Lesson 6-Intelligent Environments PDF

Artificial Intelligence & Human Values Lesson 6: Intelligent Environments Dr. Kebene Wodajo 24 October 2024 Topics What is the interplay between AI and the environment? The environmental history of computing Present day AI’s environmental contexts and consequences Key takeaways 24....

Artificial Intelligence & Human Values Lesson 6: Intelligent Environments Dr. Kebene Wodajo 24 October 2024 Topics What is the interplay between AI and the environment? The environmental history of computing Present day AI’s environmental contexts and consequences Key takeaways 24.10.2024 2 AI systems in context Data E-waste Organis ations Infrastructure: computing power and labour hardware Rules, principles and machine regulations learning models, neural networks, and algorithms Natural Software resource Systems and Platforms Norms and Automation Practices Processes Please identify/write 2-3 environmental concerns related to AI systems and share why these issues pose challenges in pairs or groups of three people. Historical backdrop: Industrial revolution Steam engines & coal-powered industrialisation → Electrical grid→ digital infrastructure, ICT Industrial revolution, CB Media The environmental history of computing Present day AI’s environmental contexts and consequences The interplay and limitations of the problem-solution binary framing Promises Risks/challenges Materiality Immateriality - (“data is the new oil”) Stories/Narratives of Promises “Icebergs are melting – AI knows where and how fast” “AI is helping communities facing climate risks” “Mapping deforestation with AI” “Using AI to recycle more waste” “AI helps predict climate disasters” World Economic Forum, February 2024. Contestation Friends of the Earth, March 7, 2024 Discourses around risks/challenges Resource and energy use “Electricity consumption from data centres, AI and the cryptocurrency sector could double by 2026. After globally consuming an estimated 460 terawatt-hours (TWh) in 2022, data centres’ total electricity consumption could reach more than 1 000 TWh in 2026. This demand is roughly equivalent to the electricity consumption of Japan.” International Energy Agency projection, 2024 Risks/challenges…. Carbon footprint Bloomberg May 15, 2024 Risks/challenges…. Emission from training LLMs “OpenAI’s GPT-3 and Meta’s OPT were estimated to emit more than 500 and 75 metric tons of carbon dioxide, respectively, during training. GPT-3’s vast emissions can be partly explained by the fact that it was trained on older, less efficient hardware. But it is hard to say what the figures are for certain; there is no standardized way to measure carbon dioxide emissions, and these figures are based on external estimates or, in Meta’s case, limited data the company released.” Heikkilä, We’re getting a better idea of AI’s true carbon footprint, MIT Review, Nov. 14, 2022 Risks/challenges…. Emissions associated with AI use Risks/challenges…. Emissions from infrastructure Rest of world, May 2024 Risks/challenges …. E-waste - how does it relate to AI? The pattern of e-waste global movement, sending and receiving countries TheRoundup.org, Apr. 14, 2024 Source: Lewis 2011, The Global E-waste Statistics Partnership, 2018 Risks/challenges…. Critical minerals and raw materials Environmental and social justice at stake Environmental justice achieved "when people are living in socially just and ecologically sustainable communities" characterised by: “decent paying safe jobs; quality schools and recreation; decent housing and adequate health care; democratic decision-making and personal empowerment... where both cultural and biological diversity are respected and highly revered and where distributive justice prevails" (Pellow and Park, p. 4 quoting Bryant (ed.) Environmental Justice: Issues, Policies, and Solutions, 1995) The principle of “do no harm” as a bare minimum. How to go beyond? Key takeaways Question the framing of AI as Pay attention to values like environmental environmental problem / solution justice (links social justice with environmental sustainability) AI systems are material and socio- technical The environmental history of computing is a history of Basic resources: land, water, energy Materiality (land, water, geography) Geography (physical distance, Power (distribution of environmental geographical features of the land, costs is uneven) human settlement) matters for the Mindsets/visions ("data as new oil", sighting and environmental "growth at all cost"; "clean tech") consequences of computing infrastructures Across this history, we notice certain Making materiality of the digital visible dynamics (accumulation of industrial is an ongoing scientific and political practices of different eras) and patterns (pushing out unwanted externalities) task  Coming up Readings for next lesson: Cohn, Carol. 1987. “Slick 'ems, Glick 'ems, Christmas Trees, and Cookie Cutters: Nuclear Language and How We Learned to Pat the Bomb.” Bulletin of the Atomic Scientists 43(5), pp. 17-24. Scott, James. "Nature and Space," [excerpt] Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press, pp. 11-33.