Artificial Intelligence Is an Extractive Technology

In 2023 I referred to Kate Crawford's book, Atlas of AI, and it still remains one of the most important texts in this discussion. Crawford draws analogies between artificial intelligence and the mining industry, focusing on both the environmental and the socio-political impacts.

When Crawford writes about the environmental impact of AI, she is talking about recommender systems (Netflix, Spotify), predictive algorithms (Facebook, YouTube), and broader implications of data classification and analytics — not just generative AI.

Crawford points out key ideas:

  1. Data centres consume phenomenal amounts of electricity and rely on water for cooling. Companies like Microsoft have built data centres in cold climate countries and even attempted to build them underwater.
  2. The hardware used to run AI depends on extractive mining processes. Lithium, silicates, rare minerals and other elements have incredibly high physical costs of extraction.
  3. Where mining extracts oil, artificial intelligence extracts data. The processing of this raw material consumes the most energy.

Training-Time Impact

One of the ways in which GPT-3.5 changed the trajectory of generative AI research was by throwing more data and more computational power at the system than any researchers had before. Just as burning more coal produces more pollution, burning through more raw materials of AI training data produces more adverse effects.

A 2024 article estimated training GPT-3 burned 1,287 MWh of electricity and emitted about 552 tonnes of CO₂ — roughly the annual electricity use of 120 average US homes. The impact of training GPT-4 and newer models remains unknown due to a complete lack of transparency.

Cumulatively, data centres consume approximately 4% of all US energy, and are using so much water that they are quite literally sucking entire regions dry.

Inference-Time Impact

Inference time means "the point at which the user gets the AI to do the thing." Recently, there has been increased interest in how much energy these systems use when we use them.

Research by Sasha Luccioni, Yacine Jernite, and Emma Strubell (Power Hungry Processing) compared energy consumption across different AI task types. Their key finding:

Generative artificial intelligence systems are incredibly resource intensive compared to other technologies which may be better suited to the job.

Multi-purpose generative AI, especially image generation, consumes vastly more energy per task than task-specific models designed for the same job.

Implications of AI Energy Use in Education

An individual's use of ChatGPT may consume less energy than streaming Netflix for one hour, a Zoom call with 10 people, or driving a car for a few miles. But this "individual use" narrative obscures the bigger problem: a lack of transparency means we have no idea what energy consumption looks like at scale in education.

We need to ask: Do we need better image generation models for lesson resources, or more reliable sources of open access images? Do we need ChatGPT to analyse parent surveys, or is there an older, more efficient technology that would do the same? Should I use Gemini to write a lesson plan, or sit down with a colleague over a coffee and do it myself?

Teaching Environmental Issues

  • English: Analyse persuasive texts on digital sustainability, asking "Is this tech company truly 'green'?" or "How do tech advertisements use environmental language to mislead?"
  • Science: What is the carbon footprint of cloud computing? How do server farms affect local ecosystems?
  • Digital Technologies: How energy-efficient is the code we write? What happens to old hardware when we upgrade?
  • Geography: Where do raw materials for smartphones come from? How does digital infrastructure affect land use?
  • Civics: Who bears the environmental burden of digital consumption? Should governments regulate e-waste exports?
  • Mathematics: How much CO₂ is generated by a Google search? What do energy use graphs of tech companies reveal?
  • Visual Arts: How can digital art raise awareness of e-waste or tech pollution?