A Comprehensive Examination of AI's Environmental Impact: Challenges and Opportunities
Evaluating the pessimistic tech-dystopian discourse that often portrays Artificial Intelligence (AI) as systems of “extractive planetary computation”, this article ventures to offer an unbiased overview of the intricacies between AI systems and their known planetary consequences. While acknowledging AI's tangible exploitative potential, this article will also explore its advanced technological capabilities in addressing pressing environmental challenges. Ultimately, this article seeks to inspire further discussions surrounding the planetary implications of AI, as well as their current limitations, to foster informed, critical public scrutiny regarding the environmentally responsible development of AI.
1. Partial Accounts of AI Under Information Capitalism
A critical examination of the post-industrial economic framework within which the historical progression of AI unfolds, as described by philosophers such as Michael Hardt and Antonio Negri under the notion of "Information Capitalism," elucidates several of its disconcerting socio-political consequences. Mirroring the dualistic metaphysical perspective of Cartesian Dualism, the prevailing portrayal of AI, as subject to criticism, has been artfully manipulated to create a distance from the material underpinnings required for its very genesis. This orchestrated disconnection from the earthly foundations of AI's production has fostered a popular narrative that seemingly divorces AI as a physical technological entity from its reliance on substantial planetary resources and intricate networks of extraction. It has cultivated an enchanting optimism that has emanated from AI's seemingly linear and unobstructed historical development. This shift in public perception has effectively redirected popular attention toward only utopian visions of AI's computational capabilities, envisioning its transformative potential across a multitude of domains such as healthcare, commerce, and military applications.
2. Reassessing the Environmental Consequences of AI
Frequently overlooked and intentionally disregarded is the fact that acquiring computational power of such quasi-theological qualities relies on the utilisation of extensive material resources. Consequently, an intricate system of resource extraction spanning temporal and geographical dimensions has been born out of AI systems globally. To wit, anthropologist Steven Gonzalez Monserrate’s case study “The Cloud Is Material: On the Environmental Impacts of Computation and Data Storage,” elucidates the multifaceted ecological consequences stemming from data storage. When considering an annual consumption of 200 terawatt hours (TWh) by data centres collectively, it exceeds the energy consumption of some nation-states. The combined yearly electricity consumption of merely five technology giants - Amazon, Google, Microsoft, Facebook, and Apple - amounts to approximately the same as that of New Zealand, totalling more than 45 terawatt-hours. As the adoption of cloud services continues to expand, this number is anticipated to rise further. In addition to the utilisation of energy that directly contributes to supporting computational processes, the standardised “fail-safes” approach adopted by data centres to ensure the accessibility of services without interruption means data centres are designed with a hyper-redundant architecture. This means if one component malfunctions, another can seamlessly step in. Remarkably, it was explained by Monserrate that a mere 6 to 12 per cent of the total energy consumption is directed toward active computational processes, while the remaining resources are allocated to cooling and upholding intricate chains of fail-safe mechanisms. In business, the relationship between AI and its established environmental impacts could be effectively exemplified through the rise in commercial leases pursued by data centre operators. This surge in demand for data centres inevitably leads to increased energy costs, shedding light on the energy-intensive character of data centre operations.
3. The Environmental Promises of AI Systems
Having established that externalities generated by AI could have a profound environmental impact, subsequent sections of this article will assess AI's potential for finding solutions to environmental problems. Although AI systems rely on tech infrastructures that demand excessive resources optimism towards the environmental promises of such systems nevertheless exists. Such optimism could be captured by market survey results as well as the comments of real estate industry ESG experts. According to the results of a survey conducted by JLL and IDC, the predominant aim for 505 data centre managers in the Asia Pacific region over the next two years is to improve their sustainability and social responsibility initiatives. This could be practically achieved by effectively deploying technologies to address some of the problems outlined in section 2.2. According to Kamya Miglani, who serves as the Head of ESG Research for Asia Pacific at Jones Lang LaSalle Incorporated (JLL), advancements in technology, encompassing both hardware and software, e.g., software-driven monitoring and control systems, are currently enhancing the energy efficiency of data centres.
4. The Use of AI in Environmental Modelling
Besides optimism towards resolving sustainability concerns surrounding data centres, AI systems could be employed to positively model and monitor environmental systems, allowing smarter decision-making regarding practices such as pollution treatment. The United Nations Environment Program's (UNEP) World Environment Situation Room (WESR), launched in 2022, harnesses AI to analyze intricate environmental datasets. It provides real-time analysis and future predictions on various factors, including CO2 levels, glacier mass changes, and sea-level rise, using credible and independent data to drive informed decisions. One of the prominent AI techniques used in environmental modelling is Artificial Neural Networks (ANNs). ANNs, which are found to be the mainstream AI technologies, employ an abstraction of the human brain's information processing mechanisms. In simple terms, it is a framework consisting of multiple processing units, often referred to as neurons or nodes, which work collectively to process and analyze data. Their exceptional capability, compared to traditional techniques such as OLS regression, lies in their ability to capture the complex and nonlinear relationships between dependent and independent variables by leveraging historical data. The effective utilization of AI techniques, including ANNs, in environmental modelling offers several advantages, including the optimization of resource allocation and the facilitation of evidence-based environmental policy development.
5. Final Remarks
In conclusion, this article has endeavoured to offer an impartial exploration of the intricate interplay between AI systems and their environmental impacts. While recognizing the valid environmental concerns associated with AI, it is equally important to study its capacity for positive change in promoting environmental sustainability. By embracing a more balanced and rational discourse, the full transformative potential of AI could then be harnessed to effectively address the pressing environmental challenges of today.