Alan Turing, a British mathematician known for deciphering the enigma machine, was the first to suggest the idea that if humans can use logic and reasoning to interpret information and solve problems, why can’t machines do the same? In pursuit of that idea, in 1950 he published a paper discussing how to build an intelligent machine and how to test their intelligence. And during that very same year Isaac Asimov published his novel ‘I, Robot’. But it was John McCarthy who first coined the term Artificial Intelligence (AI) in the subject’s first academic conference in 1956.
After many decades following that conference, technology has come to see that vision become a reality. And as an artist, curator or for that matter as someone who takes interest in the art world, Artificial Intelligence should be a subject of utmost importance to you in this current time. Besides technology growing faster than we can get used to, it is now showing prowess in surpassing human capacity for imagination. It’s probably thanks to science fiction movies that the thought of humans being pitted against AI equipped with an arsenal of robots and nuclear weapons doesn’t scare us as much as it used to. But currently it is AI’s intersection with the Arts that is hair raising for many people, as it impinges at what we define as our ‘humanness’.
The use of AI in art has sprung a lot of controversies. From Jason M. Allen winning The Colorado State Fair’s Fine Arts Competition with an artwork made using an AI image generator, to artists collectively protesting on Artstation because it allowed users post AI-generated artworks on their website. It is a topic that has many layers to its discussion and one cannot simply join the “AI bad” bandwagon and hope that things will work out. To understand what AI art entails for the future of art, several things need to be understood. Firstly, any topic pertaining to AI is vast in itself, and in this essay we cannot go over all of them. Not just in art, but the ethical implications of AI in nearly every domain of our life is a topic worthy of our attention. It has done many great things for humankind by providing simulation softwares for advance research, self-driving cars, etc. But do they outweigh its shortcomings? At least in art? Through this essay I intend to discuss and dissect the various aspects and implications of AI in art and also present my comments on this subject.
One thing that needs to be clarified before we begin to discuss the cream of the topic, is the difference between Machine Learning (ML) and AI, as it can become a source of confusion. Even though the terms are used interchangeable to refer to AI, AI is a much broader concept which can simulate human cognition and behaviour, whereas ML is a subset of algorithms that allow a machine to learn from data without being programmed explicitly. It is the ML algorithms which allow an AI to be trained in a certain manner. Meaning that AI is the brain, and ML is the hippocampus. But then why is ML used as a misnomer for AI? The answer to this question lies in how crucial ML is to AI’s foundation. Ever wondered how your favourite shopping app knew that you were thinking about buying a watch? Or your food delivery app suggesting you a scrumptious burger just when you were craving one? That right there, is ML in action. Many of those apps are powered by Google Ads which use ML to learn how you think, with the help of keywords encountered while using the app – the algorithms observe your choices to deliver ads accordingly. The same is true for social media apps, suggesting you posts that match the hashtags of those that you frequently view/like.
“What are the ethical and legal issues?”
First, we need to establish a perspective. Historically every other artform has served a purpose beyond aesthetics. For example, in a time before high speed or long-distance travel, subjects living beyond the capital or at far-flung edges of the land had seldom laid eyes on their kings/emperors. Besides being an element of the ruler’s grandiose ego, sculptures came as a proxy for their presence, and as a means of introduction. The same can also be said for murals that immortalized the image of a king on church walls, or carvings/paintings of mythology to narrate stories of Gods. Similarly, the same can also be said for printmaking which would have initially served the purpose of distributing information to the citizens. For example, posters with the image of a felon wanted for treason.
From this, one can gather that mediums of painting, printmaking, sculpture, etc are merely just instruments. One can paint their whole wall a flat colour, but the act of painting does not readily make it an artwork (whether anything is an artwork or not is completely subjective). The act of painting/sculpting/photographing/etc is merely an instrument in the process. Using the philosophers favourite object of discourse as an analogy, it can be said that if a chair was the art, painting (or any other medium) is the hammer or saw used to make that chair.
Similarly, AI is just a tool that can be employed to make art, which is an argument that many from the ‘for’ side make. And the argument by itself is very valid, but the issue is that AI’s capabilities exceed beyond just being a ‘tool’. When cameras were introduced they were mere devices subjected to the actions of our consciousness. Their capacities never exceeded beyond what we as humans could imagine. But bringing in a synthetic consciousness into the mix change the dynamics of the situation. It opens up a whole new avenue of artistic practice which has the potential to be both unethical and illegal. To understand what I mean, let’s get back to our previous analogy of a chair. A question was left unmentioned – what about the wood used to make that chair? If I were to make a chair from wood that I own then, it will not offend anyone. But if I were to steal that wood to make a chair then, I can bet that the guy who got robbed will be fuming through his ears. The data sourced for the creation of AI artworks is the main cause of concern, irrespective of whether we consider AI-image generators or AI models that mimic the special stylistic capabilities of highly-skilled individuals (AI memetic models). The annoyance stems from such image generator’s capacity for plagiarism. Generators like Midjourney, DALL – E 2, Stable Diffusion, etc. were trained with images/artworks without permission or the prior knowledge of the artist. Even before the question of its ethicality, it creates a huge legal issue of copyright. It is due to the same concerns that popular image platforms like Shutterstock and Getty Images banned AI generated images from their services, also why Adobe Firefly does not utilise ‘web scraping’ for image generation. It is also the reason why in January this year, three artists – Sarah Andersen, Kelly McKernan and Karla Ortiz filed a class action lawsuit against Stability AI and Midjourney and DeviantArt (with its own AI Art generator tool – DreamUp) arguing its legality in court.
The legal forum can be one way of approaching an amicable solution. Although it is not foolproof, and neither easy, but it is still better than rampant plagiarism in the name of art. According to experts in the field of law, current copyright (CR) laws and Intellectual property (IP) laws are outdated to properly deal with piracy carried out by AI, and that lawsuits are bound to fail. In order to overcome these issues, CR and IP laws need to be reworked so as to answer questions like – can an image created by an AI image generator be copyrighted? If so, then who amongst the user and creator of the software owns it?; What is the minimum amount of data one can source from another without raising an issue of copyright infringement?; Does having legal ownership of the source material give someone the legal claim on the output image as well.
Besides legality, the ethical concern rises with AI models’ ability for impersonation. AI-models tailored to individual artists will be immensely detrimental to the livelihood of prominent designers/ illustrators. Well reputed artists will, and have become targets for training AI models that can effortlessly replicate their stylistic capacities in any situation using any prompt. It makes one ponder as to what would happen when a well reputed artist who spent their blood and sweat to reach the echelon of success, takes a fatal blow to their social capital by getting replaced with a memetic model of their own? What would happen when any lay person could replicate their work with the help of a few key words which otherwise would have taken the artist days, months or even years to create?
A similar example of this was when an individual trained a memetic model of Kim Jung Gi mere 4 days after he passed away due to cardiac arrest. Kim Jung Gi’s creations were (and still are) beyond impressive. Lay people, fellow/professional/amateur artists looked at him gobsmacked as he drew effortlessly to create his masterpieces full of intricate details. But he was more than just his artwork, to his friends and family he was a jolly and fun-loving person. Now that he has gone, would it be ethically correct to use an AI model to create artworks in his ‘style’ without the consent of his family or fellow artists? Personally, I don’t think so.
I, and many others like me feel so because such AI models strike at the core of ‘individualness’. What makes it unethical is that it acts as an instrument of impersonation to serve as a proxy of someone dead/unavailable without their consent. The very element of dehumanisation of an individual to commodify their capabilities, is offensive to all who find AI art diabolical. An artist is not merely just his/her artworks.
Besides formal rules and regulations being placed to provide solutions, I feel that part of the responsibility is ours as well. What we as art practitioners/enthusiasts, can do is to reject the use of above-mentioned image generators and AI models of other artists, but more importantly encourage and appreciate artists who are working with AI by only using data collected by themselves, or data consented for use in their artworks. This way we can appeal to the morality of those artists who make productive use of AI. By presenting examples of good practices we can allusively create a moral code of practice which other artists will prefer to follow for the sake of their own reputation. As a result we will avoid following the path calling for outright rejection of AI, which has the high potential to backfire and indirectly encourage its morally unacceptable use.
Another issue less considered in this discussion is the environmental impact. To train and create large AI models a lot of processing power is employed, consequently huge amounts of electricity is consumed for processing and air conditioning. Besides the concern of huge carbon emissions, datacentres employ cooling towers to keep systems from overheating and as a result consume huge amounts of water. According to a study, an average data centre consumes a gallon (~3.7 litres) of water for every kilowatt-hour, all of which is freshwater. For instance, the training for ChatGPT-3 consumed nearly 700,000 litres of freshwater.
I feel that in the greater context it is less talked about because our normal internet usage by itself creates massive carbon emissions and consumes huge amounts of water. For instance, in the year 2021, Google’s U.S. datacentres alone consumed 12.7 billion litres of freshwater. And If measured from the year 2014, it amounts to 626 billion litres.
“Would creativity die out and be taken over by the algorithm? Could artists really lose their livelihood?”
The first question cannot be answered (or maybe not even in the next few years) with complete certainty, but by understanding the current situation, the trajectory of a future can be anticipated. For instance, new technologies have always posed a threat to the livelihood of those who had been satisfying its absence. Like how refrigerators replaced the Icemen or how cars replaced horse driven coaches. In July 1888 when Eastman’s ‘Kodak No.1’ became available for purchase, painters felt similarly threatened. However their fears were never realised because the camera by itself could not exceed the human mind.
It is crucial to understand that creativity does not reside in a medium or an instrument but in the human mind – whether the mind creates it or perceives it. The amount of variations and reductions of ideas/data from prior knowledge to create something new and valuable is creativity. The degree of absurdity (while maintaining the aesthetics in some cases) can also be considered as creativity, as it is regarded ‘out of the box thinking’. But the problem with creativity is, that it is expensive: requiring time and effort. From the perspective of an entrepreneur, manual creativity can be detrimental to a business model. An AI can make this task easier, it can come up with far more ideas in a few seconds than a human being can come up within a day, because unlike the human mind, the AI’s cognition is not clouded by emotions but run on pure logic of 1s and 0s. It can create lyrics on the fly, assess the likings of the demographic and create content accordingly, or create hundreds of layouts for your advertisement posters, and all this with barely any investment. As an entrepreneur, if you no longer had to pay your favourite designer for your company labels or illustrations to get the same quality of work (or maybe even better?) for next to no pay, you’ll not complain but instead celebrate this ‘creative’ tool.
It is human nature to be lured by what we like, and in its pursuit we as a species have often been happy enough to step over ethics to achieve it. In a market driven by productivity coupled with maximization of profit, it is not surprising that many corporations in future will choose AI irrespective of its ethicality. The idea is nothing farfetched, very recently I came across an advertisement on Instagram whose poster was made using an AI image generator – saving the company expenses of hiring an artist for the job. So it can be said with some confidence that ‘manual’ creativity may fall out in competition to AI’s creativity. But then again if it’s all about the market, AI’s sense of aesthetics and creativity may change only ever so slightly, being entirely driven by the wants of the market. Or maybe, with constant use it can usher in a new form of design and aesthetics.
It is no wonder that artists today feel threatened by AI. Cameras posed a similar threat once, and one may feel that just like nearly 100 years ago, artists will eventually find a way around and continue with their artistic endeavours to discover something new, and AI will simply become another tool in the shed. That may as well be true to some degree, but not entirely. Artists have been commodified by the capitalist market for a very long time now. But the major difference is that AI will not capitalize deceased artists, but the living ones instead.
“Has something been done yet?”
Currently ChatGPT is a hot topic in many circles. It doesn’t directly pertain to the arts but does point out the terrifying capacities of a synthetic consciousness in outshining a human consciousness. The rate at which AI technologies are developing is astounding. For example a few months ago I heard a person mention how animators are safe as AI cannot animate yet, and then only a few weeks ago Runaway introduced ‘text-to-video’ generation which creates a 3-second video clip based on a prompt.
On 22 March of this year, giants like Steve Wozniak, Elon Musk, and thousands more, called for a pause on AI research in an open letter. Back in 1996 after cloning became a reality with Dolly the Sheep, a similar pause led to the formulations of legal rules pertaining to genetic research which in hindsight were very prudent and made sure that no such genetic experimentations are carried out in humans. Similarly, proper legislation is needed to see that no AI system/model becomes disastrous to human kind in art or otherwise.
The Asilomar Ai Principles11 developed at the Beneficial AI 2017 conference, was one such step in the right direction. And on the bases of those principles, AI Art made by Image generators and AI-memetic models are in violation of principles 912, 1013, and 1114. But sadly they are still only principles and not proper legal rules exiting in any countries legislation. As of yet no substantial step has been taken and neither achieved.
As a species, our ability to foresee the future of technology has become weaker and weaker, and its psychological and physiological implications even harder to predict. For instance, Ken Olsen15 said in the 1970s, “There is no reason anyone would want a computer in their home”. With how fast AI research is growing, the image of the future is going to become less and less predictable with each passing day. Even though AI is very much in its infancy it has already twisted our vision of the future of arts in unexpected ways. AI (Primarily ML) has given us many gifts, but, then again, history is ripe with instances of great things being used in undesirable ways. In the end, it all falls on us as to how we use it, and what we promote. It is too soon to say anything with complete certainty, but this much is sure, that the playing field of art is about to change in inconceivable ways.
- A.M. Turing (1950) Computing Machinery and Intelligence. Mind 49: 433 – 460
- Turning Test (Originally known as Imitation Game)
- The Movie I, Robot (2004) starring Will Smith was based on this book.
- Dartmouth Summer Research Project on Artificial Intelligence (1956)
- A curved seahorse-shaped organ on the underside of each temporal lobe (Hippocampal Formation). It supports memory, learning, navigation and perception of space
- Adobe’s own Image Generator know to be very strict in regards to its data sourcing for creating images.
- Web Scraping is the process of using bots to extract content and data from web pages.
- Reid M.-Y., Jon K., Siddhartha S., Solon B., Ashton A., Mimetic Models: Ethical Implications of AI that Acts Like You. arXiv:2207.09394v1 [cs.AI] 19 Jul 2022
- Pengfei L., Jianyi Y., Mogammad A.I., Shaolei R., Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv:2207.09394v1 [cs.LG] 6 April 2023
- A list of principles pertaining to research issues, ethics and values and longer-term issues pertaining to AI’s creation.
- Responsibility: Designers and builders of advanced AI systems are stakeholders in the moral implications of their use, misuse, and actions, with a responsibility and opportunity to shape those implications.
- Value Alignment: Highly autonomous AI systems should be designed so that their goals and behaviours can be assured to align with human values throughout their operation.
- Human Values: AI systems should be designed and operated so as to be compatible with ideals of human dignity, rights, freedoms, and cultural diversity.
- Founder of Digital Equipment Corporation 1957