Garbage In, Garbage Out: Elevating AI with High-Quality Data
summary
Three real failures ground the claim: Google Flu Trends doubles case estimates for many weeks and collapses by 2015, Air Canada’s chatbot gives wrongful bereavement guidance that triggers a refund and reputational damage, and Cambridge Analytica–style misuse shows perfect models still fail ethically. The solution stack uses Confluent streams, schema-backed data contracts, stream catalog and lineage, plus Apache Flink + Bedrock for real-time RAG-style inference; a demo joins loyalty and program streams as context and shows responses fail without real-time, governed data but improve when context is added.
AI-written summary · about
chapters (13)
- 0:00High-quality data in AI
- 1:29AI and data’s pillars
- 2:32AI models need trustworthy data
- 4:21Google Flu Trends failure
- 5:49Air Canada chatbot disaster
- 7:36Cambridge Analytica breach
- 9:19Poor data impact on decisions
- 9:41RAG for generative AI
- 11:29AI application building blocks
- 12:11Data contracts for quality
- 13:40Stream catalog for governance
- 14:47Live AI inference demo
- 19:30Summary for high-quality AI
related talks
58mIt AI-n't What You Think!
26mArtificial Intelligence? - More like Artificial Stupidity!
26mThe Hidden Foundation: How Effective Data Management Determines AI System Success
40mData Strategy for Generative AI Applications — Why Your Data is the Differentiator
33mDataEngBytes 2023 - MEL-T1-02 - Kendra Vant
44mStreamlining Your Approach to Data
35mHigh-performance data foundation for AI success
26mSafeguarding AI: How to Manage AI and its Potential Risks