Advancing Responsible AI: Unveiling the Software Carbon Efficiency Rat...
summary
Green Software Foundation's SCER (aka Shear) defines a reproducible four-part rating: categorization, benchmarking, rating, labeling, and requires CO2e metrics rather than kWh. Two certification levels require publishing readings and usage to GitHub/model hubs; a Green Coding demo shows per-request CO2e and a QR-linked label. Salesforce reports ~74 tCO2e across models, a training-phase 68.8% emissions reduction after right‑sizing and shifting regions, and a 7B-parameter Coen Cogen that outperforms a 16B predecessor — urging integration into hubs like Hugging Face for apples‑to‑apples comparison.
AI-written summary · about
chapters (8)
- 0:00Responsible AI and SCCER
- 3:17SCCER framework for AI
- 8:42SCCER certification programs
- 10:49Green AI and standardization
- 13:45Salesforce’s sustainable AI
- 19:05AI transparency and compliance
- 27:01Next steps and collaboration
- 28:24Q&A on AI regulations
related talks
01h 02mJoint Keynote Address: Scaling AI Sustainably: An Uncharted Territory
57mThe Next Decade of Software Is About Climate - What Is the Role of ML?
47mCodeCarbon: How to Measure the CO2 Impact of M...
44mThe Invisible Monster: The Carbon Footprint of AI- 01h 02mJoint Keynote Address: Scaling AI Sustainably: An Uncharted Territory
41mSustainable Digital
45mGreen AI: Making Machine Learning Environmentally Sustainable
49mDoing for Sustainability, What Open Source Did for Software