▶ Interview
Byoung-Tak Zhang, Director at the Artificial Intelligence Institute Seoul National University (AIIS)
At the present time, the eyes of the world seem to fall on AI alone. Could AI technology be a revolution that changes our daily lives in a positive direction? We met with Director Byoung-Tak Zhang, a professor in the Department of Computer Science and leading the AIIS, and posed a question regarding the coexistence of humans and AI.
“ChatGPT differs markedly from AlphaGo, an AI go program developed by Google.” Director
Byoung-Tak Zhang started his remark with a smile. “Although AlphaGo attracted people`s
attention, they accepted AI development as a technology in the domain of experts. Language is
open to use by anyone, isn`t it? Despite the previous presence of chatbots, none of them was
successful. The key difference of ChatGPT from conventional chatbots is a super-giant AI capable
of writing a long text by learning copious amounts of literature data. The chatbot technology
with language as a medium has evolved into one that is directly applicable to our daily life and
work. The involvement of AI in our everyday realm has increased the sense of immersion and
familiarity.” Director Zhang said with laughter that although the GPT in ChatGPT is a jargon,
meaning “Generative Pre-trained Transformer,” everyone is already used to it.
“Some portions on AI will be disappointing. Because the visual ability of AI is at the
elementary level, it cannot grasp the details of a situation. AI cannot determine at a glance
whether our current conversation is an interview, a seminar, or a discussion. Although ChatGPT,
which was trained with text, appears to understand everything, it did not reach an understanding
of the context, with its mere focus on text generation. It is on the level of playing with
words. Nevertheless, machines have reached the level of reading certain texts and composing new
texts.”
After completing his bachelor`s and master`s degrees at the Department of Computer Science and
Engineering (CSE), Director Zhang received his doctorate degree at the University of Bonn,
Germany. Since his fellowship as a senior researcher at the German National Research Centre for
Information Technology (GMD), he has been devoted to AI research. He was a visiting scholar at
the Computer Science and Artificial Intelligence Laboratory and the Department of Brain and
Cognitive Sciences, Massachusetts Institute of Technology MIT, and Chair of the Artificial
Intelligence Society, the Korean Institute of Information Scientists and Engineers(KIISE),
leading the AIIS since 2019.
The AIIS, opened in 2019, was founded to integrate and support AI research and expand the
research base. As the largest AI research institute in South Korea, with numerous, world-class,
AI source technology manpower, it plays a central role in discussing and leading the development
and commercialization of AI technology, as well as evaluating the effects of AI on society.
Currently, 300 or more full-time faculty members from 62 departments of the SNU are working as
joint researchers, conducting research on connecting AI to all academic disciplines along with
creative AI source technologies.
“We are making strenuous efforts for the commercialization of AI. We are conducting joint
research by signing membership agreements with seven companies. We are dedicated to industrial
development by conducting research on source technologies tailored to the needs of companies. We
intend to conduct practical research to enable AI to be more applicable to our daily life.”
Director Zhang added that he is extending his attention to a multidisciplinary approach along
with commercialization. “A two-hour meeting with a philosophy professor sometimes resolved years
of inquiry. In the humanities sector, philosophy and anthropology professors are collaborating
in language research. Researchers from various fields, including computer science, engineering,
and voice recognition, are gathering to discover overriding trends.”
We asked Director Zhang to elaborate on the field of research to which he personally pays the
most attention.
“It is human-level AI. Human-level, rather than machine-level, or virtual-world AI!” His answer
was intelligible. “I would like to overcome the limitations of ChatGPT. From a technological
viewpoint, people say that AI does not need to resemble a human. However, I am developing a
human-like cognitive robot with information-processing structures and functions through a
cognitive science program. I am exploring Embodied AI, AI robots with a physical body, that act
and learn in the real world.”
Currently, AI can solve College Scholastic Ability Test problems, but has difficulty brewing
coffee or cleaning a space. However, in contrast to previous functions, AI can contribute
substantially to industrial development if it can work with a physical body. “ChatGPT speaks
only words. Moreover, although it was trained with big data on language, it is difficult to say
whether it has completely understood language. For AI to distinguish between cups and buckets,
it needs to be able to differentiate cups and buckets visually as well as verbally. It should be
trained with deep-learning images. To reach a proper understanding beyond classification, a
level of cognition is necessary that incorporates various types of experiences, including sound
recognition. This is currently impossible. Thus, it requires a physical body. I am devoting my
attention to research that allows AI to learn by itself through sensors. This research intends
to enable AT to perceive and act.”
As he himself has described, Director Zhang is employing his best efforts to incorporate
comprehensive knowledge into AI so that it can distinguish a banana among several fruits on the
table and bring it to him. In this respect, Director Zhang is building up efforts to create
human-level AI.
Early AI researchers had predicted that AI would beat humans at chess in 10 to 20 years. In fact, it took 47 years. AlphaGo was different. Although go may be a greater challenge than chess, AI swiftly learned it. This was thanks to the evolution of “AI that experiences” through deep learning, growing from the method of adding existing knowledge. AI has grown as well. Without realizing it, we have drifted to the center of the world where AI can learn by itself when given the opportunity to gain experience.
It will not be easy to accelerate the evolution of AI. This means that humans and AI will evolve together. This process will be co-evolution along our co-existence. Because those who exclude AI will fall behind, mature people will mull over how to best utilize AI by engaging in higher value-added work using AI as a tool. We will be living together. There are more surprises to come, even though AI will become a mundane being”.
Ending this conversation, we asked for advice from Director Zhang for students or citizens who are interested in AI research. He emphasized “understanding humans.” "I recommend that AI be investigated based on a deep understanding of humans. Despite previous ignorance, we should do this from now on. AI will enable us to grasp other aspects of humans that the conventional academic fields have not touched. AI research could become an academic field that enables us to understand people in a completely different manner.”