Training and running AI systems takes electricity, and the data centres that host them also use water for cooling and rely on chips and hardware with their own manufacturing footprint. Whether that matters depends on how large the demand grows, how it is powered, and where the costs land, which is why the ethical questions include transparency and who bears local impacts as well as the totals.
The lessons begin with a short BBC explainer on how AI uses water, followed by researcher Sasha Luccioni on the energy used by AI models. A longer talk then looks at AI and climate change from both sides, the costs of the technology and the ways it may help, and the final lesson digs into the research behind energy ratings for AI models.
Estimates of AI's environmental footprint vary widely and change quickly, and experts disagree about how to measure and compare them, so treat the figures in these videos as a snapshot. Watch them in order, then continue on to Surveillance, Military & Dual-Use AI.
BBC World Service explains why AI uses water, directly for cooling data centres and indirectly through electricity and chip production, and why the true figures are hard to pin down.
Video 6minProof News speaks with Hugging Face's AI & Climate Lead about her research on how much energy AI models use to answer queries, and whether AI will solve climate change or make it worse.
Video 25minSasha Luccioni gives a balanced talk on the carbon footprint of AI and on projects that use AI in the fight against the climate crisis.
Video 1h 9minA podcast conversation on research comparing the energy use of general-purpose and task-specific models, and on a proposed rating system to help people choose more efficient models. A long, advanced, optional watch.
Video 50min