Our activities in Fiscal Year 2023 (from April 2023 to March 2024 — the eighth year) include educational and R&D businesses, implemented according to our policy for FY2023.

Educational Business

The goal of our educational business is to help people who conduct research on the WBA approach on a long-term basis.  In FY2023, WBAI held the eighth WBA symposium, two WBA seminars, and a WBA lecture (all in Japanese).  Videos of the paid WBA seminars are distributed on Vimeo for a fee, and those from free events are distributed on YouTube.

The Eighth WBA Symposium

The symposium was held (in Japanese) online on August 1st, 2023 with the theme of “Development and Significance of Whole Brain Architecture Approach in the Generative AI Era” with speakers Shigeru Taguchi (Hokkaido University) and Kyuuri Yamada (Comic writer).  There were also presentations by Hiroshi Yamakawa and Naoya Arakawa (both from WBAI).

It was supported by the MEXT project: “Deciphering and Manipulating Brain Dynamics Underlying Emergence of Behavioral Change: a Multidisciplinary Biology Approach” and cooperated with Paramita, the art group.

In the symposium, Sūn, Yùwěi was conferred the WBA Incentive Awards, and awards for Meritorious Service were presented to Sūn, Xiǎobái for his contributions as the coordinator of the WBA seminar executive committee to organize events such as the symposium and WBA seminars and to Atsushi Kanbayashi for his contributions as the auditor (FY2019-2020).

WBA Seminars

In FY2023, WBAI held two seminars (in Japanese) with the following themes and speakers.

• 38th Seminar: May 11, 2023, Use of ChatGPT and other Tools in Neuroscience

with Takeshi Kojima (the University of Tokyo), Yuta Ashihara (WBAI), Ken Nakae (the Exploratory Research Center on Life and Living Systems, National Institutes of Natural Sciences), moderated by Shin’ichiro Terada (the University of Tokyo))

• 39th Seminar: November 28, 2023, Thinking the Brain, Society, and ‘Intellect’ of AI from the Self-Constructive Nature of Memory, with Hiroki Kurashige (Tokai University), and discussion with Masahiro Suzuki (the University of Tokyo), Hiroshi Yamakawa (WBAI), and Yoshimasa Tawatsuji (the University of Tokyo)

WBA Lecture

A WBA Lecture was held to provide content more technical than WBA seminars.

• 5th Lecture: March 28, 2024, The Role of Cognitive Models in Realizing Whole Brain Computational Models, with Junya Morita (Shizuoka University), Yoshimasa Tawatsuji (Waseda University), and Akira Taniguchi (Ritsumeikan University).

R&D Business

The goal of WBAI’s R&D business is to support research activities in the WBA approach.  Since 2018, WBAI has been promoting research and development of brain-inspired AGI through Brain Reference Architecture (BRA)-driven development, a methodology for developing brain-inspired software, whose specifications are described in Brain Information Flow (BIF) diagrams, which are based on the mesoscopic anatomy of the brain associated with human cognitive behavior, and the Hypothetical Component Diagrams (HCD), which describe computational functions consistent with the BIF (Fig.1).

Fig.1

Currently, BIF and BRA-DB creation, BRA automation, HCD design, and HCD integration are being conducted for BRA-driven development design, and neuroscientific verification and developing functions and environments are being conducted for implementation. We are also working on establishing processes as methodologies as described below. Below, activities financially supported by the MEXT project ”Deciphering and Manipulating Brain Dynamics Underlying Emergence of Behavioral Change: a Multidisciplinary Biology Approach” are noted as (Behavioral Change), and activities supported by the Muhammad bin Salman Center for Science and Technology for the Future at the University of Tokyo are noted as (MbSC).

[Methodology: Process Establishment]

We began establishing the methodology, workflow, and HCD-based implementation process related to BRA-driven development. Through this effort, we established a technical roadmap for the completion of the whole brain architecture and documented its progress. In particular, we proposed a methodology to complete the whole brain architecture through BRA-driven development (Behavioral Change).

We also worked on extending the SCID methodology to apply it to reverse engineering a wide range of brain regions [CNS2023].  A technical map that combines general information processing and entification capabilities as a study to assess the feasibility of human-level AI was also proposed.

[Design: BIF Creation]

We have been working on the creation of the BIF, a brain information flow based on the mesoscopic anatomical structure of the brain. We estimated forward and backward hierarchical structures in the human neocortex (Behavioral Change). We also conducted research on how to use LLM to automatically extract neural projection information and incorporate it into databases to support BRA-driven development (Behavioral Change).

In addition, we worked on and reported on a tool environment to comprehensively extract anatomical neuroprojections from neuroscience literature. It aims to further the work of extracting anatomical neural projections from images as a basis for brain-morphic AI (Behavioral Change).

[Design: BRA-DB]

We have been working on the construction, operation, and manual maintenance of a database for BRA.  The project has successfully developed a database (BDBRA) for extracting neural projections and brain region descriptions from neuroscience literature [CNS2023] (MbSC). The database uses interactive prompts to extract information and aims to leverage neuroscience knowledge for brain-morphic AI development.

We also worked on a method to extract anatomical structures from neuroscience literature and evaluate the reliability of the information, as well as a proposed database to store the results (MbSC). Through the construction and operation of the database, we have successfully automated the extraction of information from the neuroscience literature and made it readily accessible to researchers, providing an important infrastructure to facilitate new advances in brain-morphic AI research.

[Design: HCD Design]

We began to create HCD data focusing on various brain organs. As part of this effort, we conducted research on how to implement higher-order functions of the human brain by linking canonical neocortical loop circuits according to the neocortical connectome and explored its potential. We published the results of this research (Behavioral Change).

We also conducted a survey of information extraction techniques on dysfunctional animals from the neuroscience literature (Behavioral Change). We also proposed a roadmap for affective interaction using large-scale language models, focusing on the mechanisms of fear and motivation in the amygdala.

We also worked on the construction of a cognitive architecture by coupling large-scale language models that simulate brain regions. The project successfully prototyped a conversational AI that can interpret and intervene at the brain region level. We also began developing a biologically plausible amygdala circuit model of context-dependent fear disappearance and recurrence (MbSC).

While computational functions consistent with BIF have been described by hierarchical HCD, it was inflexible and inappropriate for reverse engineering. Thus, we have been studying a Function Realization Graph (FRG) to describe functional hierarchy diagrams on top of HCDs.

[Implementation: Architecture Implementation]

We undertook R&D efforts to test whether specific tasks that humans and other animals are capable of solving can be performed by computational mechanisms similar to those in the brain, including correlations with neural activity.  As part of this project, we successfully implemented and validated a probabilistic generative model inspired by the hippocampal formation.  This work represents a novel approach that combines neuroscience theory with computational modeling (Behavioral Change).

[Operation: BRAES]

We worked on the development of BRAES (BRA Editorial System), which includes the BRA database, submission review flow system, information publication portal, automated review tools, and data version control (MbSC). The system aims to provide a platform for storing and sharing hypotheses on neuro-functionality.  BRAES integrates neuroscience findings with AI techniques to provide a valuable resource to the research community.  The development of BRAES aims to optimize the process by which researchers can efficiently submit, review, and publish hypotheses about neural function, to accelerate the progress of neuroscience research and open new avenues for a better understanding of the brain [presentation at INCF].

[Other: AI Alignment]

We explored approaches with the goal of making AGI behaviors and decision processes interpretable in a way that is understandable to humans based on brain-based interpretability. This effort is based on the idea that the behavior of AGI must be understandable and predictable to humans in order for the technology to be accepted and trusted by society.  Our research seeks ways to make AGI’s decision-making process more transparent by mimicking brain functions and structures, making the results easier for humans to interpret. This approach offers a new solution to the important technical and ethical challenge of AI alignment.

Forming open AI development communities

We exchange information about WBA in a community on Slack and have semi-regular online meetings (in Japanese). We are also working with Ritsumeikan University on the implementation and validation of a probabilistic generative model inspired by the hippocampal formation, as described in the architecture implementation section above.

WBAI Activities and Volunteering

As in previous years, activities such as the organization of the event were realized under the leadership of the WBA-Seminar Executive Committee, consisting mainly of volunteers.

Financial Statements for FY2023

The balance sheet and cash flow for FY2023 are presented below (Table 1 and Table 2).

The revenues were 0.84 million yen and the expenditure totaled 0.80 million yen for administration and 0.99 million yen for operation (total expenditure was 1.78 million yen to yield 0.94 million yen deficit).

The businesses of WBAI have been financially supported by supporting members.  (As of March 2024, there were seven supporting members consisting of enterprises and individuals.)  The other incomes include fees from the participants of the WBA seminars and crowdfunding (in Japanese).

Honoraria were paid to the speakers of the WBA Seminars.  The subcontractor/outsourcing expenses for operation includes the cost of R&D,  which was mostly carried out with the budget of exterior research institutions.

The expenses for secretariat personnel (shown in Subcontracting/outsourcing Expenses) were apportioned by 50% (50/50) for operating and administrative expenses.  The remuneration was paid to an accountant office.  The rent for the office was free of charge by courtesy of Garm Inc.

Table 1: Balance Sheet (as of March 31, 2024)

Items					Amount (JPY)
Ⅰ	Assets
	1.	Current Assets
		Cash and Saving Account			7,427,783
		Accounts Receivable			26,520
		Total Current Assets			7,454,303
	Total Assets				7,454,303
Ⅱ	Liabilities
	1.	Current Liabilities
		Advance Received			80,000
		Deposit			4,084
		Total Current Liabilities			84,084
	Total Liabilities				84,084
Ⅲ	Net Assets
		Retained Net Assets at the Beginning of Period			8,314,896
		Net assets variation			-944,677
	Total Net Assets				7,370,219
	Total Liabilities and Net Assets				7,454,303

Table 2: Cash Flow

Items					Amount (JPY)
Ⅰ	Recurring Revenues
	１	Fees
		Fees from Regular/General Members			142,000
		Fees from Supporting Members			560,000
		Total Fees			702,000
	２	Other Revenues
		Interest Income			71
		Other Income			111,409
	Total Recurring Revenues				840,120
Ⅱ	Ordinary Expenses
	１	Operating Expenses
		⑴	Total Personnel Expenses		0
		⑵	Other Expenses
			Honoraria		90,000
			Subcontractor/outsourcing Expenses		529,597
			Communication		267,110
			Prizes/Awards		100,000
			Other
			Total Other Expenses		986,707
		Total Operating Expenses			986,707
	２	Administrative Expenses
		⑴	Total Personnel Expenses		0
		⑵	Other Expenses
			Subcontractor/Outsourcing Expenses		528,000
			Remuneration		264,000
			Other		6,090
			Total Other Expenses		798,090
		Total Administrative Expenses			798,090
	Total Ordinary Expenses				1,784,797
		Net Assets Variation of the Year			-944,677
		Net Asset brought forward			8,314,896
		Net Asset carried forward			7,370,219