HRI-JP HOME > Research > Projects > Uncover the high adaptability of perception and recognition.

Research

Projects

Current projects in HRI-JP

Intelligence Science

Uncover the high adaptability of perception and recognition.

HRI-JP Senior Scientist Shigeaki Nishina

Uncover the high adaptability of perception and recognition.

1. The brain is characterized by its highly developed adaptability.
The brain is characterized by its highly developed adaptability.

We are continuously bombarded with various stimuli information that comes from the outside world. The information that flows through the sensory system is not only immense but also continuously changing, and no two moments are the same. The brain picks up and processes necessary information, and orders our body to take appropriate action. This is startling by itself, but the brain is not just processing information. It changes the processing of information to be more efficient when necessary. There are important pieces as well as unimportant pieces in the flow of information. Processing in the brain will be unstable if the brain also adapts to the unnecessary information. The brain is skillful in distinguishing what information to adapt to and what not to adapt to.

Currently, it is impossible to make an artificial intelligence system that has this advanced level of adaptability. We can set specific situations and make a machine learn how to optimize its actions in those situations. However, it is a very difficult task to adapt to unknown situations while maintaining stability. If we can understand how to do this, we cannot only increase our understanding of how the brain works, but also evolve artificial intelligence that acts in different environments. At this point we cannot reach this goal quickly, but we think that pursuing the nature of adaptive information processing is the most important approach to understand the intelligence.

2. The key to true intelligence is in the adaptability of human perception and recognition.

We can obtain clues to understanding the high adaptability of the brain by conducting carefully controlled experiments. We do not gain much by simple experiments like just showing something and obtaining scores. Experiments should take into account insight from neuroscience and information science, and computational validity must be also considered. Detailed hypotheses on information processing by the brain need to be tested with various parameters, including conditions on space and time or task load. We clarify the details of the human’s behavior by comparing the obtained input-output relationships to the hypotheses. Brain activity obtained by methods such as fMRI, are used in addition to behavior measurement to check consistency with internal parameters in the hypothesized models. The goal is to understand the mechanism of the human and the brains as an advanced information processing system. We hope to elaborate on insight accumulated with these methods to build a fundamental theory for machines with true intelligence that can act like people in a changing environment.

Learning of Unattended Visual Information
3. Focusing on the interaction between the brains and machines in an environment.

This project focuses on humans as the target of study. However, we also consider how the human build societies in an environment; where the individual brains interact, directly or indirectly. Perception and behavior act as the interface of interactions, and information flowing in the interface is consciously or unconsciously processed.

There are recent reports on the importance of unconscious information processing. However, the full details are not understood. We investigated the interaction in vision between unconscious adaptation and conscious information processing, and found that unconscious, or automatic learning is restricted when conscious information selection is spatiotemporally nearby. We believe that this is one mechanism for avoiding conflict between stability and adaptability, but the neural mechanism is still unknown. We need to deeply investigate from various directions to understand the complex system of the brain.

Considering the relation between people and machines in an environment is also important. As machines are already a member of this environment, how machines should behave is a very significant issue. Understanding how people recognize and behave is extremely important in thinking how machines should interact with people.

4. What is the problem? It is important to constantly question the fundamental cause.

In fundamental research, the applications of results are often difficult to visualize during the course of the research, therefore the importance of a research is evaluated by how intrinsic the issue is. We need to ask fundamental questions as to what characterizes the human, and what are the important characteristics of the mysterious brain, when we try to understand how people recognize the world.

There are not many industrial research institutes for pure fundamental research. Our institute shows our determination to not only contribute to academia but also to clarify the nature of things. HRI provides a research environment where ideas are more unbound compared to universities that depend on public research funding. I hope to make the best of this unique environment and generate results for a new future.

Shigeaki Nishina
Shigeaki Nishina joined HRI-JP as a senior scientist in 2009, after seven years at ATR Computational Neuroscience Laboratory, Kyoto, Japan and three years at Boston University in Boston, USA. He received Ph.D. degree in experimental psychology from Kyoto Univeristy in 1999. His research interests include visual attention, object recognition, surface representation, and perceptual plasticity.
Shigeaki Nishina
Research projects in intelligence science
Voice interface builds a new relationship between humans and robots.
Understand the mechanism of human “behavior” and “cognition” from the neural network in the brain.
Building an information processing mechanism to understand the environment and situations based on sounds.
Uncover the high adaptability of perception and recognition.