“At all events my own essays and dissertations about love and its endless pain and perpetual pleasure will be known and understood by all of you who read this and talk or sing or chant about it to your worried friends or nervous enemies. Love is the question and the subject of this essay. We will commence with a question: does steak love lettuce? This question is implacably hard and inevitably difficult to answer. Here is a question: does an electron love a proton, or does it love a neutron? Here is a question: does a man love a woman or, to be specific and to be precise, does Bill love Diane? The interesting and critical response to this question is: no! He is obsessed and infatuated with her. He is loony and crazy about her. That is not the love of steak and lettuce, of electron and proton and neutron. This dissertation will show that the love of a man and a woman is not the love of steak and lettuce. Love is interesting to me and fascinating to you but it is painful to Bill and Diane. That is love!”.
The above is an excerpt from the book ‘The Policeman’s Beard Was Half Constructed’, which was touted as ‘the first book ever entirely written by an artificial intelligence (AI)’. This book was published in 1984 and the author of it is an AI computer program called ‘Racter’, which is a foreshortening for Raconteur. William Chamberlain and Thomas Etter are the programmers who wrote Racter’s program. Racter was fed with vocabulary and English grammar rules, and is capable of creating poems and short prose randomly and independently. The book created by Racter is not pre-programmed. Racter, in fact, creates it on its own, based on the vast array of vocabulary it discovers in its files. The words in Racter’s files are classified in a particular way of what Chamberlain refers to as a ‘syntax directive’, which tells Racter how to utilize the words to construct a sentence.
Apart from Racter, there are also other AI systems that are able to generate literary works. More recently, in September 2020, The Guardian published an article written by an AI system, Generative Pre-trained Transformer-3 (GPT-3), dubbed ‘A Robot Wrote This Entire Article. Are You Scared Yet, Human?”. GPT-3 is a sophisticated language generator of OpenAI that can create human like texts quickly with minimal human input by using deep learning. GPT-3 has been trained on an extremely huge amount of text data, virtually all available data on the internet. Being the largest neural network with 175 billion parameters, GPT-3 is a hundred times bigger than its brother GPT-2. GPT-3 processes approximately 45 billion times the quantity of words that an average human perceives in a lifetime. GPT-3 requires a prompt to start writing. It is then able to recognize and repeat word patterns to anticipate the following words after the initial prompt. To generate the aforementioned article, The Guardian’s staff provided GPT-3 with instructions and some introductory lines, which then enables it to produce eight different articles. The Guardian’s editor then selected parts and merged the eight articles into a single piece by deleting lines, paragraphs and rearranging orders. The editor even went on saying that “editing GPT-3’s op-ed was no different to editing a human op-ed…. it took less time to edit than many human op-eds”.
Apart from literary works, AI systems have also been involved in the creation of artistic works. In the 1970s, Harold Cohen, an artist and an art professor, created an AI program called ‘AARON’. AARON has the ability to create art works autonomously, based on the ‘teachings’ of Cohen. Throughout the years, Cohen has worked to improve AARON’s coding and its artistic knowledge like forms and colors. In 1995, AARON successfully created its first color image by drawing its form and coloring it. AARON’s works have been exhibited in leading art galleries and museums across the world. Furthermore, in 2018, an AI-generated art piece named ‘Edmond de Belamy’, was sold for an incredible amount of USD 432,500 at an auction. This artwork was generated by an AI system called GAN (Generative Adversarial Network), which was created by Obvious, an art collective based in Paris. Another prominent instance would be ‘The Next Rembrandt’, which is neither a recently discovered work of the famous Dutch artist, Rembrandt Harmenszoon van Rijn, nor is it a knockoff. It is an AI-generated masterpiece based on the AI’s analysis of the aesthetic components and the styles of Rembrandt’s works.
As discussed above, we already have machines that can generate creative works that include literary, musical and artistic works. This demonstrates that the production of automated content is becoming a commercial reality. Historically, the act of creating is equated with a human being. The advancements in AI have, however, called this notion into question. AI has become ubiquitous in today’s world and is actively involved in various sectors, but for the purpose of this essay, we will only focus on creative works generated by AI systems. The rise of this new breed of creators has raised issues in copyright law. The development of AI in the realm of creative works not only changes the way in which creative works are generated but also poses new challenges for the copyright regime. This has prompted legislators, scholars and legal practitioners to reconsider the fundamental concepts in copyright such as authorship, creativity and originality as well as the rationales that underpin copyright protection in the first place.
The thesis of this paper is as follows: are artificial intelligence systems really creative in the context of copyright?
There is no agreed-upon definition of what constitutes AI, but it refers to machines with human-like behavior and is the ability of a digital computer to perform tasks that entails intelligence when completed by humans. Nonetheless, intelligence is beyond a monolithic concept, because there is no definitive standard to measure human intelligence. Human intelligence, according to mainstream psychology, is considered as a collection of distinct components rather than just one single trait. However, it has been agreed by the majority of psychologists that creativity is one of the components of human intelligence. Racter, as well as other AI writer and artist described in the introduction, have demonstrated that AI systems is capable of generating works that are ostensibly creative. This assertion is, however, predicated on the premise that the aforesaid AIs possess one of the aspects of human intelligence, namely, creativity.
Next part of the essay will touch upon creativity within the context of copyright law. It will then move on to discuss how both humans and AI systems generate creative works. Accordingly, this section will reach the conclusion that the functioning of AI is not the same as human creativity due to its mechanical nature. As a result, AI-generated works, which do not involve direct human participation, cannot be protected by copyright. It should be noted that though the author believes that AI-generated works are different from those created by humans, the author does not negate the importance of AI-generated work and its societal, cultural and scientific value.
Copyright protection’s focus is on creative works as it was primarily designed to safeguard the fruits of human’s creative endeavors. Thus, it is vital to understand creativity within the context of copyright. In the author’s opinion, creativity is a fundamental issue that determines whether an AI-generated work can enjoy copyright protection alongside creative works produced by humans. There is, however, a scarcity of literature that addresses creativity in relation to copyright law. Creativity is a general concept rather than a legal one. Creativity can be defined as the ability to produce new, valuable and useful ideas or artefacts. From the perspective of computational system, for an AI to be deemed creative, it must strive to generate new solutions that are not carbon copies of the AI’s previously known solutions; and those solutions must be appropriate for the task at hand. Furthermore, within the context of AI, creativity is a question of degree. For instance, Pereira claims that an AI is deemed to be highly creative if it has the capacity to deal with various abstraction levels for the same problem; the ability to assess and criticize its own productions; and the ability to function in more than one domain without having to reprogram.
Albeit having the ability to carry out various functions, it should be reminded that AI system is limited by the functions human creators programmed into it. It is doubtful that the outstanding AI systems that produced the aforesaid ‘Edmond de Belamy’ and ‘The Next Rembrandt’ can ever learn how to create poetry or music on its own as its program codes and data do not encompass such functions. Similarly, Racter, which wrote the excerpt in the introduction, can never start painting without being reprogrammed to do so. Therefore, an AI system cannot go beyond its programmed functions to perform tasks that are not covered by its codes.
Furthermore, AI systems can only produce outputs based on the data that was loaded to it by humans. This indicates that AI’s creation of works will be dependent on existing works or data. The aforesaid portrait ‘Edmond de Belamy’, for example, was produced by an AI system based on 15,000 portraits from WikiArt, an online art encyclopedia, that were fed to it by its programmers. Additionally, in the case of ‘The Next Rembrandt’, the AI program was able to create the artwork in the style of Rembrandt, over three centuries after his death, only because it analyzed hundreds of Rembrandt’s works, focusing on the general characteristics and aesthetic elements. Following a thorough analysis, the AI established the following requirements for generating an artwork in the style of Rembrandt: a portrait of a male Caucasian in his 30s, with facial hair, wearing dark clothing, a collar and a hat with his head facing to the right. The AI and its developers also paid attention to the facial features of the portraits, in particular, the AI analyzed the eyes, nose, mouth and ears of the portraits. In addition, analysis of the face proportions such as the distance between the eyes, nose, mouth and ears was performed by the AI system. Hence, the output of an AI will depend on the information and data it was exposed to. This indicates that AI system itself is not capable of both thinking and inventing. Its ‘creativity’ is the results of its analysis of vast amount of data (human-created works) it was given to. It is incapable of generating an output where it has no relevant data about it. Hence, theoretically, AI is only able to produce output akin to works created by humans.
As discussed above, works generated by AI are determined by its programmed functions and the data it was fed by its human creators. Its functions cannot deviate from its program code and it can never generate works without having access to data related to its output goals. For instance, an AI designed to produce images cannot begin to compose music without much human intervention as the AI system would require human programmers to alter or construct new program codes for it, as well as loading it with music-related data, all of which would require a significant amount of human effort. As opposed to AI systems, humans can create and invent independently without the need of any sample but only inspiration. In the case of humans, internal stimulation can prompt a person to select another area of creativity impulsively. As an example, a human writer can start creating paintings and a songwriter can write autobiographical works without any special training. Therefore, despite the fact that AI has developed to the point that it can produce works that are difficult to tell apart from those created by humans and continues to evolve and become more complex in terms of its functions and the mimic of the human brain operation, its process of creating works is entirely mechanical. With all due respect, AI creativity is not the same as human creativity.
AI systems lack consciousness, subconsciousness or the usual mental states like emotions, beliefs and desires that are immanent in us humans. Albeit artificial neural networks imitate the operation of human brains, it is dubious that they will ever learn to have feelings, sensations, emotions, sentiments and consciousness. Computers, unlike humans, have no need to express themselves as they merely execute commands. Contrarily, each creative work generated by humans comprises the unique mental, spiritual and emotional contribution of its author; consequently, a work mirrors the personality of its author. This implies that feelings, aspirations, inspirations, experiences and emotions are incorporated within human creativity.
Moreover, AI lacks one of the key components of creativity, namely imagination. Taking the aforementioned AARON as an illustration, it needs to be fed with the relevant knowledge it depicts in the works it produced. That is done via a generative system where AARON was fed with a set of abstract rules on the human body including the fact that humans have a pair of legs and arms; as well how body parts appear from different angles, such as just having one arm visible at certain points. Thence, AARON is able to draw a man with only one arm can be seen (the other arm being obscured by someone or something else), but is not capable of drawing a one-armed man. This shows AARON’s inability to imagine things that it has never seen before. Similarly, AI that generate literary work like BRUTUS, a storytelling machine, lacks consciousness and experience derived from senses such as touch, which would have allowed it to better immerse itself into the lives of the characters it creates if it had it. Thus, AI does not have the ability to imagine or fantasize like humans do, which some consider to be a key distinction between machine and human creativity.
Additionally, it has been argued that humans create to meet needs, either to fulfil the author’s own needs for self-expression or to satisfy other people’s needs for cultural, aesthetic and spiritual development. However, it can be argued that AI systems exist and create to satisfy human needs due to the fact that they do not have consciousness, emotions or attitude towards the works they produced and merely follows human instructions. In other words, AI is just a tool in the hands of humans. Humans control and set the goals for the type, form and style of works the AI should aim to generate; the AI then use machine-learning algorithmic calculations to choose from a vast amount of pre-fed data and input to produce an output that is similar to the human-created works contained in its database. The activities of the AI system are completely mechanical and operates on fundamentally different principles. Thus, it is fair to suggest that the nature of the AI’s activities, which create works that are akin to human-generated and copyright protected works, is not the same as human creativity. It can even be further argued that AI’s functioning does not even resemble creativity because such activities are not based on what stimulates humans to create. It therefore follows that works generated in this manner cannot be regarded as the products of creativity, despite those works are of tremendous societal worth. AI-generated works cannot and should not be equated to those created by humans as they are fundamentally different in nature.
Hence, the arguments presented above show that AI’s ability to generate output that has the same expression as works created by humans is not a form of creativity. As demonstrated previously, the reason that AI’s activities resemble human creativity is that it analyses a vast amount of data and synthesizes the findings, which is an imitation of the human brain’s logical reasoning. AI systems do not have consciousness, subconsciousness, emotions, beliefs, desires, imaginations and other mental states inherent in humans. Due to the lack of these ‘prerogative’ mental states innate in humans, AI systems do not seek to express or convey message of any sort through their work. A lot of people have misperceived the functioning of AI system as creativity, when what it actually does is strictly subordinate to the commands of its human creators and strive to achieve the aim set by them by analyzing the data made available to it and produce an output in style of the human-created work contained in its database. An AI does not generate works to satisfy its own needs for self-expression nor to fulfil its aspirations (because it does not have any), but to meet the needs of humans. Hence, the functioning of AI cannot not be considered as creativity. Consequently, as copyright primarily protects the fruits of creative endeavors, it can be argued that AI activities that lack creativity supports the argument against copyright protection for AI-generated works where no direct human intervention is involved. Nonetheless, this does not negate the societal worth AI-generated works have brought us, as well as their significance and greatness.