Jim McHugh, NVIDIA | Big Data new York City 2016
summary
Company demos a DGX one deployed to Jan LeCun and a custom car called bb-8 outfitted with 8 cameras that learns: after 100 hours it drives around cones, after 3,000 hours it navigates Jersey country roads. DGX hardware pairs 8 GPUs NVLink (claimed as equivalent to 250 x86 servers) to cut model training and query time from days to hours or minutes. Accelerated databases like Kinetico, mapD and scream plus Graphistry visualize hundreds of millions of events and drill to hundreds of thousands interactively.
AI-written summary · about
