The AI stack from chip to analytics

on YouTube ↗09m

" summary

Intel presents an end-to-end AI stack: client and server silicon (Xeon, Xeon Phi, FPGA) plus optimized software (MKL, inference SDKs) that runs on existing clusters. New Xeon Scalable chips claim 3x Hadoop/Spark throughput, 2.7x decision support and 93% microbenchmark gains, improving real-time and streaming responsiveness. BigDL embeds deep learning training into Hadoop/Spark so models train in place on current infrastructure instead of migrating to new clusters, reducing cost and latency.

AI-written summary · about

" chapters (5)

1. 0:00Intro
2. 1:32Where to start
3. 3:06Performance and optimization
4. 4:10Moores Law
5. 6:27Whats next

" related talks

• 09mTackling edge computing with Intel's AI Builders programFiaz Mohamed · Artificial Intelligence Conference 2018 New York
• 04m4 real-world AI deployments at enterprise scaleGadi Singer · O'Reilly Artificial Intelligence Conference 2019 - New York, New York
• 30mAI: unleashing the next waveMilind Hanchinmani · Anthill Inside 2017
• 13mCombining algorithms and quality data to build better AI platformsMa Ziya · Strata Data Conference 2018 - New York, NY
• 50mAI acceleration at the edgeJonathan Lee, Michael Kleiner, Allan Gagnon, Alex White, Mohan Potheri · AWS re:Invent 2023
• 24mTools & Techniques for a cost optimized Analytics & AI pipelineAnand Kulkarni · MachineCon 2022
• 06mCollaboration and openness drive innovation in artificial intelligenceMartin Hall · Strata + Hadoop World 2016 - New York, New York
• 11mGadi Singer (Intel) interviewed at the O'Reilly Artificial Intelligence Conference NY 2019Gadi Singer · O'Reilly Artificial Intelligence Conference 2019 - New York, New York