Open Source AI and Granite
summary
Granite positions itself as an auditable open model family with Granite time series and vision/guard rails and a promise that IBM will defend attribution in court; demos run in Jupyter notebooks, use local models and runs locally on CPU (no cloud GPU), projectings include a diabetes insulin-usage forecast (accurate on validation) and Nagios disk forecasts, an abstract classifier surfaces 8 of 12 salesy CFPs from ~120, and there are ~1.4M Hugging Face models complicating selection.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:05Open source AI
- 3:54Model Openness Framework
- 5:17Fine-tuning and Instruct Lab
- 6:20RAG vs fine-tuning
- 7:00Agent frameworks and IBM B
- 8:25Granite open source models
- 10:53Granite licensing
- 11:14Granite use cases
- 12:38AI for medical education
- 13:40STEM stories for kids
- 14:23Time series demo
- 15:48Insulin usage projection
- 19:03Infrastructure monitoring
- 20:30Disk usage projection
- 21:58Q&A
related talks
15mArtificial Intelligence Needs Community Intelligence
16mKeynote: Open Source for AI and AI for Open Source
32mHumans In the Loop: Collaborating in the Age of AI
38mBuilding AI responsibly with a GRC strategy
29mOpen Source Sovereign AI: Building Enterprise GenAI Platforms for Control & Comp...
22mThe Imperative for Open Source AI - The Path Forward to Take Down the AI Oligarchs
38mScaling Specialist Knowledge with AI
23mAI/ML + open source observability