The era of the AI Copilot
summary
Microsoft and OpenAI deliver an end-to-end platform on Azure plus Windows for building copilots, offering hosted models, fine-tuning (3.5 live, GPT‑4 soon), and open-source options with easy deploy from the Azure model catalog. They describe plugins as an open standard, orchestration via Semantic Kernel/LangChain/prompt flow, grounding with retrieval-augmented generation, and safety tools including cryptographic media provenance. Cost drops cited are 70 percent price reduction then 90 percent in a year. Example: a working social media copilot uses Whisper, Dolly 2.0, Bing API, DALL·E and a LinkedIn plugin to post automatically.
AI-written summary · about
chapters (24)
- 0:00Intro
- 6:55Chat GPT
- 15:20Open AI
- 15:50Copilots
- 18:20Bill Gates Quote
- 18:45Platform vs Copilot
- 19:45Reusable Foundation Models
- 20:10Plugins
- 21:27Anatomy of a Copilot
- 22:18Build a great product
- 23:15The model is not your product
- 24:05The Copilot Stack
- 25:43User Interface
- 27:03Orchestration
- 28:24Langchain
- 29:29Prompts
- 31:45Grounding
- 32:58Plugin execution
- 33:38Foundation models
- 35:26Copilot Sack
- 36:28Copilot Demo
- 39:20AI Safety
- 40:22Media Provenance
- 40:58AI Copilots
related talks
02h 24mBuild opening keynote
56mBuild your own CoPilot
36mWhat's new with Copilot Studio
23mEnabling ISVs to transform industries in the age of copilots
36mCustom copilot Teams Integration in Government applications
45mBuilding a copilot: Azure AI Studio or Copilot Studio
12mA conversation with Charles Lamanna on building custom copilots with Microsoft
47mSafeguard your copilot with Azure AI