Research Interest Statement
Attention: This page is only for INTL1830 course work!
Beyond the Screen: Can Machines Ever Truly Understand Us?
Imagine having a conversation with a librarian who has read every book ever written, in every language, and can summarize them for you in seconds. It is not a dream or fantasy nowadays. You must be using ChatGPT, Gemini etc. These are what we call Large Language Models, or LLMs—the invisible engines behind the above tools. Simply put, they are computer programs trained on massive amounts of human text to predict and generate speech that sounds remarkably human. In the daily life of some people, they are utilizing these tools to summarize their email, write essays and deal with complicated computer programming tasks. Moreover, image/video generative model(Sora) have been deeply integrated into artistic creators’ work.
Like human beings, LLMs output sentence word by word. They will predict the next word and then select which has the largest probability based on previous generated words. If you have mastered to use LLMs, you will notice that you can upload files by yourself to let LLMs analyze them. LLMs could base on external unseen information to generate the answer you may need. Evidently, LLMs like a professor or an encyclopedia around you. Have you ever wondered why ChatGPT could give you an amazing answer within a few seconds? Have you tried to figure out why the answers sometimes have wrong information? You must be shocked if you know the truth.
Before joining Auburn University, I was researching LLMs at an AI institute. LLMs was still in its infancy at that time. I became fascinated by this field when I realized that these models don’t just ‘copy’—they seem to ‘reason.’ I wanted to understand the boundary between a machine’s logic and a human’s intuition. Now, I dive deeper into the underlying principle of basic techniques for LLMs. My research could lead to a future where your local clinic has a ‘digital specialist’ capable of spotting rare diseases that might otherwise go undiagnosed for years. Imagine a world where your doctor spends less time typing on a computer and more time looking you in the eye, because an AI is handling the complex data in the background. For instance, an LLM could monitor an elderly patient’s symptoms at home and alert a doctor the moment a pattern suggests a potential heart issue, preventing an emergency before it happens. This topic is critically important because our healthcare systems are currently stretched to their limits, with medical errors often occurring simply because information is buried too deep for a human to find. It involves multidisciplinary knowledge(mathematics, computer science and linguistics) and excellent strategic foresight for a researcher. So, nothing is an easy task for me and my peers. You should support this type of research because health is our most valuable asset. By investing in smarter, safer medical AI, we are not replacing doctors; we are giving them a powerful new set of eyes.
I believe we can enter the Artificial General Intelligence(AGI) era, everyone would have a real-time personal assistant in the future.