Q. Please briefly introduce yourself.

Hello, my name is Hae Beom Lee, and I joined the Computer Division of the Korea University School of Electrical Engineering as an assistant professor in March 2025. My research interests lie in artificial intelligence and machine learning. Before joining Korea University, I received my Ph.D. under the supervision of Professor Sung Ju Hwang at KAIST and subsequently worked as a postdoctoral researcher with Professor Yoshua Bengio at Mila, Canada.
 

Q. What kind of research will you be conducting at Korea University?

Our research focuses on developing System-2 machine learning, which aims to build AI systems capable of complex reasoning processes similar to human conscious thought—beyond the intuitive, automatic responses of conventional System-1 models. While System-1 refers to unconscious recognition models (for example, distinguishing between a dog and a cat), System-2 refers to models that can filter, combine, and reason over interrelated facts to derive conclusions based on given situations and questions. Recently, our main interest has been world model–based AI. Current large language models (LLMs), such as ChatGPT, often suffer from hallucinations and inconsistent responses, largely because they lack an explicit model of the real world. To address this issue, our lab is exploring how to integrate explicit world models into existing LLMs while maintaining their strengths.
 

Q. What led you to this field, and how do you view its future prospects?

During my Ph.D., I mainly studied meta-learning—how to transfer knowledge effectively across multiple tasks. However, around 2020, with the advent of LLMs, a new way to perform meta-learning naturally within the language space emerged, causing the traditional meta-learning field to lose momentum. Nonetheless, the question of what structure meta-knowledge should take remains unsolved. To explore this, I conducted postdoctoral research under Professor Yoshua Bengio, one of the world’s leading AI scholars, where I delved into System-2 deep learning and the significance of world model–based machine learning. At my lab, we aim to investigate the structure of meta-knowledge from various perspectives to address the limitations of current LLMs. Although this research requires fundamental changes in model structure and reasoning mechanisms—making it a challenging endeavor—successful outcomes could provide crucial insights toward developing the next generation of AI systems beyond current LLMs.
 

Q. Do you have any advice for students?

As a new assistant professor, I cannot yet claim deep academic insight compared to my senior colleagues. However, based on my experience, I believe that in engineering, survival depends on adapting to constant change—and predicting those changes is extremely difficult. When I entered university, computer engineering was often undervalued, yet it later gained massive popularity with the rise of AI. More recently, however, the enthusiasm has waned again with the emergence of LLMs. Such shifts are nearly impossible to foresee. Likewise, meta-learning, the topic I devoted countless nights to during my Ph.D., was once a hot research area but has since seen its demand plummet.
 

In AI, I believe there are two main ways to thrive in this ever-changing landscape. First, by cultivating strong theoretical and mathematical foundations during graduate studies, enabling one to analyze and adapt to new models and algorithms as they evolve. Second, by developing an experimental and intuitive research sense to identify emerging, trend-relevant topics quickly. Either path requires readiness for change and the ability to respond to it.
 

I understand that many undergraduates feel anxious about the rapidly shifting job market and research landscape. While such concerns are valid, they are also part of what it means to pursue engineering. I encourage students to spend their undergraduate years exploring what genuinely interests them and where their strengths lie. In doing so, you will naturally find a direction that aligns with your skills and individuality. No matter which path you choose, I hope you will not be swayed merely by appearances but instead build the confidence and capability to follow your own path here at Korea University. I wish you all the best in your future endeavors.