The impact of AI on the environment

This briefing on the impact of AI on the environment aims to address the knowledge gap surrounding AI's environmental impact and inform stakeholders about ongoing research efforts, to enable better understanding of the technology and reduce fear surrounding implementation.

There are two prevalent AI models – generative AI (used in creative applications, chatbots, and new drug discovery), and discriminative AI (used in medical diagnosis with pattern recognition, decision-making, and speech recognition). This briefing clearly defines the different energy intensities of the different models contrasting large, energy-intensive models used for medical diagnosis with smaller, local models used for tasks like note-taking.

The exact environmental impact of AI is largely unknown as tech companies like Microsoft and Google are yet to publish comprehensive sustainability data for their AI models, thereby making precise impact assessments difficult.  Ongoing work, such as Defra's research on Microsoft Co-Pilot for public sector use, promises to bridge this gap and improve understanding in the future.