Winter 2025

Baked in

Its ethical problems may be nothing new, but as ‘AI’ invades graphic designers’ workspaces, its excessive energy consumption, resource extraction and exploitation of creative labour need tackling head on. By J. P. Hartnett

Marian Bantjes concludes ‘Artificial idiot’ (Eye 105), her analysis of generative AI, with an argument that the ‘moral, ethical and political problems’ associated with these technologies are neither new nor exclusive to them. So, while the widespread use of generative AI is undoubtedly a cause for concern due to well documented issues – bias, stereotypes and corporate greed are the examples listed – there is not much to distinguish generative AI, ethically speaking, from technology in general. This is hard to argue against. Authors such as Ruha Benjamin and Safiya Umoja Noble have shown how racist biases are programmed into commonplace computer algorithms, while the wealth amassed by technology corporations depends on routine flouting of environmental, labour and privacy laws and standards. Following this line of thought, it makes little sense to be any more concerned about the ethics of AI than an Adobe or iCloud subscription or even a simple Google search – to the extent that modern technology implies ‘Big Tech’, there are no ethical options available. Meanwhile, the once unsettling proposition of AI fades into the background of computationally mediated life. But, what if, in place of fatalism, the rapid adoption of the new tools might instead provide a spur to reconsider the relationship between graphic design, technology and the environment?

Just as public protest against data centre proliferation continues to highlight the materiality of the ‘cloud’, the notion of AI as a magical entity animating our computers crumbles with each headline about the exponentially spiralling energy consumption driven by the AI ‘arms race’ between the world’s major technology companies. A report in Jacobin notes that Google’s recent integration of generative AI within search uses ten times more energy than its predecessor, while the company’s carbon emissions have risen by 50 per cent in the past five years as a result of investments in AI technologies. The International Energy Agency estimates that global energy demand from data centres, essential for powering AI, will double (from 2022) by 2026. These developments are occurring when demands for decarbonisation and degrowth have never been more urgent: 2024 saw many heat records broken globally. The junking of once proudly touted environmental policies in favour of AI expansion has shown yet again the hollowness of corporate environmental commitments and the futility of faith in ‘green economics’.

How might we situate graphic design within these issues? First, it is necessary to dispense with the idea that graphic design can be thought about as distinct from other aspects of the designed world. A crucial lesson comes from the ‘decolonial’ critique of design – led
by the Decolonising Design collective, the Futuress publishing platform, Arturo Escobar (see review in Eye 99) and others – which has radically problematised design’s role within Western modernity’s drive for ‘progress’. In canonical accounts of European graphic design history, there is a rare reference to the environment in the conclusion of Modern Typography, where Robin Kinross cites ‘irrevocable and disastrous damage to the natural world’ as ‘the great negative of the modern.’ From a decolonial perspective, environmental destruction is not an unfortunate by-product of the otherwise noble project of modernity, it is inseparable from its entire logic. The economic, scientific and technological advance of Western powers – which supported emergent cultures of design in Europe and North America – could not have been achieved without the colonial violence underpinning it, the impacts of which continue to be felt.

Defuturing the way we work

Frequently cited among such critiques, design theorist Tony Fry (see Eye 102) has long argued that the discipline of design history has failed, given the scant attention it has paid to the destructive effects of design: ‘A great deal of what has been brought into being by design, from coal-fired power stations to asbestos, from herbicides to jet-skis, from cigarettes to cluster bombs, all combine to take the future away. They “defuture”. In particular, industrial society has brought these, and a myriad other defuturing things and forces, into being. As members of such a society, we find ourselves at the currently comfortable epicentre of the condition of unsustainability that defuturing animates.’ A critical point for Fry is that ‘whatever is designed goes on designing’ – the damaging consequences of design must be lived with long after the fact of creation. They are ‘baked in’, to use a popular phrase, though unfortunately more literally with respect to design’s impact on global heating.

AI brings Fry’s concerns into sharp relief, since ‘the future of AI’ is a significant component of what, in the present, is being bought into by technology companies as they scramble to gain market supremacy. But in order for that future to be realised, another possible reality – of sustainable and ethical use of resources and energy – must be rejected. Kate Crawford’s Atlas of AI attempts to map the various dimensions of this rejection: from unsustainable mineral extraction (used to power computing), to exploitative labour (used to train AI), to unethical data gathering and classification, to invasive state surveillance. The book ultimately highlights how inextricable AI is from the internet’s sprawling operational structures.

But for Tiziana Terranova, ‘the internet’ no longer even exists – it was an infrastructure originally governed by public and non-profit organisations. That has been replaced by what she terms the ‘Corporate Platform Complex’ (CPC), an entity governed by a relatively small number of very large technology companies. This transformation has meant that the internet now ‘operates in the background [of everyone’s lives] but it no longer signifies for most people the possibility of a different, better world.’ It has been subsumed completely by capital. Yet the ‘CPC’ was not always the only version of the internet that existed. And by implication, it is not necessarily the only version that could exist.

A seminal use of AI within graphic design occurred during the 1980s at the Visible Language Workshop (VLW) in MIT where Muriel Cooper, her colleagues and students were busy experimenting with new AI technologies under the heading ‘Intelligent graphics’, as documented in David Reinfurt and Robert Wiesenberger’s monograph on Cooper (see Eye 96). Under the influence of AI researcher Henry Lieberman, Cooper believed that ‘a more responsive, intuitive interface combined with machine intelligence could produce software that worked symbiotically as a designer’s assistant.’ Anticipating a future in which managing vast quantities of information would be a designer’s primary task, many of the projects explored automation as a means to enhance the designer’s workflow. Lieberman developed software that could ‘learn’ how to construct layouts based on examples, though Reinfurt notes that the success of this kind of learning was dependent on the quantity of data available, foreshadowing what would become one of the central concerns for AI today. The VLW projects can be understood as theoretical exercises to determine what the technology might be capable of – AI as an idea. New tools are overcoming the limits of such experiments as a result of huge increases in the resources – labour, money, data, minerals, energy, water, computation power – supporting their growth.

A notable historic intersection between theories of design and artificial intelligence can be found in the work of Herbert Simon, who authored one of the most cited definitions of design: ‘Everyone designs who devises courses of action aimed at changing existing situations into preferred ones.’ Simon (1916–2001) was a consultant at the RAND Corporation, an influential think tank that shaped US military policy during the Cold War, and a pioneer in artificial intelligence research during its early development in the 1950s and 60s. In The Sciences of the Artificial he articulates an elaborate argument for understanding the artificial world – the world made by human beings, such as economic systems, businesses, social plans, engineering – in scientific terms. If an accurate enough picture of this artificial world can be constructed, then the problems of society can be solved. Design comes into play as the ‘science’ of solving these problems – the changing of ‘existing situations into preferred ones’. Computers and AI are essential, since being modelled on human intelligence they can extend its capabilities; while human intelligence can equally be better understood through the simulations of AI, creating a feedback loop between the human and the artificial.

A deadly fiction

Central to Simon’s theory is the idea that artefacts made by humans – from simple objects to complex systems – represent an interface between an ‘inner’ and ‘outer’ environment. The inner environment is the ‘substance and organisation’ of the artefact itself. The outer environment is the ‘surroundings’ in which that artefact operates. The aim is to design the former in such a way as to attain goals in the latter – to adapt what is in here to change what is out there. But this theory depends on being able to neatly compartmentalise what are, arguably, non-compartmentalise-able components. For example, what is the ‘inner’ substance of an image generated by AI, when it is an aggregate of millions of images drawn from the ‘outer’ environment and is generated by a system reliant on resource extraction, human labelling, and vast energy consumption in that ‘outer’ environment? In AI, the inner is the outer environment.

Simon’s argument is not simplistic, but his view of the world, as something that can be reduced to independent components and binary oppositions, is at odds with recent discussions in design theory that argue for the necessity of understanding the interconnectedness of things, such as Escobar’s new book Relationality, co-authored with Michal Osterweil and Kriti Sharma. In 2006, Eye magazine published an article on ‘relational aesthetics’ (‘Part of the process’, Eye 59) that documented the emergence of art and design practices concerned with creating spaces for social interaction that would work against the behaviours prescribed by consumer capitalism. Although, for Escobar, Osterweil and Sharma, forms of design that create such encounters would be welcome, a more profound lesson about ‘Earth and humans’ constitutive relationality’ must be grasped.

Building from a rich assemblage of different perspectives, the three authors argue for the indivisibility of the earth and its inhabitants. They ask: ‘If we really saw ourselves as part of the earth … would we scour, and scar, the earth in search of minerals so we can live half-absorbed in front of our digital screens?’ Western modernity – which incorporates not only ‘Big Tech’ but also the global design industry – has advanced a form of ‘reductionism’ that has sidelined many ‘aspects that are crucial to life and to the making of social reality’ and has resulted in widespread ‘ecological devastation’. If design really means changing existing situations into preferred ones, and AI is seen as one means for designers to achieve such goals, it is difficult to see how the current trajectory of AI development – one of vastly increasing energy consumption in the Global North and ever more intensive resource extraction and labour exploitation in the Global South – could constitute a ‘preferred situation’ in any rational analysis.

For computer scientist Timnit Gebru, the person ‘who creates the technology determines whose values are embedded in it.’ Gebru, allegedly fired from Google for co-authoring a research paper critical of the ethics of popular AI models, has called for AI researchers to ‘learn about the ways in which their technology is being used, and question the direction institutions are moving in.’ Should the same not be said of designers working with AI? If reports are true, the majority of networked technologies will employ AI soon enough, if they do not already, and so the question of ‘whether or not to use AI’ is fast becoming moot. But this is not the same as saying that there is nothing to be contested. For Escobar, Sharma and Osterweil this is no less than the question of what is real. For them, the belief that we can act on the world as if we are not constituted by the world is a deadly fiction. Joanna Boehnert (see Eye 105) has argued that graphic designers are well placed to help expose this, but such actions will require confronting AI’s ‘moral, ethical and political problems’ head on.

First published in Eye no. 107 vol. 27, 2025

Top. Detail from Be Happy by Steven Ryan, Substance, part of a series of posters called Br(ai)ve New World’ shown at the Design Museum of Chicago, 2024. See pages 68-69 in Eye 107.

Eye is the world’s most beautiful and collectable graphic design journal, published for professional designers, students and anyone interested in critical, informed writing about graphic design and visual culture. It is available from all good design bookshops and online at the Eye shop, where you can buy subscriptions and single issues.