NEC Corporation has been awarded a contract to conduct research on underwater acoustic foundation models using self-supervised learning. This project is part of the 'Innovative Breakthrough Research' program by Japan's Defense Innovation Science and Technology Institute, under the Agency for Defense Equipment.
The project aims to develop a world-leading foundation model for underwater acoustics by fiscal year 2027. NEC plans to train a model on extensive underwater acoustic data to create general-purpose generative AI, akin to large language models, for underwater phenomena.
The foundation model is expected to enhance the accuracy and speed of understanding underwater environments, supporting applications beyond defense, including marine life exploration, resource exploration, environmental monitoring, and earthquake prediction.
Underwater sound waves, captured as acoustic data via sonar and hydrophones, require significant analysis by experts. However, the current process is time-consuming and limited in accuracy. NEC aims to address these challenges by leveraging its 90 years of experience in sonar technology and its proprietary AI technology, 'cotomi,' along with self-supervised learning methods.
The project will utilize large-scale underwater acoustic data collected through public-private partnerships. This approach aims to overcome the difficulties associated with collecting labeled data in underwater acoustics, which has hindered the development of large-scale AI models in the field.
NEC will collaborate with the Defense Innovation Science and Technology Institute to demonstrate the foundation model's effectiveness and complete its development by the target deadline. The initiative is expected to strengthen defense capabilities and advance the maritime industry through dual-use applications.
The 'Innovative Breakthrough Research' program is part of a broader initiative by the Agency for Defense Equipment to foster innovative technologies that significantly enhance defense capabilities or transform societal functions.
Comments are closed.