日本語 
Yokohama Rubber Co., Ltd. and dotData, Inc. have begun jointly developing an AI agent designed to support hypothesis development, interpretation, and decision-making by engineers involved in manufacturing research and development (R&D).
Before applying the technology fully to actual R&D projects, the two companies will verify the functions required and assess its technical feasibility. The development will combine technical documents, such as design information and experimental reports, with knowledge accumulated by engineers and features identified from measurement data through dotData’s automated feature engineering technology.
The AI agent will consider the objectives, circumstances, and constraints of each R&D project and provide engineers with information to support their decisions. Rather than presenting a single “correct” answer, the system will operate on a Human-in-the-loop basis, giving engineers multiple possibilities, comments from different perspectives, and potential risks to consider.
Manufacturing research and development creates a lot of data from experiments, simulations, manufacturing processes and testing.. Changing that data into new technical ideas takes engineers to understand it using their knowledge of the research goals the conditions of the experiments the limits they face and their past experience.
The companies also highlight the value of abduction which is creating hypotheses in research and development. Since new research might not have a right answer engineers must come up with many possible explanations for what they see and then test them to find the best options. The new AI agent is meant to help with that process by allowing engineers to look at data and information, from angles.
The system under development will provide four main forms of support. First, it will reference design information, experimental reports, research reports, and engineers’ accumulated knowledge to understand the background and history of an R&D project.
Second, dotData’s automated feature engineering technology will explore patterns in measurement and experimental data without limiting the analysis to factors engineers have identified in advance. These automatically generated features can provide quantitative insights that, when combined with technical knowledge and design information, may reveal relationships that are difficult to identify through existing hypotheses or experience alone.
Third, the AI agent will provide comments and potential risks from multiple technical perspectives, including materials, design, analysis, production, and evaluation. It will also raise questions that can help engineers examine whether alternative interpretations exist or whether relationships could change under different conditions.
Finally, engineers will remain responsible for interpretation and decision-making. They will develop hypotheses based on the information presented, validate them through experiments and analysis, and feed the results back into discussions with the AI agent. The companies will also examine ways for engineers to use AI-agent interactions to communicate their analytical objectives and areas of interest, including support for operations and settings required for automated feature engineering.
Also Read: Carbon EX Merges with exroad to Expand Carbon Market Support
The project builds on Yokohama Rubber’s HAICoLab AI utilization framework, which the company has promoted since 2020 to encourage collaboration between people and AI in developing products, processes, and services while supporting employee growth. Yokohama Rubber has already used dotData in areas such as tire design and rubber mixing, where engineers interpret automatically generated features from multiple perspectives to discover new insights and hypotheses.
The new joint development will extend this approach into an AI-agent model. Engineers will compare the features and comments provided by the AI with their own expertise and experience rather than simply accepting the results. Through this process, the companies aim to help engineers expand their perspectives and develop new approaches to thinking and decision-making.
Going forward, Yokohama Rubber plans to apply the AI agent to actual manufacturing R&D themes and further develop HAICoLab practices that combine human creativity with data utilization. dotData will continue developing technology that combines quantitative context generated through automated feature engineering with AI agents, with the goal of supporting applications across a wider range of manufacturing R&D environments.
English