The class was busy learning how to implement the best way to proof their model, making sure their AI could work to perfection. (Photo courtesy of Vivaan Vora)

The highly anticipated presidential debate on Sept. 10 left a hopeful or doubtful impression of the presidential nominees on Americans. It was quickly followed by mass media outlets fact-checking the claims. However, AV’s Advanced Computer Science class used this opportunity to create an AI-powered fact checker to fact-check the statements made in the debate. 

“There is [an] AI similar to what powers ChatGPT, and we will take the audio from the debate, convert that to text, feed that text into the AI, and have it return true or false,” said Noah Small (‘25). 

Types of AI

Different groups used different algorithms to check the statements using AI. Groups used open-source tools like Google’s Gemini or OpenAI’s ChatGPT.  Though most groups struggled, some groups got the fact-checker to be accurate. 

“The biggest challenge we faced [was] converting the YouTube Subtitles, putting them into ChatGPT, and then creating a real-time response. That’s the main goal of the AI fact checker,” said Mithunn Balaji Sivakumar (‘25). 

The idea originally came from the teacher, Kevin Kiyoi, who wanted to use the debate to progress the student’s knowledge. 

“He gave us the idea for this because one day, he just randomly started talking about it. And everyone agreed that it was a good idea to pursue,” said Reyansh Samanta (‘25). 

Whether the project succeeded, students taught themselves how to integrate Application Programming Interfaces (APIs) and work as a team. 

Learning Process

“This project was mainly for us to learn about how we can create a project as a team. How we can have teamwork skills among our class and how we can work with our classmates,” said Vivaan Vora (‘25). 

Students learned heavily about how AI works in this project, especially for students who took AP Computer Science A last year. Some students also used this opportunity to upgrade their systems for expanded use in the future.

“It’s not very applicable but it’s something that could definitely be used later if it gets to a more workable state. For example news networks over debates,” said Small.