“Creativity and AI”;
At first glance, creativity may seem like a uniquely human endeavour. People often view it as spontaneous and uncontrollable, perhaps a fleeting moment of inspiration that, once gone is gone forever. But creativity is not born from nothing. If you speak to creative people, they will agree that it is more of a process. If one comes to view it as such, then the creative process can be replicated by machines.
Music is perhaps the most receptive the form of creative arts. Ancient philosophers like Plato (probably) even wrote that music is a moral law. Inspired by the process of evolution and natural selection, a team of AI professionals at Melomics Media took to creating a software, which they named Iamus. Iamus had been able to independently compose thousands of pieces of contemporary classical music, some of which were later performed by the London Symphony Orchestra. But Iamus’ limits do not end with compositions of the modernist classical genre. The computer can compose music of various styles, including rarely used instruments and scales in different cultural contexts like Hindu or Arabic music.
Music is not the only creative art form that is being automated. When speaking of fine arts, one can not possibly discount the visual arts and painting. Intent on seeing whether robots could be accepted as creative artists, professor Simon Colton at the University of London built an AI programme called The Painting Fool. The Painting Fool software first detects emotions in people, through photographs as well as a real life model, and then paints a portrait in order to convey a particular emotional state. Some of the software’s artworks have been displayed in galleries and art exhibitions, such as one occasion where 222 of its portraits were showcased in Amelies Progress Gallery. It uses a number of techniques and styles to achieve its objectives as it paints, and is capable of self-criticism to decide whether it has met those goals using another incorporated software application. The Painting Fool is only one ‘creative’ software among many. There are other notable examples, including AARON by the artist Harold Cohen over thirty years and e-David by the University of Konztanz in Germany.
In the mid 1980s another team at Google produced the first version of RKCP (Ray Kurzweil’s Cybernetic Poet), an algorithm that writes authentic poetry using models and analysis of natural language. In 2013, Benjamin Laird and Oscar Schwartz launched an online Turing test for poetry, using poems generated by the second, more sophisticated version of RKCP. In his essay The Imitation Game, Alan Turing said if a computer could deceive a human 30% by making them believe they were interacting with another human being, then it successfully passes the test. After analysing the answers from thousands of participants, they found that some poems written by RKCP were actually able to fool human readers 65% of the time into thinking it was a human poet. Although the results may arguably be down to the judge’s taste, it was found in earlier tests that the judges’ level of experience with poetry did not enable them to identify the poems correctly any more than an inexperienced judge.38
The Turing Test is particularly important because it gives us a measure of not only how humans interact with one another, but also of what sort of qualities we expect in human beings and in our interactions. A central idea in the debate about art and aesthetics is how the former offers a reflection of the human condition, and a great part of the human condition is shaped by how we treat others and ourselves. This is why passing the Turing test is regarded as a crucial threshold in determining the ‘intelligence’ of computers.
But if the hallmark was to pass the Turing test, many creative computational software (including Iamus and RKCP) should have already been considered as, or close to ‘intelligent’. But many are still not convinced. Critics have pointed out that regardless of what behaviour and algorithm are built into the software, a machine will never be a person; for it has no awareness of what it is truly doing, no sense of purpose, no intention. Although later researchers have attempted to fill in the aspects of humanity the test neglects by modifying it, it remains that the Turing test is limited because its original intention was to test whether a machine could think in a fashion that is indistinguishable from human beings, and not at all for whether or not a robot can pass for one. Indeed, there are evidence that human poets who fail the Turing Test, meaning that they or their work were mistaken for, or viewed as less human than a computer’s!42 And there are instances when subjects actually preferred a computer-composed sonnet over one written by Shakespeare. Along the lines of the Turing Test, OZ Paradigms and Searle’s Chinese Room Argument were developed to challenge the idea that human and computer minds could ever be comparable or even remotely similar.
Ultimately though, the Turing Test still serves as very useful thought experiment and contributes heavily to how we debate about AI and human-computer interaction. When we ask whether a computer can truly create art, what we are really asking what are the human qualities and values we want to see in art. But human values are not always constant. It is an idea that changes over time. What we build into our machines, be it natural language or behaviour, is going to reflect our values back at us. In the future, as computational technology continues to change and evolve, one thing is almost sure to occur, if it is not already occurring. That is artificial creativity will invoke many new and interesting questions, such as ‘How would the works of art of a robot or computer be sold? Who would want to buy it, and if not, why would the art be seen as not having as much aesthetic value as work produced by a human?’ and force age-old ones to be revisited about what gives an artwork a monetary value and what does it mean to be an artist.
Conclusion
As explained above, the question of whether AI and computers can write poetry, paint and compose music is no longer in doubt. It is possible to programme robots to teach themselves how to do any task we want them to do, and they can do it well. The arts only make up a small part of the economy, so the pressure on them is perhaps less. But the situation is much more urgent for the larger population employed in low-skill level jobs (and soon enough, even white-collar, high-skill level ones). A study by the University of Oxford put the estimate of 47% of jobs in the US are vulnerable to automation within 20 years. This is why it is critical that we think now about what work and employment in general means to us and what might we and future generations do about it.
On a practical level, new technology will force the demand for labour to come down. Companies will start replacing workers once automation technology becomes more cost-effective than employing human labour, so the demand for labour will be less than before. In contrast however will be the rise in demand for labour servicing new technology. In particular, some of the new jobs created will be in the domain of interpersonal interactions that are particularly difficult to automate, for example areas like nurturing, caring, persuasion, sales and coaching. Social and creative skills will also become a vital part of what future generations will need to learn. As a result of this, it is possible to deduce that a larger part of the population will be going through higher education, especially in the coming decades there will likely be an extension in demand for technology experts and programmers.
Although the effects of this within future years are also dependent on population change, economic theory predicts that wage rate will also fall along the number of jobs available in the labour market. But workers are also consumers. It only takes people to start anticipating lower wages for them to adjust their expenditure to fit long-term anticipation. And when the wage rate is low, consumption also becomes stagnant. If consumption contracts rapidly, then it is possible foresee two outcomes. The first is another wave of global stagnation, deflation or even recession. The second is the fall in price of many goods and services across the economy. On an economic model one would see this as an inward shift in aggregate demand, causing downward pressure on price and output to be less, whilst on the Phillips curve the economy would be at the point where inflation is low while unemployment high, creating a downward spiral in the economy. If there are doubts about whether companies would want to prevent the fall in price in order to protect their own profit, then it is important to bear in mind that corporation behavior also follows the principles of economics, and that no company can be persistently powerful without consumers.
During the early years of recovery from the Great Recession, 95% of US income growth was down to the top 1%. The idea is that the top percent of earners will become increasingly marginalized in the future while the gap between rich and poor broadens. Income inequality will probably increase before it can plateau or fall. If unemployment is high and income low, then it is easy to precipitate that tax revenues will be low too, and governments will have to cover its expenditure by increasing borrowing and public debt. An increasing portion of government debt will likely go to the top few percent of the population whose consumption and spending will be more immune to the changing economic patterns. But relying on the rich while income inequality is high is an unsustainable strategy, especially if we consider work disincentives and capital flight.
However, the outlook for the post-automation economy does not necessarily have to be bleak. The impact it has on the economy depends on the rate of automation. If the progress happens slowly, the economy can take time and adjust itself to an increasingly unemployed population. Many experts are suggesting a universal basic income guarantee in preparation for a new economic paradigm, and this is an idea that is currently being embraced by some countries. This basic income will be paid out of tax revenue, but where will this revenue come from? Some are pointing towards the top percent of earners in industrialised economies. The highest paid 3,000 people in the UK are currently contributing more income tax than the nine million poorest paid workers, according to government statistics. It is said that a moderate increase in tax rate, for example about 1% for the wealthiest households could exclusively produce substantial tax revenue, while still letting them bring home a majority of their income. Historically, the tax band for the wealthy has been raised in times of economic difficulty, like in the US during World War Two (77% in 1918) and following the Great Depression (when income tax was 63%). While how effective these policies may be in reality is still debated, the main idea is that provided scientists and policy-makers plan and think carefully about policies to safeguard the future, it is possible cushion the worst of a grave, not improbable economic downturn. Many people would still be vulnerable to the brunt of the economic cycle, as millions have been in the past. However, human nature is surprisingly resilient in times of greatest adversity. Once the storm in the markets has passed, there is reason to hope for a better future for humanity.
In a world where labour is fully automated, cost-effective machinery will be very efficient at generating consumer goods at a very low price. This is one of the reasons that led to some AI experts and futurists to believing that AI will usher in a period of abundance for humankind. Projecting forward to this future is perhaps difficult, much in the manner of how people two generations ago would have been astonished at the profusion of energy, goods and services people have today. Prices may seem much higher than the present term, but when adjusted for inflation, people’s purchasing power will have grown. Income inequality will likely to shrink eventually, as technology allows people to access more and more resources such as healthcare and education.
Others, like Google’s chief technologist Ray Kurzweil, are even more optimistic about their projections. In a post-work world, there would be so much materialistic abundance, and so little work that humans may be in danger of getting depressed and miserable. Without our work, where else could we look for a sense of contribution? It is interesting to imagine how people would lead their lives when there is little they should strive and work hard for. If one is stripped of the obligation to work and the need to pay bills, what is left? When life is not a road but an open plain, where does one go? Where else can one find a purpose in life? Yet there is also the hope that, with abundance and the fact that most people no longer have to work out of a necessity, people would be able to direct their energies to better use. There would be more time devoted to leisure and recreation, to understanding ourselves and our motivations as well as investing in relationships with others. The time may be spent travelling, writing a novel, teaching a yoga course, learning how to trim bonsai trees, meditating, engaging in more sports or embroidering among other things.
Regardless of what our experts and our own views and projections about the future may be, however, no one can know whether any assertion about the future can be right or wrong. The only way anyone can know the answer to the future is to wait for it to pass. But for the present moment, it would be a mistake to dismiss the prospect of automation and Artificial Intelligence entirely as a fantasy. Indeed there are people back in the 1950s who thought computers beating humans at chess would become a reality within only a decade, and these people are looked upon today as foolish, for they had believed the speed of technological change was faster than it really was. However, when you compare them to the people who thought that such a prospect would be impossible, it is easy to see who looked more foolish in hindsight. A discussion about Artificial Intelligence today may be an invitation for ridicule and dismissal for many. But understanding its implications and using logic to think decades ahead may be essential to what that future turns out to be. Like the philosopher Alain de Bottom said, we are now poised right before the tipping point of potentially one of the greatest technological revolutions in human history, which is why there is no time like the present to start building up our wisdom to control which way we will tip.