Azure AI, Power new possibilities for every organization
summary
Demo shows Form Recognizer extracting invoice fields and image recognition flagging a windshield crack at 92% confidence, and Chevron automates hundreds of PDF reports via RPA. Azure Machine Learning AutoML runs many algorithms (about 50), produces a best model in ~24 minutes with model explanations highlighting engine size and mpg as top features, and deploys to AKS or Power BI. Heathrow prototype runs 3D VTK scans (~50 MB) on Azure Stack Edge using four containers every 200 ms and flags likely illegal wildlife parts in seconds. Responsible AI tools include Fairlearn, differential privacy and homomorphic encryption.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 2:55Two fundamental trends
- 4:48Making AI accessible
- 6:05Cognitive Services
- 7:35Example
- 10:05Case Study
- 12:47Create your own model
- 13:55Data preparation
- 15:05Azure ML Service
- 16:00Azure Machine Learning
- 16:57Machine Learning Designer
- 17:53Automated Machine Learning
- 19:32Machine Learning Demo
- 19:48Machine Learning Workspace
- 23:32Create new experiment
- 29:00Data Flows
- 34:02Real Life Example
- 38:12Responsible AI
- 41:02Fairness in AI
- 42:32Differential Privacy Toolkit
- 43:45Future of AI
- 45:57Azure Supercomputer
related talks
32mUnleash your AI potential
01h 00m5 Levels of AI in Microsoft Azure
31mAzure AI Innovations at Microsoft
57mEasier AI allows everyone to build smart apps
29mAI for All
30mAI: Empowering us all
48mBuilding and using AI models responsibly
40mBuilding secure and efficient AI solutions in Azure with Partners
