In Prompts We Trust
summary
Live prompt demos use GPT‑4 and 3.5, Azure OpenAI and Azure AI Studio Prompt Flow with LangChain‑style chaining to show concrete gains: few‑shot and role prompts improve style, Chain‑of‑Thought and Tree‑of‑Thought fix arithmetic and reasoning errors, and tweaking temperature/top‑p/penalties changes creativity versus reliability. Retrieval‑augmented generation pulls indexed PDFs from blob storage and returns sources to reduce hallucinations. The presenter builds a RAG-backed TikTok trend assistant and a LinkedIn recommendation generator live — both work but require iterative prompting and grounding to be dependable.
AI-written summary · about
chapters (22)
- 0:00Introduction and personal journey
- 2:49Understanding large language models
- 5:17The trust gap in AI
- 8:49A trust model for generative AI
- 14:59Iterative prompt engineering
- 16:42Considerations for trusting outputs
- 17:01Factors influencing prompt effectiveness
- 18:05Common prompting techniques
- 26:24Combining prompting techniques
- 26:53Styling, tone, and formatting
- 28:31Advanced reasoning techniques
- 36:38System message demonstration
- 37:26Adjusting inference parameters
- 39:30Retrieval Augmented Generation (RAG)
- 45:59RAG flow implementation
- 47:38Exploring prompt patterns
- 49:12Interactive prompt refinement
- 49:33Homework and concluding thoughts
- 49:44Session wrap-up
- 52:36Audience Q&A prep
- 53:38Final closing
- 56:06Post-event follow-up
related talks
01h 04mIn Prompts We Trust
44mThe Science of Effective Prompting
47mThe Art of Prompt Engineering: Best Practices for Effective AI Interactions
16mPrompt Engineering for Testers: Types and Applications
46mPractical End-to-End AI Development using Prompty and AI Studio
29mSupercharge Your Productivity: Mastering Prompt Engineering
01h 03mKeynote: Red, Blue, and Purple AI
43mBuild and Maintain your Company Copilot with Azure ML and GPT-4 Q&A