Catalog & Cocktails: The Honest, No-BS Data Podcast What is Data + AI Observability and Why It's Part of Your Competitive Moat with Barr Moses
May 1, 2025
Barr Moses, CEO and Co-Founder of Monte Carlo, dives into the world of data and AI observability as a competitive advantage. She asserts that managing proprietary data is the true moat, not just models. The discussion highlights the importance of reliability in AI products and the necessity of human oversight. They explore the integration of data governance and machine learning practices, the evolving relationship between data and AI, and the vital role semantics play in observability. Networking insights and learning resources for data professionals are also shared.
AI Snips
Chapters
Transcript
Episode notes
Data + AI Observability Essential
- AI observability must include data observability to ensure end-to-end AI application reliability.
- Shipping AI products without observability risks producing unreliable or harmful outputs.
AI Fails in Unexpected Ways
- A chatbot once advised using organic super glue to fix pizza cheese, illustrating AI output failure.
- Another example involved a chatbot selling a car for $1 due to a clever hack.
Four Causes of AI Issues
- Four root causes disrupt reliable AI: model output errors, data source issues, code changes, and system failures.
- Observability must address all these areas, not just model outputs, for true AI reliability.