[Edaily Reporter Park Jung-Soo ] #Simplatform Co., Ltd., a company specializing in physical artificial intelligence (AI), is embarking on the development of next-generation AI technology that identifies the causes of phenomena based on the laws of physics and mathematics.
Simplatform Co., Ltd. announced on the 25th that it has begun research and development as a joint R&D partner in the Ministry of Science and ICT’s “Next-Generation Artificial Intelligence (AI) + Science and Technology (S&T) Basic Technology Development Project.”
This project is being pursued to secure scientific AI technology that goes beyond the limitations of existing data-driven AI by incorporating principles of physics and mathematics into the design and training processes of AI models. The Ministry plans to support the development of highly reliable analysis and prediction technologies, as well as next-generation AI architectures, from 2026 to 2031.
Among the four projects, Simplatform Co., Ltd. will collaborate with the Korea Advanced Institute of Science and Technology (KAIST) and Ajou University to develop a “Mathematics and Physics-Based Causal AI Analysis and Prediction Model.”
While existing AI has focused on predicting results based on training data, the core objective of this research is to enable AI to identify the causes of phenomena and the physical laws governing them, even when environmental conditions or circumstances change.
Simplatform Co., Ltd. will be responsible for verifying whether the developed AI technology functions effectively in real-world industrial settings. Using its proprietary “Thing Driver” and “NUBISON Datalake,” the company plans to collect industrial data—such as equipment, sensor, and process logs—and refine and standardize it for use in AI training and evaluation.
The company will also establish a benchmarking and industrial validation system to verify the AI’s predictive accuracy and reliability under various operating conditions and environmental changes.
This technology differs from existing data-driven AI in that it incorporates physical and mathematical laws into the AI training process. Through this approach, the research team aims to develop models that are lightweight and fast, achieving accuracy levels comparable to existing models even with relatively small amounts of data.
By interpreting industrial data based on the laws of physics, the team expects to enhance explainability, reproducibility, and consistency.
Ultimately, the research team plans to build an integrated “Causal AI Research and Validation Platform” that links data, AI models, evaluation code, and protocols.
In the future, they intend to expand the scope of application to include semiconductor and manufacturing processes, as well as biosignals, the environment, and climate. Through this, they aim to reduce the time and cost associated with repetitive simulations and verification, and to accelerate the industrial adoption of reliable AI-based analysis and prediction technologies that remain trustworthy even when conditions change.