Hanseok Ko, a globally recognized artificial intelligence expert, has returned to his alma mater to bring his three decades of experience to help make The Catholic University of America a center for the ethical development of the technology.
As a professor at Korea University, he led development of the core audio/speech interface technology for Hyundai/KIa and man-machine interface initiatives for robotics, consumer electronics, and healthcare at the research center he ran at the university. He is an IEEE Signal Processing Society Distinguished Lecturer, International Speech Communication (ISCA) Fellow, and Institution of Engineering and Technology (IET) Fellow. Among his many honors, he is the inaugural Newton-Bennett Endowed Chair for the School of Engineering at the University.
Ahead of his installation on April 14, he spoke about his vision for the University including the new artificial intelligence graduate and undergraduate program.
What brought you back to your alma mater?
This is my school. I earned my PhD here at Catholic University, and the Navy gave me a fellowship to complete my studies. After that, I became very successful as a professor and researcher. I want to give back what I have received from my University and my country.
What is your vision for artificial intelligence at the University?
I am envisioning Catholic University as a proving ground for advancing artificial intelligence science and technology, especially for the ethical aspect of artificial intelligence.
Who is going to place the guardrails to ensure that AI will benefit humans? We are. There’s no better environment than Catholic University with its enthusiasm to address the scientific issues of AI so that it can be used to fulfill our mission in Washington DC, which is arguably the center of the world.
What about the University makes it the perfect place for artificial intelligence research?
Whenever I walk around campus, I see all of these light poles with signs with messages about serving God. It is so refreshing. I have not seen anything like that in many years in academia. I am so touched and inspired by these signs. They remind us of our fundamental humanity and calling to become closer to God.
I’m coming from the technical side, but the guardrails need to be thought out together. We have the School of Theology, the Busch School of Business, the Columbus School of Law, and a liberal arts tradition. We can mobilize our different resources to make the University a center of interdisciplinary AI research.
The University is launching an artificial intelligence program for undergraduates and graduates this fall. What do you plan on sharing with students?
They need to first learn the skill set to handle a server equipped with GPU, short for graphic processing unit. A CPU processes data one at a time, but a GPU is designed for parallel processing so it’s useful for developing deep learning models.
With deep learning models, scaling is always an issue. Large language models are measured by the number of parameters, which refers to how much the model has to learn. The number of parameters for OpenAI is an estimated 1 trillion, which is going to require at least 10,000 GPUs.
When students come in, I want them to understand the gravity of the scaling. They need to be able to look at billions and billions of parameters and develop models through the training process. GPUs are very expensive. I’d like 10,000 GPUs to handle large scale language models but it would cost millions of dollars. We need to have the infrastructure to explore and develop some of the learning models here on campus. We are going to have some of the servers with GPUs locally, physically on campus, and also plan on using a cloud server. We want students to be able to have 24 hour access to develop their skill set. Once they have the skill set, I think they can develop the machine learning models to advance science and apply to various AI use cases that benefit humans and society.
What aspect of AI are you most excited about?
There are two types of AI making a lot of headlines today. One of them is generative AI. Generative AI is when you input some text or a description, and the system outputs things like videos, images, or other content. That’s an example of reactive AI. There are many algorithms out there that support this, such as ChatGPT. But the name of the game today in AI is agentic AI. Agentic AI has the humanlike ability to solve complex problems beyond an immediate response. It breaks a task down into different subtasks, and each of those can be handled by a specific agent. Agentic AI is a combination of different agents all working together — just like how the human brain would solve a complex problem. It’s not about replacing people, but about helping with time-consuming processes so we can spend our time better.
