The Cutting Edge of AI and IoT
summary
Claims that Ed impul builds edge AI pipelines deployable to Linux MCU and NPU devices, shows integration with generative models like OpenAI Whisper 11 Labs and Nvidia Omniverse for synthetic labeled data, and cites a cost example that running ChatGPT on 512×512 1 FPS costs about $172 per day. Demonstrates an on‑device anomaly detector that sends only anomalous frames to ChatGPT over LoRaWAN and reports it working in the recorded demo.
AI-written summary · about
chapters (11)
- 0:00AI and IoT integration
- 1:32Traditional vs AI for sensors
- 3:37Edge AI pipelines and deployments
- 6:45Five years of Edge AI with Ed Impul
- 7:48Synthetic data for Edge AI
- 9:34Digital twins and 3D data
- 11:20LLMs for data quality
- 13:48Foundational and domain models
- 15:13Smart anomaly detection demo
- 16:37AI, IoT, and cloud synergy
- 17:00Closing and resources
related talks
14mEdge AI: Real Solutions for Real World Problems
33mAI on the Edge – Unlocking New Opportunities in IoT
15mSimplifying & Scaling Edge AI
13mHow Edge AI Will Save The World
24mWhat does AI mean for IoT development?
05mEdge AI: Because the Cloud is Too Slow...
14mTwo Keys to AI's Future
44mArtificial Intelligence in Small Embedded Systems