Karl Marx predicted the rise of artificial intelligence (AI) over 170 years ago. Buried within a dense, unfinished manuscript that he began in the 1850s but that remained unpublished until 1939, Marx predicted that “the means of labour passes through different metamorphoses, whose culmination is the… automatic system of machinery… set in motion by an automaton, a moving power that moves itself.” A complete and utter change in both form and function to the means of production.
AI can be loosely defined as the capacity of a machine to imitate human intelligence, whether in intensive pattern recognition, decision-making, or visual perception. It is set to reshape our society in ways we can barely imagine.
A 2016 study from the University of Oxford suggests that AI will eliminate about half of the United States’ human labour. In countries like Ethiopia where agriculture composes a substantial part of the economy, that number rises to 85 per cent.
This massive replacement of human labour by AI will radically change the structuring of our economic institutions. Some, including Elon Musk and Jack Ma, have argued that AI poses a cataclysmic threat to humankind — one that must be curbed.
As opposed to the approach laid out by billionaires — who may have very good reasons to resist change to a system that has worked quite well for them — I believe that the introduction of AI gives us an unprecedented opportunity to radically transform our economic and social systems for the better. Working not against AI, but with it, will allow us to build a truly just society from the upcoming ‘fourth industrial revolution.’
So what is justice, anyway?
Navigating this crossroads is no easy task, so I opt to employ a guide that will assist in building a just society. John Rawls, widely considered to be one of the most influential philosophers of the twentieth century, presented an influential model of a just society in his aptly titled book, A Theory of Justice. In it, Rawls argues that justice is, in essence, fairness.
The proposed system is elegantly simple: it includes only two basic principles. The first is the “Liberty Principle.” In a just society, each person should have an equal set of basic liberties, familiar to those of us who live in liberal societies: freedom of speech, assembly, conscience, and thought; the right to vote for and hold public office; and so on. However, as Marx noted, while individuals can be religiously or politically free in a liberal state, they may not be able to act on those freedoms due to material constraints.
This brings Rawls to his second “Difference Principle.” This rule is egalitarian in nature; it prohibits all positions of inequality that are not in principle available to all — think equality of employment — and ensures that the inequalities benefit the least advantaged.
Having this roadmap in mind can help us assess the two questions that I think are at the heart of the debate over AI in the context of a just society: should AI be embraced, and who should own it?
Rise of the robots
Many are convinced that we should be worried about AI. The headlines write themselves. A Mother Jones article warned against job replacement, saying that “mass unemployment is a lot closer than we feared.” A piece in The Guardian discussed a “hollowing out” of the global middle class. Entrepreneur published a piece criticizing the European Union for its lax guidelines on AI.
Calls for halting the progress and implementation of AI are understandable, given that labour change has always been feared in capitalism. The iconic example of this are the Luddites, who destroyed textile machinery in protest of the replacement of traditional labour in the early nineteenth century.
However, maybe the Luddites weren’t necessarily protesting the machines as such. I imagine they would have been perfectly content with the implementations of textile machinery, which made their jobs less strenuous, safer, and more efficient, if their quality of life was not affected in the process.
The main concern people have with machinery, past or present, is not whether they will replace human beings. It is about income security — but we will get to that later. Disregarding the loss of human labour, there is insurmountable potential for AI to create a more just society by improving upon our civil liberties, including our freedom of expression, assembly, fair trial, voting, life, and privacy.
It may, however, be tempting to ask whether liberal freedoms can be improved at all. After all, it seems that in an ideal liberal-democratic society, everyone already possesses these freedoms. There are no bars stopping racial or sexual minorities from voting, no Ministry of Public Security to thwart peaceful protests or democratic dissent, no ‘disappearance’ of dissenting journalists.
It is true that we enjoy an overwhelmingly inclusive set of civil liberties. However, the problem lies in the fact that their potential for application is unequal across the board. For example, a homeless man may very well have equal freedom of expression as the editor-in-chief of The Globe and Mail, but they do not have an equal chance to employ this freedom.
This pattern is much the same with other basic civil liberties: someone who lives in a rural community has a much harder time going to vote as opposed to urban dwellers; a person who holds major office can disseminate their voice much more effectively than a regular Joe; a Christian has an easier time practicing their religion than a Muslim. When applied to Rawls’ theory, this “scheme” of civil liberties is incompatible across the board, and is therefore, at least to some degree, unjust.
AI could act as the great equalizer for social liberties. It can make the world more accessible and freer for all users. The ability for advanced and logical problem solving can help overcome many barriers to civil liberties.
A self-directed and corporeal algorithm can, for example, help reach out and organize a protest, direct people to polling stations (or better yet, allow them to vote directly from their personal devices), disseminate voices to a large and diverse audience, and ensure that there is a fair legal process by representing clients.
It can allow for more accessible and integrative methods for religious practice, enhance our healthcare system, and protect our personal information and data from malicious software. A capable, accessible, and sophisticated intelligence has almost unlimited potential for the encouragement of positive liberties.
However, I would like to caution that current use of AI can also corrupt, rather than amplify, civil liberties. Predictive algorithms are used by police to disrupt peaceful protests, screen social media posts, or censor political opposition. Nevertheless, I believe that there are two sides to the AI story, and AI has the potential to ensure that our society is all the more just.
Some may scoff and point out that the picture laid above would be simply impossible to realistically implement. Why, they may ask, would the companies that create these algorithms be interested in enhancing personal liberties?
Indeed, it may be very well in their interest, and the interest of their shareholders, to limit our personal freedoms. They may do this by keeping us in echo chambers, blocking us from consuming content that is critical of companies or governments, or by selling personal information gathered by the algorithms to maximize profits.
How do we trust that AI will not exploit our civil liberties in the name of profit? The answer lies in the nature of ownership and control of the technology.
All ‘bout the money
Here, we arrive at the more pressing question: who should own AI? I mentioned previously that the main concern about machinery is not its nature — rather, it lies with a potential decrease in quality of life. Here, Rawls’ second principle can come in handy.
There is a lot of money to be made in AI. The McKinsey Global Institute estimates that AI has the potential to create between $3.5 trillion and $5.8 trillion annually across nine business functions in 19 industries and boost global productivity by around 1.2 per cent.
That may not sound like much, but the introduction of the steam engine and the spread of information technology only increased productivity by 0.3 and 0.6 per cent, respectively.
This outpour of wealth, however, would not satisfy Rawls’ second principle of justice under our current capitalist system. There would inevitably be an unequal distribution of the benefits and burdens of AI, one which would make the rich richer and the poor poorer.
A fully automated capitalist state would have two main outcomes, both of which would produce financial inequalities which would neither be accessible nor beneficial to all. First, there would be a marked loss of jobs.
Many middle-class jobs are classified as routine abstract work, such as accounting or customer service. Given that AI could easily perform most desk jobs, those slots would eventually disappear. Additionally, routine manual work, such as package delivery or agriculture, would also be gone.
The two fields that machines would presumably not replace are non-routine manual work, such as childcare or healthcare, and non-routine abstract work, such as architecture, coding, or researching.
This lack of jobs would be coupled with low money circulation. The number of transactions would fall, and money would instead be used for investments or savings. Given that only a select few would be privileged enough to hold their jobs, this would result in the gross accumulation of wealth in their hands because only they would have a disposable income.
This would, in turn, contribute to economic shrinkage, a terrifying concept for capitalists since it means that resources would be harder to acquire. The shrinkage would, in turn, result in more automation, since the capitalist class would benefit most from it, and more jobs would be lost. On and on the cycle would go, until the system presumably collapses.
The ‘just’ invisible hand
An automated capitalist state cannot uphold any of Rawls’ principles of justice. First, civil liberties could not be upheld in an equitable manner.
Superfluous economic power often translates into political power. Lobbying groups for the capitalist class would inevitably spring up, and with them a host of candidates who latch onto the rich for their own benefit.
White-collar crime is rarely prosecuted as severely as blue-collar crime, and the rich are often given much more leeway in the political and judicial system. The fair value of civil liberties would be controlled by the handful who managed to keep their jobs, leaving large swaths of the population with little to no political power.
This heightened inequitable distribution of civic freedoms would then render this society unjust. Decisions of societal importance would be within the scope of the market, and not by democratic process, to a much greater degree than today.
Jobs would also not be equally accessible to all. Today, we can see that inequalities drive inaccessibility. This effect would be multiplied by the extreme inequalities caused by automated capitalism. The recent college admission scandal shows this on the micro scale. The number of admitted positions at elite American colleges should, in theory, be egalitarian. The fact that students from affluent families could blatantly pay to get the spots is obviously unjust.
On the macro level, accumulated wealth allows systemic inequalities, where high-paying jobs are reserved for a limited demographic. Upward social mobility has been on the decline in the United States since the 1980s — coinciding with the introduction of Reaganomics — with middle- and low-income children struggling for upward mobility while children of the upper class inherit fortunes.
This perpetual cycle, through which high-income jobs are reserved for a chosen few since others could not hope to acquire the needed skills, would only be amplified by the dramatic inequalities that would stem from AI.
Lastly, any benefits that AI might incur would disproportionately go to the privileged. The means of production are currently held by a small group of individuals, and the loss of jobs and money circulation by AI would only accentuate this.
The Economic Policy Institute reported that in 2017, the average CEO at one of the 350 largest firms in the United States saw a salary increase of 17.6 per cent from the previous year, equating to an $18.9-million USD jump.
This jump could only be deemed just if it benefits the least advantaged in society. However, just by looking at employment compensation, we can deduce that this is not the case. While the rich got richer, the average employee compensation remained flat, rising by only 0.3 per cent.
This is not just a one-off year — this pattern has been observed in the capitalist system for some time. The gap between executives and workers has widened for decades, with the CEO-to-worker compensation ratio jumping to 312-to-one from the 20-to-one ratio in 1965.
Clearly, the inequalities are not benefiting those who are least advantaged, which points to a systemic flaw. AI would increase the wealth of the few, but the current system will ensure that the wealth does not trickle down.
An automated capitalist state, then, would not fulfil any of Rawls’ principles of justice, and would therefore be unjust.
If not capitalism, then what?
I’ll be blunt. I think that the only way that a more just, fair society can emerge from this would be the implementation of a fully automated democratic socialist state.
Democratic socialism is defined as a system by which all major means of production are publicly owned. This means that as a constitutional right, all citizens own and administer the assets they require in order to be cooperating members of society. In an automated society, AI used in major companies would be administered and profited from by citizens.
Many may point out that historically, socialism has not always reached its full potential. In countries such as China, Venezuela, or Cuba, the administrators of the system have amassed vast amounts of wealth through illegitimate means, which has led many skeptics to conclude that human nature itself rules out the possibility of true socialism.
However, in an automated world, human nature may not need to play a part in the distribution of the means of production.
In an article for The Washington Post, Feng Xiang wrote that, “If AI rationally allocates resources through big data analysis, and if robust feedback loops can supplant the imperfections of ‘the invisible hand’ while fairly sharing the vast wealth it creates, a planned economy that actually works could at last be achievable.”
Moreover, a nationalized AI system would mean that the elimination of wage labour would not pose a problem, unlike under a capitalist system. Human capital would be largely replaced by resource and information capital. Whereas this would spell out the inevitable collapse of capitalism, it would help a socialist system flourish.
The work and wealth that would stem from AI would be owned and administered by society at large, ensuring that the fruits of production would benefit all. This means that the first clause of Rawls’ second principle would be upheld. Since there would be no “offices” to be held, as human labour would be essentially eliminated, there would be no inequalities in attaining different offices.
This may sound like a frightening idea. After all, we are used to dedicating a majority of our lives toward a profession, whether out of a sense of duty, need, or passion. Eliminating that requirement could make life seem strangely barren.
However, I think that if we really internalize what the abolition of mandatory labour would mean, we can come to an understanding that we would, in fact, be much happier people.
Lack of work would make us much freer than we are today. We would have time to finally get to the book we’ve been meaning to write, to spend more time eating a nice brunch with friends, or to finally master juggling.
A world with little to no required work would mean that the work we do perform would be all the more meaningful. It would be done solely for our own personal enjoyment, or for the benefit of others, not for a perceived duty.
Positions would be defunct, only replaced by equally accessible activities which create no tangible inequalities, thus fulfilling the first clause of the second principle.
In an automated socialist state, natural resources and machinery, the main means of production, would belong to and be administered democratically by the public.
Each citizen would then have a guaranteed source of income, which would mean that any inequalities that arise out of luck, such as in the capitalist case, would be eliminated. The remaining amount of administrative labour would be equally divided among the population as part of their civic duties.
Thus, the benefits and burdens of AI would be equally distributed. Collaboration, rather than competition, would rule the economic market, with an increase of output meaning an equal rise in profits for all. Any fluctuations in capital would equally impact all, fulfilling Rawls’ second principle.
There are, of course, many other benefits of AI that I didn’t touch upon. Better healthcare, decision-making, research, forecasting, disaster response, energy distribution, and much more can emerge from the implementation of AI.
Most of all though, AI provides us with the unprecedented opportunity to transform our society into one that is more just.
Following Rawls’ classic theory of justice, we can safely conclude that implementing a fully automated democratic socialist state could create a truly just society. Some may argue that doing so would be like building castles in the sky — however, we’ll never know until we lay down the first brick.