The Colorado River is a living ecosystem of self-sustaining waterways stretching from tens of thousands of feet high in the Rocky Mountains traveling through the Sonoran and Mojave deserts and flowing into Mexico. These beautiful and unruly waters comprise rivers, tributaries, confluences, and bends flowing in from snow melt, groundwater, and aquifers. Over the past century, the US empire has violently undermined Indigenous lifeways and water sovereignty on the Colorado River through calculated legal manipulations and environmental degradation.1 The prevailing “Law of the River,” which mandates that Native American tribes collectively own 30% of the Colorado River’s waters,2 has amounted to a form of colonial datafication predicated on overallocated and depleted waterways and the repudiation and disavowal of Indigenous governing structures.3

The Colorado River’s widely publicized “megadrought” crisis provides the most recent excuse for the expansion of colonial datafication or artificial intelligence (AI) of its waterways. Since the year 2000, an 86% depletion of snowmelt runoff has contributed to declining water levels and cracked and dry lands. White anxieties over this water shortage have inflected a hydropolitics of exclusion during a time of crisis.4 As part of these political currents, venture capitalists and tech companies are responding with promises of new innovations in AI and machine learning (ML) to solve impending drought and scarcity.5 Undergirding their AI and cloud-based innovations is a colonial water and land allotment grid, the legal-quantitative pretext for the ongoing extraction and overallocation of the Colorado River.6

The idea that the Colorado River is a singular, quantified water system derives from colonial history. Over the past two centuries, white settlers have terraformed the Colorado River Basin guided by violent beliefs of manifest destiny–that westward expansion was justified and inevitable–and terra nullius–meaning “territory without a master.” What began as religious drivers to conquer and expand transformed, over time, into the scientific prowess of the US empire. Where religious settlers used biblical imagery to justify dispossession of lands and waters, modern settlers mobilized the capitalist values of efficiency and optimization to occupy land and commodify and trade water, a mastery through numbers that codified extant ideologies of racial and religious superiority. 

In the nineteenth century the USGS violently occupied and datafied land, water, and Indigenous communities, defining and occupying the Colorado River Basin in the process. This historic big data project catalyzed several key laws partitioning Native land for settler ownership: the Homestead Act, the Railway Act, and the Dawes Act (also known as the General Allotment Act).7 These laws dispossessed Native land and water and broke up collective property systems that predated settler systems. Corresponding laws – the “Law of the River” galvanized in the 1922 Colorado River Compact – transformed water from an abundant ecosystem to scarcity-driven units of acre/feet trade.8

The colonial water grid primed the Colorado River for AI expansion and for the continued use of AI as a water commodification system. The circularity extends to the fact that AI systems deplete water resources–they require enormous amounts of water to build and maintain servers and data centers.9

In broad definition, AI is a technical and political program that consolidates resources and decision-making power within the hands of a few entities through computational algorithms.10 These entities hold power and license over massive databases, material infrastructures, and digital technologies. Visions of Manifest Destiny are taken for granted and frequently invoked by AI figureheads like Alex Karp, Jeff Bezos,11 and Elon Musk12 and by companies like IBM, Google, and Microsoft.13 AI solutions are not just technological solutions dreamed up by idealistic engineers. They are a land and water grab by new contenders for settler power. These proposals intentionally blur distinctions between water innovation and water law. Artificial intelligence reifies the settler colonial water and land allotment grid.14

While computational and predictive resource governance often takes up the guise of numerical objectivity, it has historically been anything but, encoding instead the social ideologies of colonial capitalist modernity:15

  1. AI and ML water programs are neither neutral nor objective. These are tech innovation nodes that consolidate money, power, and decision making authority over water resources.
  2. AI is neither autonomous nor sentient; AI and ML systems are the Frankenstein’s monster of water data management, avatars of a continual manifest destiny approach to transforming water into a scarce resource with discrete, calculable use-value.
  3. AI cannot “solve” any water crisis because (it is neither autonomous nor sentient and) AI developers’ leading definition of water crisis is a crisis of inefficiency, a framework that reifies extractive and deleterious water governance on the Colorado River. 
  4. In AI/ML R&D, strategic partnerships between tech companies, water governors, and research institutes are blurring distinctions between water innovation and water law in order to consolidate wealth and power. 

In practice, AI/ML is foremost a mega tech R&D funding category primarily organized around profit-seeking without public accountability; governments, venture capitalists, and scientific houses are pouring money into AI and ML climate and environmental management innovations, including janus-faced “AI for good” programs. By citing the heading of AI/ML, water research institutes and university laboratories can rapidly acquire large sums of money. 

In these innovation hubs, companies such as Intel and Microsoft are working with local water governors such as the Central Water Conservation District and the Nevada Water Authority to develop and employ AI water solutions without public input and consent nor sustained conversations with Native governments. Their suite of innovations include SMART Water Grids, AI to Optimize Utility Processes, and water financialization tools, primarily Blockchain, for trading water futures.16 Hubristic generative AI proposals that would alter local water allocation and distribution contexts are projected to reach 1.5 trillion USD in revenues by 2033. 

Reinforcing these innovations is the myth that the Colorado River water crisis is a crisis of inefficiency. A 2019 Water Foundry report citing the Secretary of Interior as “watermaster”17 describes the Colorado River Basin as “a strategic testbed” that is “used to determine the feasibility of emerging and novel digital technological solutions for the water sector.” This view of the Colorado River Basin as a datafied laboratory reinforces the spatial arrangements of the white settler state, that the waterways are a singular, datafied system feeding innovations for profitable water trade. Within this virtual system, AI/ML proponents acknowledge the problems of scarcity, overallocation, and poisoned water. However, they state the cause of these problems to be “inefficiency” rather than direct outcomes of the stratified histories of colonial occupation, and thereby avert engaging with what actually caused the water crisis.  

The myth of inefficiency is thereby both a catalyst for the expansion of AI and a distraction from addressing the colonial water laws driving the devastation of overallocation and overdevelopment of the Colorado River. Colonial data systems and predictive technologies are predicated on capitalist drivers of profit and growth; for this reason, the pursuit of “efficiency” via these systems can only exacerbate the devastation. There is danger of capture by the “adaptation regime,” a “historically specific configuration of power that governs the landscape of the possible intervention in the face of climate change.”18

There is further epistemic confusion about the teleological promises of AI–water system designers commonly define it as fully sentient, able, predictive, and intelligent.19 AI is neither sentient nor autonomous. AI often refers to patchworked data analysis procedures and computing processes used to make predictive interpretations. Similarly, machine learning or ML refers to compilations of algorithmic models such as clustering algorithms, CNNs, and LLMs that analysts use to shape data patterns. In reality, the data systems and procedures comprising AI and ML are often clumsy guesswork misaligned with human judgements of reality.20 Akin to the patchworked corpse of Frankenstein’s monster, the technical terms “AI” and “ML” are referents for a jumble of datasets, algorithms, and computing procedures sewn together for specific and particular ends. 

Furthermore, cloud computing cannot supersede land relations. All AI and ML systems abide by extant economic assumptions and data processes that reinforce colonial and capitalist relationships to land and water.21 Even innocuous uses of AI and ML in water management (such as salinity level mapping) assume political frameworks including legal codes, data ownership, and economic frameworks such as optimization, that are often unexamined in their calculated presentation. Additionally, all AI and ML systems depend on human labor and natural resources. Data centers alone consume large amounts of water and electricity and often alter existing land and water policies to justify these scales of resource consumption.22

A danger in AI/ML water systems is in how easily experimental design can alter water allocation and pricing policy. For example, researchers at Penn State recently designed a sorting algorithm to establish “archetypes” of water consumers.23 While seemingly innocuous, this system design places the burden of conservation on individuals instead of demanding accountability from the colonial and capitalist systems that exacerbate drought and scarcity.24 It is an example of an algorithmic design mechanism that can have lasting policy consequences. Across localities, AI is being developed in ways that confuse data concepts and technological PR with political and environmental truths. Artificial intelligence in the water domain is a convoluted technical and political program where unexamined data innovations and data acquisitions can quickly augment decisions and policies about living resources. The premise of AI innovation in water is that land and water are inefficient, necessitating an endless and devastating pursuit of efficiency and predictive accuracy. By chasing an abstract frontier, AI powers are designing an artificial future for the Colorado River to avoid contending with its living history.

  1. Recently: Arizona v. Navajo Nation, 21-1484, 2023; “Supreme Court: US Not Responsible for Water Rights; Navajo Nation Still Battling for Water,” Native American Rights Fund, June 22, 2023, https://narf.org/scotus-az-v-navajo-amicus/ ↩︎
  2. Please read: Andrew Curley, “Our Winters’ Rights”: Challenging Colonial Water Laws,” Global Environmental Politics 19, no. 3 (2019): 57-76 ↩︎
  3. Globally, water protectors and Indigenous communities continue to lead resistance against colonial-led water devastation, fossil fuels, and extractive economic systems, see: Leanne Betasamosake Simpson, As We Have Always Done: Indigenous Freedom through Radical Resistance (University of Minnesota Press, 2020); Nick Estes, Our History is the Future: Standing Rock versus the Dakota Access Pipeline, and the Long Tradition of Indigenous Resistance (London: Verso, 2019) ↩︎
  4. Matthew S. Henry, Hydronarratives: Water, Environmental Justice, and a Just Transition (University of Nebraska Press, 2023) ↩︎
  5. For example: Catherine E. Richards et. al., “Rewards, Risks and Responsible Deployment of Artificial Intelligence in Water Systems,” Nature Water 1 (2023): 422-432, https://www.nature.com/articles/s44221-023-00069-6; “Experts advise governors on how artificial intelligence and other tools can allow water in Colorado River Basin and elsewhere to be better managed.” “Advancing data to better manage Western water,” Big Pivots, July 5, 2023, https://bigpivots.com/advancing-data-to-better-manage-western-water/, accessed September 3, 2023; “Water rights, artificial intelligence, wildfire resiliency among highlights at American Bar Association meeting Aug. 3-8 in Denver,” ABA, July 19, 2023, https://www.americanbar.org/news/abanews/aba-news-archives/2023/07/ag-garland-to-speak-at-annual-meeting/, accessed September 3, 2023 ↩︎
  6. For further reading: Theodora Dryer, “Settler Computing: Water Algorithms and the Doctrine of Equitable Apportionment on the Colorado River, 1950-1990,” Osiris 38 (2023): 265-285, https://doi.org/10.1086/725187. ↩︎
  7. Teresa Montoya, “#WeNeedANewCountry: Enduring Division and Conquest in Indigenous Southwest,” Journal for the Anthropology of North America 22, no. 2 (2019): 75-78. For further history of how theft of Indigenous water was critical to settler land policy in the nineteenth century, see: Nick Estes, Our History is the Future (New York: Verso, 2019), chs 2-3 ↩︎
  8. Theodora Dryer, Amrah Salomón, Andrew Curley, Teresa Montoya, Erika Bsumek, “The Colorado River Compact at 100 Years,” panel at the Association of American Geographers, Denver, CO, 2023 ↩︎
  9. “Report: Data centers guzzling enormous amounts of water to cool generative AI servers,” siliconANGLE, September 2023, https://siliconangle.com/2023/09/10/report-data-centers-guzzling-enormous-amounts-water-cool-generative-ai-servers/ ↩︎
  10. My analysis of AI in water is informed by my sustained research into histories of data architectures and algorithms in environmental contexts. See: Theodora Dryer, “Seeds of Control: Sugar Beets, Control Algorithms, and New Deal Data Politics” in Algorithmic Modernity, ed. Morgan Ames Oxford University Press, 2023; “Big Data Stream,” Logic Magazine, September 1, 2021, https://logicmag.io/kids/; Designing Certainty: The Rise of Algorithmic Computing in an Age of Anxiety (PhD diss, University of California, 2019); “Algorithms under the Reign of Probability” IEEE Annals of the History of Computing 40, no. 1 (Jan.-Mar. 2018): 93-96 ↩︎
  11. “The High Frontier and Jeff Bezos’s vision of American Imperialism,” Political Economy Research Centre, November, 2022, https://www.perc.org.uk/project_posts/the-high-frontier-and-jeff-bezoss-vision-of-american-imperialism/ ↩︎
  12. “Elon Musk and a Breakdown of his Global Empire,” Orange County Register, June 4, 2023,https://www.ocregister.com/2023/06/04/elon-musk-and-a-breakdown-of-his-empire/ ↩︎
  13. “Frontier Model Forum,” OpenAI, https://openai.com/blog/frontier-model-forum ↩︎
  14. My critique of AI is a critique of the dominant white supremacist and colonial power structures of AI and does not extend to anti-colonial data and Indigenous data sovereignty research. Indigenous AI and Indigenous Data Sovereignty are respective movements and research directives, see: “Indigenous AI,” Indigenous AI Working Group, https://www.indigenous-ai.net/; “Indigenous Data Sovereignty and Governance,” University of Arizona, Native Nations Institute (NNI), https://nni.arizona.edu/our-work/research-policy-analysis/indigenous-data-sovereignty-governance ↩︎
  15. Please read: Yarden Katz, Artificial Whiteness: Politics and Ideology in Artificial Intelligence (Columbia University Press, 2020); Iván Chaar López, “Sensing Intruders: Race and the Automation of Border Control,” American Quarterly 71, no. 2 (2019): 495-518; Neda Atanasoski and Kalindi Vora, Surrogate Humanity: Race, Robots, and the Politics of Technological Futures (Durham: Duke University Press, 2020) ↩︎
  16. Each one of these innovations involves policy changes related to water pricing, water usage, water ownership, and water access ↩︎
  17. “The Secretary of the Interior acts as watermaster of the Lower Colorado Region, managing the delivery of all water below the Hoover Dam.” ↩︎
  18. Kasia Paprocki, “Threatening Dystopias: Development and Adaptation Regimes in Bangladesh,” Annals of the Association of American Geographers 108, no. 4 (2018): 955-973 ↩︎
  19. For example, a recent AI/ML water design report states: “Here we define AI as a machine-based ‘intelligent agent’ capable of interacting with its environment with the aid of sensors, interpreting information for decision-making and autonomously taking actions to achieve goal-oriented outcomes via a human or robotic actuator, while ML refers to the subset of algorithmic models that learn and predict outcomes through passive observation of the environment.” ↩︎
  20. Tega Brain, “The Environment is Not a System,” Research Values 7, no. 1 (2018)          https://doi.org/10.7146/aprja.v7i1.116062; Luke Stark, “Artificial Intelligence and the Conjectural Sciences,” BJHS Themes 3 (2023): 1-15, doi:10.1017/bjt.2023.3; Cindy Kaiying Lin and Steven J. Jackson, “From Bias to Repair: Error as a Site of Collaboration and Negotiation in Applied Data Science Work,” Proceedings of the ACM on Human-Computer Interaction 7, no. 131 (2023); https://doi.org/10.1145/3579607 ↩︎
  21. Xiaowei Wang, Blockchain Chicken Farm: And Other Stories of Tech in China’s Countryside (MacMillian Publishers, 2020); Krista Chen and Cindy Lin, “Myth of Tech Equalizer: Labor and Environmental Implications of Data Centers in Taiwan and Singapore,” Presented in ICA Conference, Toronto, CA, May 27, 2023 ↩︎
  22. Mél Hogan, “Data Flows and Water Woes: The Utah Data Center,” Big Data & Society 2, no. 2 (2015) https://doi.org/10.1177/2053951715592429; “AI Programs Consume Large Volumes of Scarce Water,” University of California, Riverside, May 11, 2023, https://www.enn.com/articles/72524-ai-programs-consume-large-volumes-of-scarce-water ↩︎
  23. Matthew Carroll, “Colorado River Basin: Machine Learning Approach May Aid Water Conservation Push,” Penn State, January 26, 2023, https://www.psu.edu/news/research/story/colorado-river-basin-machine-learning-approach-may-aid-water-conservation-push/, accessed September 3, 2023 ↩︎
  24. Renee Obringer and David D. White, “Leveraging Unsupervised Learning to Develop a Typology of Residential Water Users’ Attitudes Towards Conservation,” Water Resources Management 37 (2023): 37-53, https://doi.org/10.1007/s11269-022-03354-3 ↩︎