Scott and Mark learn AI
summary
Live demo builds a MAUI C# app that uses Semantic Kernel to orchestrate cloud and local models, local execution via ONNX/DirectML and a Phi-3 Q4 mini on port 5272, and Azure OpenAI (GPT-4o) for higher-level reasoning. The app counts and summarizes desktop files, logs traces to an OpenTelemetry/Aspire dashboard, requests user consent before moving files, and exposes a surprising off-by-one: it reports 36 then moves 37 text files, showing why deterministic kernel functions matter.
AI-written summary · about
chapters (13)
- 0:00Introduction to desktop chaos
- 1:08AI reactions to desk clutter
- 4:07Developing the organization app
- 7:15Meta prompts and constraints
- 10:05Red teaming and safety filters
- 11:53Limitations of LLMs in counting
- 14:04Implementing deterministic plugins
- 17:24Function calling and reliability
- 23:26Observability with Aspire
- 25:41Orchestrating cloud and local AI
- 28:36Summarizing files with local models
- 33:28Group move and file management
- 36:49Future AI reasoning and wrap-up
related talks
27mQ&A #2
33mHow to enhance your own AI services using Semantic Kernel
53mUnleash the power of AI: Automate tasks with OpenAI and Semantic Kernel
01h 00mBuilding your own AI Agent using Semantic Kernel
01h 02mDevelop AI agents with Semantic Kernel
54mKeynote: AI-Powered App Development
59mOrchestrate your AI with Semantic Kernel
58mKeynote: The past, present, and future of AI for application developers