By Dave DeFusco
Manish Kumar Thota, a student in the Katz School鈥檚 M.S. in Artificial Intelligence, is the recipient of 麻豆区鈥檚 $25,000 Sacks Impact Graduate Fellowship in Ethics and Entrepreneurship for his proposed work on developing a machine-learning chatbot that can assist with students' academic work.
The chatbot, utilizing a newly developed Generative Pre-trained Transformer (GPT) model, aims to provide better personalized and interactive learning experiences for students with machine learning algorithms, natural language processing and a vast assortment of educational resources. It can offer enhanced conversational abilities, personalized learning experiences, comprehensive subject coverage, real-time assistance and continual learning and improvement.
鈥淭he Sacks Impact Graduate Fellowship, known for its commitment to fostering leadership and creating positive change, aligns perfectly with my passion for making a difference in the world,鈥 said Thota, who is interning as a machine learning engineer at S&P Global in New York. 鈥淭his fellowship will provide me with a unique platform to enhance my professional skills while addressing critical issues that require immediate attention.鈥
The most common way people experience artificial intelligence is through chatbots, which work like an advanced form of instant messenger, answering questions and formulating tasks from prompts. These bots are trained on troves of internet data and are adept at finding patterns and imitating speech. They don鈥檛 interpret meanings, however, rather acting like a high-fidelity version of autocomplete.
Since its debut, ChatGPT has stunned users with its ability to produce complete novels, computer code, TV episodes and songs. Thota explained that while online learning is an effective way to acquire knowledge, it can be challenging to learn from recorded videos without the ability to interact with the instructor or other students. To address this challenge, he proposes a course video that uses GPT to extract relevant information from video content and to generate automated question-and-answer interactions.
The process would involve three steps: Video analysis using a GPT model to first analyze video content to identify key concepts, topics and discussions; the generation of a set of questions based on the identified information; and the creation of answers to the generated questions.
鈥淭he goal of this project is to train a large language model that can recognize and teach course content, answer questions and provide interactions with audiences,鈥 said Dr. Youshan Zhang, an assistant professor in the M.S. in Artificial Intelligence program and Thota's advisor. 鈥淭his model could be used to improve the learning experience for students in online courses by providing them with a more interactive and engaging learning environment.鈥
The model, he added, allows students to learn at their own pace, review the material as many times as they need, interact with the material in a more engaging way and get immediate feedback on their understanding of the material. Thota said he chose to study at the Katz School because of Dr. Zhang and his work in this area.
鈥淚'm an AI and data science enthusiast specializing in machine learning,鈥 said Thota. 鈥淚n a rapidly evolving technological landscape, I believe it's crucial to stay updated with cutting-edge techniques and trends, and the Katz School is enabling me to do that.鈥