Dr Matt Wood, AWS | AWS Summit NYC 2018
summary
Announcements show SageMaker adding streaming from S3 so models consume terabytes without local copies, and a batch transform API for bulk offline inference; training time drops roughly 10–25% and Snap trains on over 100 TB of images via the platform. Transcribe adds multi‑channel splitting and 12 new Translate language pairs. A deployable camera with onboard accelerator processes over a billion ops per second and enables demos like rep counting, skeletal mapping, book reading, and species detection.
AI-written summary · about
