Atlas of AI: power, politics, and the planetary costs of artificial intelligence.

By M. Can. 2022

Subfields: Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Safety, Robustness & Interpretability

Citations in the corpus

Evaluating Large Language Models Trained on Code (2021, background): “See, e.g., (Crawford, 2021) 9 Given that code generation (and other forms of AI) might be deployed widely throughout the economy as discussed above, these considerations suggest additional urgency in adopting renewable energy. features that exist as features of other tools of authorship (e.g.,…”
AI adoption and diffusion in public administration: A systematic literature review and future research agenda (2022, unknown): “Crawford (2021, p. 8) argues AI in the current version is far from being artificial or intelligent but depends on a “ set of political and social structures … designed to serve … dominant interests [and] in this sense a registry of power ” .”
The Forgotten Margins of AI Ethics (2022, background): “” [18] When algorithmic systems fail — and they often do — such failures result in concrete negative outcomes such as job loss or exclusion from opportunities.”
GPT-3 and InstructGPT: technological dystopianism, utopianism, and “Contextual” perspectives in AI ethics and industry (2022, background): “Technological dystopianism is the view that technology threatens authentic human life, social values, and societal relationships (Colman [11], p.”
DALL-EVAL: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models (2022, background): “), race is not a biological concept and should be understood as a socially constructed and political concept [22, 10].”
Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model (2022, background): “More transparency is needed regarding the environmental impacts of manufacturing computing equipment given the large quantities of chemicals and minerals required (Stephens and Didden, 2013; Crawford, 2021), the significant quantities of ultra-pure water and energy needed to manufacture it (United N”
A Framework for Deprecating Datasets: Standardizing Documentation, Identification, and Communication (2021, background): “Datasets necessarily represent a worldview, including how the data is collected, labelled and applied, and these particular representations and classifications of the world have social and political values embedded within them [15, 29].”