Korea Cyber University Lab Pioneers Personalized AI Research

At Korea Cyber University, the PIKA (Personalized Intelligence & Knowledge Adaptation) research lab, led by Professor Lee Gyu-jeong, is embarking on a significant journey into the field of personalized artificial intelligence (AI). Professor Lee, a seasoned practitioner with experience in the military and tech industry, brings a wealth of practical knowledge to the academic setting, focusing on 'practical AI' that can be readily applied in real-world scenarios. The lab's core mission revolves around 'personalized intelligence,' aiming to create AI models that provide optimal reasoning tailored to individual users and their environments.

A key challenge the PIKA lab is tackling is 'catastrophic forgetting,' a phenomenon where AI models lose previously learned information when trained on new, often limited or biased, user data. To combat this, the lab is employing federated learning and continual learning as primary methodologies. These approaches allow for personalized AI development while also emphasizing efficiency and, crucially, privacy protection. This focus on privacy is particularly relevant given the sensitive nature of data often used in personalized AI applications, such as healthcare and psychological counseling.

“

The lab is not a special space, but a process of deeply exploring and solving problems encountered in each person's field.

The lab is actively pursuing the implementation of personalized AI using small language models that operate directly on devices, like smartphones. A notable project involves collaborating with a startup to create a personalized health coach chatbot, fine-tuning a small language model with user-collected healthcare data. Furthermore, they are exploring the therapeutic potential of AI in psychological counseling through personalized models. To bolster trust in AI systems handling sensitive information, the PIKA lab is also researching 'unlearning' techniques to selectively remove personal data traces from AI models.

Currently in its nascent stages, the PIKA lab is fostering a collaborative research environment, welcoming external researchers and actively recruiting student members. This offers early participants a unique opportunity to shape the lab's research direction and culture. Recognizing that many students balance work and study, the lab operates with flexibility, primarily utilizing online platforms for communication and collaboration. Looking ahead, the lab plans to leverage industry partnerships, particularly with AI startups, to access GPU infrastructure for advanced research, including personalized embedding models for retrieval-augmented generation (RAG) and the development of federated unlearning techniques. Professor Lee emphasizes that the lab is not just a physical space but a process of deep inquiry and problem-solving, welcoming anyone with the passion and persistence to advance their field.

“

We are waiting for researchers who will create personalized intelligence that truly understands users together.