- Professor Yoshiyasu Takefiji
Introduction of the discussion
Deep learning, which dramatically increases the learning ability of artificial intelligence (AI), is a technology that realizes that machines have the ability to learn by themselves based on data. Deep learning uses an algorithm called a “neural network” for the machine to learn.
A neural network has a multi-layered structure in which nodes (corresponding to one nerve cell) are connected like a network, inspired by the biological mechanism of the human brain. Increasing the number of layers and nodes enables more advanced judgment.
There are several types of neural networks that realize deep learning. Since the problems and strengths that can be handled differ depending on the structure and nature of each type of neural networks, it is necessary to consider what kind of data should be given to AI for what it wants to do.
Finding the problem
When thinking about deep learning, it is important to understand the “black box problem”.
The black box problem refers to the problem that no one can explain why the decision was made when AI that makes some decision is realized by deep learning.
In the case of deep learning, it is necessary to read the constructed network in order to understand how the machine perceives the explanatory variables. If we cannot explain why AI thinks so and why we have reached this conclusion, no one will be able to notice it even if we make a wrong conclusion or estimate.
As a solution to the black box problem, I would like to deepen my research toward the realization of “AI that can fulfill accountability”.
Examining the conclusion
By studying with Professor Yohiyasu Takefuji, who is a SFC’s only neural network researcher, I am sure that I could obtain various perspectives regarding my research and it would open up new horizons for me.
Therefore, I believe that Keio University’s Faculty of Environment and Information Studies is the most suitable place for pursuing my research and social contributions, and I am aspiring to enter your school and study in your laboratory.