Skip to content. | Skip to navigation

Personal tools

Home > News & Publications > Events > From Pixels to Artificial Intelligence: The Parallel Computing Journey of GPUs

From Pixels to Artificial Intelligence: The Parallel Computing Journey of GPUs

Start: 11/7/2017 at 3:30PM
End: 11/7/2017 at 4:30PM
Location: DeBartolo Hall, Room 140
Add to calendar:
iCal vCal
From Pixels to Artificial Intelligence: The Parallel Computing Journey of GPUs

ABSTRACT: Translating languages. Making homes and cities intelligent. Diagnosing cancer. Teaching autonomous cars to drive. These are just a few things being enabled today by AI, and specifically, deep learning. Deep learning uses massive amounts of data to train complex "deep" neural networks that can detect, classify, translate, and make decisions. The changes brought about by AI are accelerating at a pace never seen before in our industry. But the approach demands that computers process oceans of data at precisely the time when Moore’s law is slowing. The NVIDIA GPU (Graphics Processing Unit) architecture — designed for processing massive parallel workloads — can reduce the time required to train complex deep neural networks from months to days, and it can provide massive  inference performance on networks deployed to low-power devices at the edge.

How did the GPU, which started as a special-purpose accelerator for rendering 3D computer games, become the platform of choice for AI? In this talk I'll take you on a journey from pixels to parallel computing to show you the past, present and future of computing on GPUs. I'll use examples from my work and the work of others to illustrate the progression of GPUs from graphics, to General-Purpose computing on GPUs (GPGPU), to the NVIDIA CUDA parallel computing platform and the acceleration of deep learning.

SPEAKER BIO: Mark Harris is chief technologist for GPU Computing Software at NVIDIA, where he works as a developer advocate and helps guide NVIDIA's GPU computing software strategy. His research interests include parallel computing, general-purpose computation on GPUs, physically based simulation, and real-time rendering. Harris earned a B.S. in computer science at the University of Notre Dame in 1998 and an M.S. (2000) and Ph.D. (2003) in computer science at The University of North Carolina at Chapel Hill where he applied early GPUs to fluid dynamic simulation and rendering clouds. In 2002 he recognized a nascent trend in computing and coined a name for it: GPGPU (General-Purpose computing on Graphics Processing Units), and founded to provide a forum for those working in the field to share and discuss their work. He now lives and works off-grid (solar power, rainwater, and 3G broadband!) with his wife and daughter in the mountains of the north coast of New South Wales, Australia.

« April 2018 »