Risks to the environment

International Scientific Report on the Safety of Advanced AI

Bengio et al. (2024)

Source

Sub-category

Risk Domain

6Socioeconomic & Environmental

6.6Environmental harm

The development and operation of AI systems causing environmental harm, such as through energy consumption of data centers, or material and carbon footprints associated with AI hardware.

"Growing compute use in general- purpose AI development and deployment has rapidly increased energy usage associated with general- purpose AI. This trend might continue, potentially leading to strongly increasing CO2 emissions."(p. 59)

Entity— Who or what caused the harm

Human

Due to a decision or action made by humans

AI system

Due to a decision or action made by an AI system

Other

Due to some other reason or is ambiguous

Intent— Whether the harm was intentional or accidental

Intentional

Due to an expected outcome from pursuing a goal

Unintentional

Due to an unexpected outcome from pursuing a goal

Other

Without clearly specifying the intentionality

Timing— Whether the risk is pre- or post-deployment

Pre-deployment

Occurring before the AI is deployed

Post-deployment

Occurring after the AI model has been trained and deployed

Other

Without a clearly specified time of occurrence

Supporting Evidence (3)

"The recent rapid growth in demand for computing power (‘compute’) used for AI, and particularly general- purpose AI, development and deployment could make AI a major, and potentially the largest, contributor to data centre electricity consumption in the near future. This is because compute demand is expected to far outpace hardware efficiency improvements."(p. 59)

"Today, data centres, servers and data transmission networks account for between 1% to 1.5% of global electricity demand (578); roughly 2% in the EU, 4% in the US, and close to 3% in China (69, 579, 580). AI likely accounts for well under half of data centre electricity consumption currently, but if the rapid growth of AI's computational requirements continues, AI could become the primary consumer of data centre electricity over the coming years and increase its share of global electricity demand. I"(p. 59)

"The CO2 emissions resulting from AI development and deployment depend on the extent and sources of its energy consumption as well as several factors. The carbon intensity of the energy source is a key variable, with renewable sources like solar power contributing substantially less CO2 emissions throughout their life cycle compared to fossil fuels (581*). AI firms often rely on renewable energy (76, 78), a significant portion of AI training globally still relies on high- carbon sources such as coal or natural gas (581*). Other important factors affecting CO2 emissions include the geographic location of data centres, their efficiency, and the efficiency of the hardware used. As a result, the actual CO2 emissions for a given amount of energy consumed in AI can vary considerably."(p. 60)