Opinion piece
AIdeology: on the environmental politics of Artificial Intelligence systems
Paulan Korenhof
Environmental Policy Department
Wageningen University and Research, the Netherlands
The matter of AI
We are in an environmental polycrisis of climate change, extreme weather, biodiversity loss, pollution, and its implications in particular for the world's poorest. This poses challenging questions for environmental governance. To cope with the complexity of the problems, many have turned their attention to the development of Artificial Intelligence (AI) systems to support us in making well-informed decisions on how to counter the current tide. By making implications of environmental change and interventions visible and by offering scenario-testing and predictions, AI-applications are expected to give insight into which sustainability strategies will work, when, and by how much.
While AI, as per its name, thus suggests to offer extra "intelligence" in environmental governance, the denominator "Artificial Intelligence" draws our attention to AI's noetic function. This deflects our attention from the often vast and complex material, technical, and social infrastructures required to make AI work in the first place. AI systems come with significant environmental and social costs—and AI itself may thereby contribute the sustainability challenges it aims to address (Crawford 2021; Van Wynsberghe 2021). The hardware, materials, energy and the humans mining, building, training and maintaining these systems matter for what a system is and does to the world, and in the world. I argue that this materiality of AI systems already embodies an environmental political approach which in turn affects the environmental governance the AI aims to support, and thereby ultimately, the future of our planet.
AIdeology
AI is not just a software program. It is a large complex socio-technical system that consists of different technologies, resources, data sets, sensors, infrastructures, protocols, sciences and human labour, which together are pivotal for machine learning algorithms to be able to exist, function, and be used. In this socio-technical system, human decisions play a key role: an AI system is shaped by how its developers perceive environmental problems, what futures they envision, what role AI should play in this, and what values and norms they hold. By deciding on when, how, and why to develop and implement AI, this set of views is materialised in the system — this is what I call the 'AIdeology' of an AI system. In the realm of AI development for sustainable futures, this AIdeology intersects with views on environmental politics. Currently, a dominant environmental political approach in the European Union is that of ecological modernisation. Ecological modernisation is founded on the premise that economic development and environmental sustainability can go hand in hand. Ecological modernists argue that with the use of technological innovations like AI-systems for improved efficiency in resource and energy consumption, and market-based approaches like the Europe's Emission Trading System, environmental challenges can be addressed without hampering economic growth. By active collaboration between governments, businesses and society, and by integrating ecological concerns into economic processes, environmental protection can become an opportunity for modernisation instead of an obstacle to progress. This is in particular reflected in the Europe's 'Twin transition' strategy: the transition towards in increasingly digitised as well as sustainable society, where digital innovation plays a key role in the pursuit of environmental sustainability.
The environmental politics of Europe's Twin transition shape the AIdeology of AI systems developed under this paradigm and the potential futures they can pave the way for. With digital innovations as key strategy point, we see a push — and commitment to large financial investments — to advance AI development in Europe. This places a focus on AI development for the sake of AI and triggers wide deployment of the technology. Moreover, in the context of advancing economic growth and innovation, the corresponding AIdeologies are often focused on AI development for efficiency and optimisation. With this, the future visions AI can offer remain strongly embedded within the dominant environmental political approach: efficiency and optimisation are focused on improving an existing system, and are unable to look outside of the current status quo for alternatives.
While AI in its prime hype is easily thought of in terms of increase, acceleration, a must for progress and profit, and an answer and improvement for everything, it is important to ask: can we imagine a different AIdeology? As a brief thought exercise, let us imagine an AIdeology based on the political principles of degrowth. Contrary to ecological modernisation, the degrowth movement argues that the pursuit of continuous economic growth cannot be detached from an overconsumption and resource depletion that surpasses planetary boundaries. Degrowth is therefore focused on the reduction of overconsumption and unnecessary production to promote environmental sustainability, social equity, and well-being.
Applied to AI development, 'degrowth' AIdeology would need to take the form of a system developed and used in moderation to prevent overproduction as well as overconsumption, and its environmental and social impact should be proportional to its goal (e.g., no generative AI use for tacocat images). AI development would only justified if it is the best fit to solve the problem at hand, and if it makes a fundamental contribution to sustainability. Ideally, this contribution should transcend the increasing of efficiency and optimisation. Moreover, systems should be decentralised, use minimal data and energy, and successful software should be shared worldwide to prevent unnecessary development and training. Realising such a degrowth AIdeology would require a significant paradigm shift in contemporary thinking about AI development and use from governments, industry, and consumers alike.
Can we change the system?
“The master’s tools will never dismantle the master’s house” (Lorde 1984). This famous quote by Audre Lorde refers to the limitations of using oppressive systems to achieve liberation. It also carries relevance for our case: the use of unsustainable AI systems to achieve sustainability. Contemporary AI systems, with their focus on efficiency, optimisation, and ‘green growth’ seem designed to maintain and push the most out of the status quo. These systems thereby reinforce the current focus on capital growth and its unequal power distribution in societies on a national and global level. On a practical level, these AI systems seem to function as delay for making hard and difficult choices: how much can we still squeeze out of the current status quo by ‘knowing’ more and increasing efficiency before we truly need to alter our course?
We know some hard decisions need to be made by doing: by bringing about large societal changes in resource use, consumption patterns, and food production. Real transformation calls for creative and out-of-the-box approaches that rethink directions for the future and the role technology can play in this. Sustainability is not achieved by unsustainable technologies. We therefore need to rethink AI beyond the dominant ecological modernist AIdeology towards an AIdeology aimed at transcending the current status quo, an AIdeology of proportional and moderate development and consumption where AI is only developed when meaningful and strictly necessary for the well-being of the planet and its inhabitants.
In the end, AI will not save our future, only we can.
References
Crawford, K. (2021). The atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
Lorde, A. (1984). Sister outsider: Essays and speeches. Triangle Classics. Van Wynsberghe, A. (2021). Sustainable AI: AI for sustainability and the sustainability of AI. AI and Ethics, 1(3), 213-218.