ABCI – Envisioning High Performance Human-Centered AI for Industry
TimeWednesday, November 14th3:30pm - 4:15pm
DescriptionArtificial intelligence (AI) is leading the world’s industrial revolution in relation to IoT, Big Data, Robot, advanced biotechnology, new manufacturing industry, and so on. Although there is an ongoing discussion of the AI singularity, there is a more immediate need to focus on practical applications of AI. In these applications, AI can be defined as sensors coupled with deep learning, which can substitute for some human sensory functions.
Thus, AI will relieve workers from the burden of some physically or mentally intensive tasks. AI algorithmic techniques have developed dramatically with a high dependence on big data sources as training data. More importantly, at a a practical level, I will highlight contributions from HPC technology that are indispensable for learning. Until now, we have supported AI as a system, but cooperation between HPC applications and machine learning will intensify and become a very important factor in the future of AI.
AIST is constructing one of the world’s largest AI focused, open, and public computing infrastructure called ABCI (AI Bridging Cloud Infrastructure). It will support not only industrial AI applications to be deployed for use in society, but also academic AI research. To facilitate this, we are connecting ABCI to a global, high-speed network research platform to enhance AI research with a broad number of institutions. One of the first grand challenge projects we are supporting is mapping whole brain-scale signal integration events in zebrafish. This will open the possibility of creating novel, neural deep learning architectures of the fish brain to be used in a wide variety of applications including robotics, sensor design, and behavioral sciences.
ABCI will be operational shortly in 2018 and will serve as a reference architecture for data centers combining HPC and AI to discover new innovations in the future.