Lakshmi Narasimhan of Intel at Machine Learning Developers Summit 2019
summary
Claims that only ~3–4% of AI projects are in production, that India rose to #1 in mobile data consumption after Jio, and forecasts of 50 billion devices and 200 billion sensors. Shows concrete tooling and results: DAL, MKL/MKL‑DNN, NVIDIA GPUs for training, Movidius VPU and Altera FPGAs for inference, Nirvana accelerators, and OpenVINO’s model optimizer + inference engine. Demo: Intel Distribution for Python yields ~20× speedup on k‑means; OpenVINO runs a retail vision pipeline at ~7–8 FPS on CPU, ~3–4 FPS on Movidius.
AI-written summary · about
chapters (13)
- 0:00AI and data revolution
- 2:31Data as new oil
- 5:25AI as next compute
- 7:55Simplifying AI/ML concepts
- 11:04Intel AI hardware
- 16:21CPU vs GPU for AI
- 18:48Intel ML software tools
- 22:21Intel Python distribution demo
- 33:26Optimized deep learning frameworks
- 36:18OpenVINO toolkit
- 40:15OpenVINO retail demo
- 45:14Intel AI ecosystem
- 48:28Q&A
related talks
24mTools & Techniques for a cost optimized Analytics & AI pipeline
30mAI: unleashing the next wave
13mAccelerating AI algorithms
05mAccelerate AI development from cloud to client
42mAI Everywhere– accelerate your development from PC to cloud
20mAI : Solving challenges and driving Innovation across Industries
16mDeveloper-ready AI PCs - out of box support for DirectML on NPU & GPU
20mOperating Frontier Intelligence at Scale