Volts Wrapping our heads around AI and climate
72 snips
Jun 7, 2024 
Alp Kucukelbir discusses the transformative power of AI in climate change, optimizing industrial processes with AI and machine learning, concerns about AI manipulation and disinformation, AI's role in decarbonization policies, advancements in data gathering technologies, and making industries more efficient in developing nations.
AI Snips
Chapters
Transcript
Episode notes
Real-Time Steel Recipe Optimization
- Faro measures scrap steel chemistry every ten minutes and uses ML to adapt the production recipe in real time.
- This lets plants raise recycled content from ~20% toward 50% while keeping quality and saving emissions.
AI Should Know When To Defer To Humans
- AI systems must quantify uncertainty and surface when human intervention is required in industrial settings.
- Good systems will say 'call someone' when encountering unfamiliar, risky batches.
Industrial Data Is Small And Bespoke
- Most individual plants have small, idiosyncratic datasets; models must adapt to local quirks rather than rely on huge pooled datasets.
- Scalable impact comes from packaged tools usable by domain experts, not requiring scarce hybrid experts.