Augmented Intelligence - A Marriage between Machine and Human
summary
Argues that augmented intelligence is a necessary intermediate step toward full automation, using explainability and human feedback to build trust. Demonstrates a production case with 12,000 unlabeled images and 520 labeled images, training a deep convolutional autoencoder to surface reconstruction-error heatmaps as anomaly cues, then concatenating latent features plus error maps for a classifier. Uses SHAP explanations and a web UI where workers inspect highlights, draw bounding boxes, and feed corrected labels back to the model for continuous learning.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:50Agenda
- 2:30Artificial Intelligence
- 6:22Do you prefer an additional human
- 8:40Do you trust your system
- 9:12Limitations of AI
- 10:13Lack of understanding
- 10:45Augmented Intelligence
- 11:51Augmented Intelligence Example
- 13:46Augmented Intelligence Collaboration
- 15:35Why Augmented Intelligence
- 15:55Building Trust
- 17:34Using Shep
- 19:32Industry 40 Hackathon
- 20:17Visual Inspection
- 22:22Data
- 27:48Summary
- 29:12Questions
related talks
25mAugmented Intelligence — The Future of GenAI?
13mHuman + Machine
08mAgents and Robots: The Human Effort Putting AI Into Production
12mAugmenting Human Capabilities to New Dimensions
24mMachine Intelligence gets real
36mFrom Automation to Augmentation: AI Coaches That Amplify Expertise
39mThe Search for Transparency and Accountability in the Age of...
37mInteractive Explainable AI