#Japan #Tokyo #AI for Science #KARAKURI #Upstage

Karakuri's Entry into the AI for Science Domain

Karakuri, a prominent Japanese company specializing in Large Language Models (LLMs) based in Tokyo, is making strides into the realm of AI for Science. With a focus on enhancing research capabilities for institutions and corporate R&D departments, Karakuri aims to catalyze innovation in scientific research through the practical application of its AI solutions.

The Importance of Entering the AI for Science Domain

The current advancement in scientific research faces obstacles in terms of accelerated speed and reproducibility due to several factors. These include the inefficiencies in data utilization, where researchers often spend excessive amounts of time on complex data analysis and preprocessing, hindering their ability to engage in creative thinking. Additionally, the implicit knowledge barrier complicates matters, as researchers may rely heavily on informal knowledge tied to personal experiences, which can lead to inefficient knowledge management and literature searches.

Karakuri is determined to tackle these fundamental issues by initially addressing the inefficiencies in data and knowledge utilization, with the intent to eventually dismantle the barriers caused by implicit knowledge.

Unique Strengths of Karakuri

Karakuri possesses a competitive edge in this arena, owing to its dual commitment to both the development of high-performance LLMs and the implementation of AI solutions that are grounded in real-world applications.

Technical Superiority

• - Transforming Implicit Knowledge into Organizational Assets: By harnessing the capabilities of LLMs, Karakuri has successfully elevated the quality of customer support through deep analysis of complex knowledge structures, significantly improving accuracy in knowledge organization. This capability can also streamline the extraction and use of expertise buried within research documents.
• - Correcting Personalization Through AI Implementation: The company has also developed AI analytical tools to standardize judgment processes in research organizations that typically rely heavily on the experience of seasoned researchers for optimal experimental condition setting. These tools aim to enhance success rates in experiments and accelerate development processes through data-driven methods.
• - In-House LLM Development: Karakuri has achieved recognition as part of the Ministry of Economy, Trade, and Industry's GENIAC project, offering top-tier performance with its domestically produced LLM series, “Karakuri LM.” This includes the development of the first-ever “Computer-Using Agent” (CUA) in Japan, showcasing their high technical capabilities in tailoring models for various applications.

Overview of the Joint Project with the University of Tokyo and Upstage

Karakuri's strategic initiative is embodied in its recent joint venture with the University of Tokyo's Professor Makoto Okada and the global AI startup Upstage Inc., launching the “AI for Science” support project aimed at revolutionizing scientific discovery processes using generative AI.

Addressing the Challenge of Non-Text Information

In recent years, scientific data has evolved rapidly, bringing diversity and volume, yet many systems struggle to effectively handle non-textual information such as graphs and diagrams. This challenge hampers the efficiency of data quality assessment, thereby affecting analysis, validation, and sharing of research results.

The Integration of Strengths Among the Three Parties

The collaboration will synergize the unique strengths of each entity:

• - Karakuri: Focus on providing AI implementation and engineering technologies.
• - University of Tokyo: Leveraging theoretical insights and leadership in research.
• - Upstage Inc.: Employing LLMs and advanced document parsing technologies for precise information extraction and feature analysis.

Utilizing the High-Performance LLM “Syn Pro”

The project aims to leverage “Syn Pro”, a highly efficient Japanese LLM co-developed by Karakuri and Upstage, recognized for its top performance in the domestic landscape. The three parties will collaboratively redefine AI’s role from a computational tool to a scientific partner that supports researchers' thought processes and reasoning. The goal is to establish a new research model that encompasses automation, transparency, and collaboration while safeguarding intellectual property rights.

Comments from Key Stakeholders

• - Professor Okada from the University of Tokyo: "AI possesses transformative power for science; however, the critical importance of structuring data has not been adequately recognized. By integrating Upstage’s Document AI with Karakuri's LLM, we can see possibilities for addressing long-standing challenges faced by research institutions and companies."
• - Hiroyuki Matsushita, CEO of Upstage: "Over the past six months, we have explored the potential for AI utilization in the scientific research domain with Karakuri and the University of Tokyo. We are thrilled to take this significant step towards the social implementation of 'AI for Science.'"

Upstage specializes in providing high-performance generative AI solutions and has secured investments from leading firms including Amazon and AMD. They recently launched a next-generation LLM tailored for the Japanese market, aimed at transforming complex and unstructured documents into practical business value.

Company Overview

Karakuri embodies the vision of “Friendly Technology,” aiming to enhance DX in industrial and research sectors through advanced AI implementation and LLM development. Since its inception, Karakuri has collaborated with NIMS and established itself in the field of LLM research, influencing various industries with its innovative AI solutions.

Location: 2-7-3 Tsukiji, Chuo City, Tokyo 104-0045, Japan
Established: October 3, 2016
CEO: Shimon Oda
Business Focus: Development and operation of AI solutions specifically for customer support.

For more details, visit Karakuri's official site.