jolla, are Regeneron Science Talent Search scholars, among 300 selected from 1,949 applicants crossed the United States and China with an oculus connected streamlining technological processes.
The Regeneron Science Talent Search, billed arsenic America’s “oldest and astir prestigious subject and mathematics contention for precocious schoolhouse seniors,” asks applicants to taxable an original, autarkic probe project, essays and recommendations.
The 300 scholars volition beryllium awarded $2,000 each, and their schools volition get $2,000 for each enrolled scholar.
Get the La Jolla Light play successful your inbox
News, features and sports astir La Jolla, each Thursday for free
You whitethorn occasionally person promotional contented from the La Jolla Light.
Forty scholars volition beryllium named Regeneron STS finalists connected Tuesday, Jan. 24. The finalists volition vie for much than $1.8 cardinal successful awards March 9-15 successful Washington, D.C.
Here’s a look astatine James and Shirley:
James, 17, made it to the database of 300 for his task “Near Real-Time Seismic Human Fatality Information Retrieval from Social Media with Few-Shot Large-Language Models.”
The task uses “the powerfulness of [artificial intelligence] and societal media successful this modern integer property to amended earthquake casualty predictions,” James said.
He collaborated with the U.S. Geological Survey “to measurement oregon foretell however galore radical are going to endure injuries oregon deaths” from earthquakes.
James said the task takes the “powerful accusation root of societal media, wherever radical are perpetually talking, and processes it with AI and modern technologies successful bid to springiness a much close and timely response. … This is truly utile for lifesaving assistance administered by authorities agencies.”
It’s often hard to get quality surveyors to earthquake sites to hunt for radical injured oregon dead, James said, truthful “social media provides a speedy model into these sites wherever radical tin beryllium reporting connected their ain phones astatine a overmuch faster complaint than going done authoritative channels that tin often beryllium burdened by a batch of this quality labour that is required.”
USGS has recovered that making usage of societal media successful that mode tin beryllium difficult, arsenic “you person radical tweeting everything and anything,” James said.
“That’s wherever AI comes in,” helium said. “I have, passim my past probe experiences and done robotics, acquired rather a bully knowing of AI and its abilities. I leveraged my cognition to span this gap.”
James was portion of a squad that took 2nd spot past twelvemonth astatine the First World Championship successful robotics.
“I’ve had a passionateness for AI since ninth grade,” helium said.
The thought for his Regeneron task came from his enactment implicit the summertime astatine Stony Brook University’s Simons Summer Research Program in New York.
He said his mentor there, Susu Xu, “was already reasoning of utilizing AI successful earthquake response. … I had a batch of acquisition moving with this benignant of exemplary that dealt with connection [and] societal biases contiguous successful connection AI, truthful that helped maine truly instrumentality connected this project.”
Being selected arsenic 1 of the apical 300 for the Regeneron contention is “a immense honor,” James said. “It’s a large designation for this task that I’m truly arrogant of.”
As helium waits to spot whether he’s 1 of the 40 finalists, helium continues to collaborate with his mentor and is “working to integrate our strategy into the existent USGS effect models.”
James credits Bishop’s machine subject teacher Marcus Jaiclin with helping to motorboat his probe and learning, and helium hopes to use AI successful civilian engineering aft helium leaves the school.
Shirley, 17, was selected for her task “A Heuristic Solution to the Closest String Problem Using Wave Function Collapse Techniques.”
The Closest String Problem, oregon CSP, is portion of the “NP-Hard/NP-Complete occupation family, regarded by mathematicians and machine scientists arsenic intolerable to lick perfectly wrong applicable clip limits,” she said.
But CSP, a theoretical machine subject computational occupation that tries to find the geometrical halfway of a acceptable of input strings, has “promising applications successful fields similar cause find and cistron coding,” Shirley said, truthful she was determined to question amended solutions.
“I developed a caller algorithm for the Closest String Problem and demonstrated that it outperforms aggregate state-of-the-art algorithms successful erstwhile publications that effort to lick CSP,” she said.
Shirley said the thought for her task came from video games, which are among her hobbies.
“My algorithm utilizes techniques inspired by Wave Function Collapse, oregon WFC, a method utilized successful video games to make analyzable visuals,” she said. “Drawing inspiration from Wave Function Collapse, I was capable to make a caller solution and mode of approaching the Closest String Problem.”
Shirley said she’s honored to beryllium among the 300 Regeneron scholars for 2023 and added that it “is highly fulfilling and encouraging.”
“[I’ve] ever enjoyed engaging with others from assorted disciplines and backgrounds, particularly successful research,” she said. “To person this accidental and to stock my enactment arsenic a Regeneron STS student is simply a imagination travel true.”
Since publishing her enactment successful a peer-reviewed journal, Shirley has received inquiries from respective researchers astir the satellite funny successful CSP.
“It’s ace breathtaking for maine to beryllium capable to sermon and collaborate with different researchers successful the field,” she said. “My aboriginal extremity is to proceed being a creator and innovator, enabling exertion to marque bigger impacts.” ◆