|▲ Huh Jeong-gyu (Assistant Professor, Major in Bigdata Financial Engineering Convergence)|
The match between AlphaGo and Lee Sedol in 2016 was an incident that made us realize that the era of AI was closer than we thought. Because the number of moves that one can make in Baduk is greater than the number of atoms in space, it is important to focus on really valuable moves and only to make the best possible moves in Baduk. Many people thought that computers were not developed enough to conquer Baduk because they were regarded as simple devices that would not be able to consider all the moves. However, AlphaGo's win proved the fact that computers can think similarly to humans, which made us think again about the boundaries between humans and computers. A craze for AI swept around the world after the incident, and AI has become the magical word for being able to solve any kind of problem. The reason for AI’s success is considered to have three parts. The first part is big data obtained from the use of the internet, the second is a huge artificial neural network algorithm, and the last one is a highly developed GPU that can process the big data.
If someone asks about my research field, I will usually answer that I work on applying AI into financial engineering. However, I have actually spent just three years in fusion research between AI and financial engineering. Before that, I used to study financial engineering with traditional mathematical/statistical methodology which was not related to AI. However, AI has developed into my main research tool because I recognized it can have a great impact on financial engineering. I will present an example based on the research I have done. Pricing a financial product is generally considered a mathematical problem. However, I presented an indirect way of solving the problem by creating AI that can estimate prices of products almost exactly. Let us assume someone is trying to solve the problem in a mathematical way at present. This is meaningful in terms of the development of knowledge. But, since there is already a much easier way, AI, to solve the problem, this type of study may no longer really be necessary.
Most people are established in their own current fields, unless they are choosing their fields for the first time such as students entering university, and these people who are established in their field are naturally unwilling to solve their problems in an unfamiliar way such as using AI. However, as mentioned before, AI has numerous possibilities to solve problems that preexisting methods cannot solve, meaning that current knowledge might become useless in the future. I will mention again a little more about Baduk. Many fine Baduk programs like EunByul Baduk existed before AlphaGo, which are known to have skill levels similar to pro players. The developers continuously upgraded the program with their knowledge, but they thought that these programs would take a fairly long time to beat a grandmaster in Baduk. Unexpectedly, however, the AI made with reinforcement learning was able to beat a grandmaster in Baduk easily. Now no one uses those preexisting methods to create Baduk programs. Of course, I do not think this is an appropriate example for all fields. However, at least once, I recommend that you investigate your problems using AI, or perhaps start studying the basic concepts of AI.
By Huh Jeong-gyu, Assistant Professor, Major in Bigdata Financial Engineering Convergence
허정규 빅데이터금융융합전공 교수 email@example.com