Sustainable AI: AI Footprint and What To Do About it

Explore the hidden side of ai by examining rapid adoption, cloud services, data centers, and the energy, water, and hardware footprints and resources depletion behind carbon emissions and sustainability decisions.

Define artificial intelligence as systems that think and act like humans, and explain how machine learning, deep learning, generative AI, LLMs generate text from prompts and their energy use.

Explore how AI drives energy and water use in data centers, the embodied carbon in hardware, and data transfer demands, and learn targeted steps toward sustainable AI.

Examine the environmental footprint of training GPT three, with 175 billion parameters, 1000 GPUs, and 34 days of compute, consuming 1287 MWh and emitting 552 tons of carbon.

Generative AI consumes energy in both training large language models and ongoing inference in data centers, with emissions rising as model parameters grow and queries scale.

Examine the paradox of AI's growing energy demands and carbon footprint, while exploring opportunities to optimize energy use and harness AI to tackle climate challenges.

Define sustainable AI as minimizing environmental footprint while maximizing impact for people. Learn environmental design and ethical governance across AI life cycle—from mining and manufacturing to training, deployment, and end-of-life.

Apply the five c's framework to AI, focusing on clean energy and efficient operations, inclusive development, culture, care for workers and communities, and corporate governance.

Explore the principles of sustainable AI, from right sizing and frugal AI to on-device AI, data minimization, and lifecycle thinking, with techniques such as distillation, quantization, and pruning.

Demonstrates running an LLM locally with LM Studio on a personal computer, loading open-source models from Huggingface. Explores sustainability across development to deployment, including environmental, ethical, and societal considerations.

Build workforce capacity for sustainable ai through cross-disciplinary training, and design environmentally friendly systems that balance transparency, fairness, accessibility, digital inclusion, and prudent ai use to minimize energy impact.

Organizations can advance sustainable AI by using smaller models, minimizing prompts and data, monitoring usage, and aligning governance and leadership with a practical roadmap.

Explore how standards like ISO 42,001 and the UI Act shape sustainable AI governance, including environmental impact assessments and transparency requirements across the AI life cycle.

Showcases AI’s positive societal impact, from improved healthcare diagnostics in mammography and faster clinical trials to precision disease insights, climate tracking, and materials breakthroughs reducing lithium in batteries.

Review Stanford AI index 2025 and studies on energy, water, and carbon footprint of LM inference, plus review of demand for language models; learn rightsizing and optimizing to reduce footprint.