Ecosystem and Environment

Evaluating the Social Impact of Generative AI Systems in Systems and Society

Solaiman et al. (2023)

Source DOI

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.

"Impacts at a high-level, from the AI ecosystem to the Earth itself, are necessarily broad but can be broken down into components for evaluation."(p. 16)

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 (2)

"Widening Resource Gaps: As described in section Financial Costs, the high financial and resource costs necessarily excludes groups who do not have the resources to train, evaluate, or host models. The infrastructure needed to contribute to generative AI research and development leads to widening gaps which are notable among sectors, such as between industry and academia [145], or among global powers and countries [19]. Externalities broadly refer to the unanticipated effects of economic activities on the social environment. Access and Benefit Distribution: Ability to contribute to and benefit from a system depends on ability to engage with a system, which in turn depends on the openness of the system, the system application, and system interfaces. Level of openness and access grapples with tensions of misuse and risk. Increasing trends toward system closedness [227] is shifting access distribution. Geographic and Regional Activity Concentration: In the field of AI as a whole, top AI research institutions from 1990-2014 have concentrated in the U.S. [164]. More recent data highlights the U.S., EU, and China"(p. 17)

"Environmental Impacts: In addition to the 4.1.6 Environmental Costs and Carbon Emissions from a system itself, evaluating impact on the Earth can follow popular frameworks and analyses."(p. 18)