New multimodal vision AI models and their practical applications
summary
Shows that GPT-4 Turbo with Vision is GA in Azure and is hooked into Azure AI Search with vision embeddings, image/video prompting, and configurable content filters (low/medium/high after responsible‑AI review). Live demos include floorplan CAD work where the model reads measurements, estimates a garage as 19×10 ft and, after object‑detection annotations, gives a 4 ft maximum dining table width; Phi‑3‑vision is a 4.2B multimodal model with 128K context that computes a chart gap of 17; in‑car demos use DALL·E 3 and report ~200,000 beta users.
AI-written summary · about
chapters (9)
- 0:00Introduction to multimodality
- 2:42GPT-4 Turbo with Vision
- 6:12CAD diagram analysis demo
- 14:13Video analytics and search
- 15:35WPP customer use cases
- 20:17Phi-3 Vision capabilities
- 24:48Phi-3 Vision demo
- 28:43Mercedes-Benz in-car AI
- 34:09Future of Azure AI
related talks
25mMultimodal Conversational Interfaces with GPT and Vision AI
51mThe Next Frontier: Azure OpenAI Models and the Arrival of GPT-4
27mThe Cutting Edge of AI: From LLMs to Multimodal Intelligence
01h 33mComputer Vision in Azure
01h 03mIt's All About AI
20mGenAI synthetic data used to fine tune a domain specific model
35mAugmented Retrieval Products with GAI models
48mGetting started with generative AI using Azure OpenAI Service