The Data Stack Show 197: Deep Dive: How to Build AI Features and Why it is So Dang Hard with Barry McCardle of Hex
Jul 10, 2024
AI expert Barry McCardle discusses challenges in building AI features, benchmarking AI models, and AI integration in products. Topics include AI development struggles, determinstic template selection, and open source database success. The podcast also covers market dynamics, enterprise adoption, and the fun side of product launch videos.
AI Snips
Chapters
Transcript
Episode notes
AI Accountability Analogy
- Barry McCardle uses self-driving cars as an analogy for AI's role in decision-making.
- He questions the accountability of AI when outcomes are negative, unlike human analysts.
Self-Service and AI
- Self-service analytics, with or without AI, presents similar challenges in data misuse.
- AI's accuracy depends heavily on context, highlighting the need for tools like semantic layers.
AI in Production
- Building production AI systems is far more complex than simple demos.
- Hex leverages retrieval augmented generation (RAG) and focuses on UI/UX for effective AI integration.