A research group consisting of Professor Kenichi Ohki (affiliated with Institute for AI and Beyond (Beyond AI), Graduate School of Medicine, and International Research Center for Neurointelligence (WPI-IRCN), The University of Tokyo), Professor Teppei Matsui (Graduate School of Brain Science, Doshisha University (at the time of this research: Lecturer, Graduate School of Medicine, University of Tokyo)), Assistant Professor Takayuki Hashimoto, and Assistant Professor Tomonari Murakami (Graduate School of Medicine, The University of Tokyo), and Assistant Professor Masato Uemura (Faculty of Medicine, Kansai Medical University (at the time of this research: Project Assistant Professor, Graduate School of Medicine, The University of Tokyo)) have discovered a new mechanism by which the cerebral visual cortex separates visual information from spontaneous activity.

It is known that the brains of organisms, including humans, exhibit spontaneous activity generated by the brain itself even in the absence of sensory input (e.g., during rest). Spontaneous activity has been observed in various animals and is considered one of the most characteristic features of the biological brain. If spontaneous activity cannot be distinguished from activity in response to sensory input, it could lead to incorrect perceptions or hallucinations. However, in reality, the brain of living organisms can accurately process sensory information. The mechanism behind this has remained largely unexplained. Particularly, it was unclear how spontaneous activity is processed in the visual cortex of mammals, which forms a hierarchical network of multiple areas. Therefore, this study used marmosets, small primates with a visual system closer to that of humans and capable of advanced neural activity measurements, to analyze in detail the relationship between spontaneous activity and visual responses in multiple visual-related areas of the cerebral cortex. The research group discovered that within the hierarchical network of the cerebral visual cortex, the patterns of spontaneous activity and visual responses are similar in lower hierarchical levels, but gradually separate as one moves from lower to higher hierarchical levels. This gradual separation along the hierarchy suggests that the hierarchical nature of the network plays an important role in the separation of spontaneous activity and visual responses. This study not only proposes a new information processing mechanism of the cerebral cortex but also has the potential to contribute to the development of artificial intelligence that incorporates the advantages of the biological brain.

This research was supported by the Beyond AI Joint Project which is a collaborative initiative launched by the University of Tokyo and SoftBank, Japan Agency for Medical Research and Development (AMED), Japan Science and Technology Agency (JST), and Grants-in-Aid for Scientific Research (JSPS）.

The findings of this research was published in the British scientific journal "Nature Communications" on December 4 2024 (UK time).

・The full press release (in Japanese only) in PDF.
https://www.t.u-tokyo.ac.jp/hubfs/press-release/2024/0207/003/text.pdf

＜RELATED LINKS＞
・nature com.
https://www.nature.com/articles/s41467-024-54322-x

・UTokyo Focus
https://www.u-tokyo.ac.jp/focus/ja/press/

A research team consisting of Professor Kenichi OHKI and Jumpei UKITA (a graduate student at the time of the research) of the Department of Physiology, Division of Functional Biology, Graduate School of Medicine, the University of Tokyo, revealed that some of the vulnerabilities in deep neural networks can be mitigated by injecting random noise that mimics the brain in deep neural networks.

Currently, artificial intelligence (AI) is evolving at an accelerating pace, and its underlying structure is based on deep neural networks. Deep neural networks are known to be tricked into producing outputs that are clearly different from that of humans by malicious attacks, known as adversarial attacks. For example, image recognition AI in self-driving cars must correctly recognize a "stop" road sign as "stop" to make the car stop. However, the image recognition AI may not be able to correctly recognize a "stop" road sign generated by a hostile attack, even if it is clearly a "stop" sign when seen by a human. As a result, the car may not be able to stop, leading to a traffic accident. Thus, vulnerability to hostile attacks is one of the major challenges of implementing AI in society.

Incorporating brain properties of animals, such as humans, into AI may help overcome such vulnerabilities. The research team discovered that certain types of vulnerability can be reduced by introducing noise that mimics the randomness of neurons in the brain into deep neural networks. By using this method, the possibility of creating AI that more closely resembles the behavior of humans and other animals is thought to be greater.

This research was supported by the Institute for AI and Beyond of the University of Tokyo, the Japan Agency for Medical Research and Development (AMED) "Project to Elucidate the Entire Functional Brain Network through Innovative Technology", the Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, and CREST-JST.
The research result was published in Neural Networks on 16 September 2023 online.

Professor Ohki is the Project Leader for the Basic Research Project entitled “Development of next generation AI by modeling brain information” at the Institute for AI and Beyond.

・The full press release (in Japanese only) in PDF.
https://beyondai.jp/contents/wp-content/uploads/2023/09/Press-Release_NeuralNetworks_20230916_Web.pdf

＜RELATED LINKS＞
・Published paper (Title: Adversarial attacks and defenses using feature-space stochasticity)
https://www.sciencedirect.com/science/article/pii/S0893608023004422?via%3Dihub
DOI 10.1016/j.neunet.2023.08.022

・Article in English at UTokyo Focus
https://www.u-tokyo.ac.jp/focus/en/press/z0508_00311.html

A research team including Professor Kenichi OHKI of the Institute for AI and Beyond, the Graduate School of Medicine and the International Research Center for Neurointelligence (IRCN) of the University of Tokyo and Assistant Professor Tomonari MURAKAMI also of the Graduate School of Medicine of the University of Tokyo has successfully elucidated a new mechanism of the brain by which a complex neural network, including interareal connections in the cerebral cortex, is efficiently formed during development. This is the first study investigating comprehensively how interareal connections among cortical and thalamic regions are formed during development.
　
This study is expected to be applied in the future to treatments for diseases such as congenital blindness and to circuit formation algorithms to improve an artificial intelligence.

The research result was published in the online edition of Nature on 3 August 2022 (UK time).

Professor Ohki and Assistant Professor Murakami are Project Leader and Researcher respectively in the Basic Research Project entitled “Development of next generation AI by modeling brain information ” at the Institute for AI and Beyond.

The full press release (in Japanese only) in PDF.
https://www.u-tokyo.ac.jp/content/400193638.pdf

Related link: UTokyo HP
https://www.u-tokyo.ac.jp/focus/ja/press/
https://www.u-tokyo.ac.jp/focus/en/press/z0508_00240.html

Abstract URL (Title: Modular strategy for the development of hierarchical networks in the mouse visual system)
https://www.nature.com/articles/s41586-022-05045-w