What will our AI-shaped future look like?

AI Morality
David Edmonds (editor)
Oxford University Press
July 2024
256 pages
ISBN:  9780198876434

---

In May of last year, The Irish Times published an opinion piece by Adriana Acosta-Cortez, a twenty-nine year old healthcare worker living in north Dublin. Acosta-Cortez described herself as “Latinx”  and a “cat mom” in her article and social media profile. Acosta-Cortez’s article was about Irish women’s use of fake tan, which it decried as “problematic”, “self-hate”, and a culturally appropriating “fetishisation of high melanin content”. The article evidently satisfied the keen eyes at the editorial desk as it successfully made it to publication and became the second most read online article on the newspaper’s website that day.

In the hours that followed, however, it transpired that something was amiss. It turned out that Acosta-Cortez did not actually exist. Her social media profile (which is still available, and contains a link to an archived version of her article) had been made up. Her profile photo, a feisty blue-haired young woman, had been computer generated. And, perhaps worst of all, the article itself had been produced using AI. The newspaper, effectively Ireland’s version of The Guardian, pulled the story from its website, and the following day a red-faced editor had to apologise for the “gap in our pre-publication procedures” highlighted by the affair.

The whole event raised concerning questions. What kind of due diligence, if any, was carried out by the editorial team? Does such minimal oversight apply to all articles and content, or only to some? On a different and more depressing note, the affair raises questions about what passes for readable material nowadays. If AI can generate an article in a matter of seconds, suitable to the tastes and values of Ireland’s aspirational classes, and gain the imprimatur of the editorial team of its premier newspaper, then one must ask how the quality of thinking and writing that feeds the nation’s minds has become so insipid. We hear much about the havoc that ultra-processed foods can wreak on our bodies nowadays, but how healthy and discerning is a country’s intellectual culture when artificial intelligence can concoct an ultra-processed article, mental convenience food, and release it into the food chain unnoticed? Perhaps worst of all, the affair laid bare the kind of anaemic and banal diet that seems to sustain Ireland’s cultural and intellectual establishment.

A recently published book brings some of these concerns into sharper focus. AI Morality, edited by David Edmonds, is a collection of essays from Oxford University Press. Edmonds is Distinguished Research Fellow at the Uheiro Centre for Practical Ethics at the University of Oxford, and most of the book’s contributors have affiliations with either this institute or Oxford’s Institute for Ethics in AI. Published in July of this year, the twenty essays in this collection cast an informed and up-to-date eye over an area that is evolving very rapidly. The book is very accessible, with each essay extending to no more than ten or fifteen pages. Many of the contributions are lively and astute. They approach the theme from a variety of viewpoints, and the best among them tease out or problematise matters we tend to take for granted, or are not even aware of at all.

In his introduction, Edmonds highlights the pervasiveness of AI, an “ever-increasing presence … in almost every aspect of life”. He sets out a working definition of artificial intelligence: “In essence … it entails the performance of tasks by computing systems that previously would have needed human brain power”. The “intelligent” part of artificial intelligence, which makes it so distinctive from earlier kinds of software and robotics, is: “the ability to ‘learn’, so that performance on tasks can improve as more information is inputted over time”. AI thus “identifies patterns that are often blind to humans and makes predictions and inferences on the basis of past information”.

Some of the book’s essays evaluate AI’s application in relatively benign situations, such as suggesting another item to view or purchase in your Netflix or Amazon account based on your previous behaviour. Other essays consider much more troubling and invasive uses, such as the Chicago Police Department’s experiment in using an algorithm to list and score people according to their probability of being involved in a homicide or non-fatal shooting. On the one hand, AI can seem to liberate us from mundane tasks such as online TV and shopping, yet on the other it implicates itself in an invasive and determinist “guilty until proven innocent” scenario of profiling and surveillance.

One of the most thought-provoking essays in the collection concerns the impact of AI on work. Daniel Susskind’s “Work and Meaning: A Challenge for Economics” argues that the challenge of AI to work “is not just that the labour market might be hollowed out, leaving some workers without work or with a different type of work, but that it might hollow out that sense of meaning as well”. A professor of economics at King’s College, London, Susskind critiques the historically “very narrow” understanding of work in the field of economics – pithily summed up as “toil and trouble” in the words of Adam Smith, the father of the discipline. Contemporary economic models tend to afford very little room for understanding work as meaningful. Rather, Susskind finds a “conflict between the healthy relationship between work and meaning appealed to by economists in public commentary, and the harmful one reflected in formal models”.

Studies have found that factors as varied as “retirement intentions, absenteeism, and skills training” are influenced by the extent to which people view their work as purposeful and worthwhile. Susskind argues that we have underestimated the extent and kinds of jobs that AI can challenge. Professional, white-collar roles, which were once considered safe due to the non-routine nature of some of their tasks, could find aspects of their work automated, their workspace closely surveilled, and unpleasantly precise performance targets set at the whim of an algorithm, for instance. Consequently, even if AI “does not reduce the quantity of jobs, it may greatly affect the qualities of jobs available”.

Susskind discusses two proposed interventions to the potential decline in work: a Universal Basic Income and a Job Guarantee Scheme. If one accepts that work and meaning are linked, “then a JGS [Job Guarantee Scheme] is likely to be [more] appealing”. Under such a scheme, Susskind argues, “everyone is provided with paid work, funded by the state”. Such an idea is not entirely convincing. Lacking any of the dynamism of a market economy, it has strong hints of a highly bureaucratised, quasi-Soviet make-work regime, which would, in the end, actually grant little or no sense of meaning to people’s work.

Overall, however, Susskind is to be credited for making the case for work as a source of meaning in the field of economics. He raises an important question, one that only the human heart, rather an a string of code, can answer: “if technological progress is carrying us towards a world that is quite different from our own, where there is not enough demand for the work that human beings do to keep everyone in a good job, then we must confront the nature of the work-meaning relationship more seriously”.

In another essay later on, John Tasioulas takes up the theme again, and further develops the relationship between work and meaning. In “Work and Play in the Shadow of AI”, Tasioulas, Professor of Ethics and Legal Philosophy at Oxford’s Faculty of Philosophy, asks what it means “to find fulfillment as a human being in a world in which AI-based systems are undertaking tasks that have characteristically given human life its point”. The implications of AI for work are significant. “There has never been a technology that has the potential to replace human work activities on anything like a comparable scale”, Tasioulas writes. “This goes well beyond carrying out ‘routine’ or ‘mechanical tasks’ to include white-collar occupations, such as journalism and legal services”. As we have seen, well constructed yet entirely fake articles can be generated by AI and appear in esteemed national newspapers.

Tasioulas observes that “one of the distinguishing features of modernity is that ordinary life, including work activities, is treated as a scene of genuine human fulfilment”. This is in contrast to earlier generations, where factors such as famine, disease, and even just the sheer physicality of labour made life and work harder and far more unpredictable. Thus, argues Tasioulas, a “core good” of work in today’s world is achievement. Particularly in the developed world, people work for reasons other than, or in addition to, survival, such as experiencing community, gaining social or professional recognition, or making a social contribution. Consequently, rather than feeling a sense of liberation, having no work poses significant risks to people’s sense of purpose, feelings of achievement, and experience of community.

Developing the concept of achievement further, Tasioulas proposes an unusual and interesting possibility – “that the post-work utopia will be one in which we are primarily occupied in playing games”. It is hard not to feel simultaneously curious, skeptical, and excited by such an idea. Children play games from their earliest years. This is how they learn about the world and socialise with those around them. Grown-ups play games too, both competitively and for fun. Tasioulas writes: “achievement is the core value of game-playing, and game-playing presents this value in a manner that vividly exemplifies the mindset of modernity, because the value primarily resides in the difficulty of the processes of achieving a goal, not the goal itself”. Playing a game is as much about the journey as the destination – be it tactics, technique, or the agonising minutes before the final whistle. A win is good, but it’s the struggle that makes it worth it.

Ultimately, whether games can truly act as a surrogate for work remains unclear. Even Tasioulas acknowledges “the distinctive form of contact with physical reality that work can afford”. Work sees humans engage with – even wrestle with – “an independent and potentially recalcitrant physical reality”. Perhaps Karl Marx, who continues to cast a powerful shadow over modernity, understood this best of all. Drawing on Hegel’s concept of dialectic, Marx saw that “work has crucial significance for human self-realisation because it involves the struggle to transform nature into a humanised domain of culture”. Perhaps one of the unique paradoxes of being human, the struggle of work, of imposing a human shape upon an obdurate physical reality, represents at once a great strain and a means of personal achievement for us. We might be counting down the years till our pensions, yet we might not realise what we’ve lost till it’s gone.

Whatever our AI-shaped future will look like, Tasioulas concludes, “it is clear that one of the gifts of AI technology to philosophy is that it requires us to reflect anew not only on the value of work and play, but on the ultimate nature of human well-being itself”.

One potential drawback of books such as this, so up-to-the-minute in terms of their subject matter, is that circumstances change very quickly. A paper that seemed very relevant today could have aged very poorly in six or twelve months’ time. One author’s proposal to establish “data cooperatives” to act as intermediaries between data buyers (big tech) and data producers (us) seems wildly idealistic and impracticable even today when one thinks of the sheer volume of data produced every second across the globe. A lawyer’s and bureaucrat’s dream, perhaps, but impossible to execute without significantly hindering the technology’s development. Other essays, meanwhile, discussed highly specific or hypothetical situations from which it felt difficult to draw broader philosophical lessons.

Overall, however, the book’s different voices offer different perspectives, both glowing and guarded, on this new and rapidly developing phenomenon. In his introduction, Edmonds notes that artificial intelligence can, in some circumstances, challenge our autonomy and “erode the self-control we have over our lives”. And for all its intelligence, it has no sense of responsibility either: “algorithms can’t be punished”. Hurling your phone upon the sofa will have no effect on the cold and indifferent algorithm that is drawing its meaning (and data) from your existence. Thinking further about the interesting complementarity of Susskind’s and Tasioulas’ papers – arguably the most thought-provoking in the collection – it is clear that AI has the power to significantly colour and influence, improve or disimprove, our physical reality, but ultimately it cannot replace it.