[CENTRA Webinar] "ABCI: An Open Innovation Platform for Advancing AI Research and Deployment" was held on April 10th, 2019 at 8-9pm US EDT by Global *CENTRA project, via ACIS Lab, University of Florida. *globalcentra.org
Dr. Ryousei Takano is a research group leader of AIST, Japan (aist.go.jp/index_en.html). He received his Ph.D. from the Tokyo University of Agriculture and Technology in 2008. He joined AXE, Inc. in 2003 and then, in 2008, moved to AIST. His research interests include operating systems and distributed parallel computing. He is currently exploring an operating system for heterogeneous accelerator clouds. Dr. Takano is a member of both CENTRA and PRAGMA (Pacific Rim Grid Middleware and Applications Assembly).
With the recent global interest in artificial intelligence (AI), Japan's National Institute of Advanced Industrial Science and Technology (AIST) is positioning itself to be one of the leaders in the development and deployment of AI applications for both industry and society. AI technologies, especially machine/deep learning, are highly dependent on computation at scale using big datasets to train AI algorithms and improve their accuracy. To address this computational challenge, AIST built ABCI - AI Bridging Cloud Infrastructure (abci.ai), the world's first large-scale, open innovation platform for AI research and development. ABCI consists of 1088 computation nodes with 4352 NVIDIA Tesla V100 GPUs and 22 PB storage system, all housed in an ultra-green data center building. This platform integrates high performance computing, big data, ultra-high bandwidth, and low latency hardware in a modern design to enable better cooperation between HPC applications and machine learning, which will become a very important factor in the future of AI. To facilitate academic AI research, we are connecting ABCI to a global, high-speed network research platform to extend research opportunities to a broad number of international institutions. In this talk, I will give an overview of ABCI and the potential applications and collaboration utilizing the computational capacity and capability.