Analysis of the relationship between and development of artificial intelligence art and human art
produced by: Yiyan Lin
Recently there has been mounting interest in artificial intelligence (AI) ‘art&design’. In 2018, Christie's sold an AI painting entitled ’Portrait of Edmond de Belamy’, which was valued at $7,000-$10,000 at Christie's in New York. Several designers’ and artists’ theoretical arguments have raised doubts about this, implicitly challenging the value of AI art. It is, therefore, worth considering the notion of art&design in AI. Meyer (1999) claims that the value and meaning of art are independent on its creator, while Ernst (1999) suggests that the story behind the artist's art and information may influence how we view the work of art. However, these discussions have tended to focus on artists and designers, questioning or rejecting AI art rather than considering whether AI art can be a form of art in the long run. The aim of this essay is to evaluate the notion of AI art and how it can be accepted and popularised in the future, and also whether it can be called a new form of art and replace traditional art. It will first examine the current state of AI art with respect to designers and artists from a specifically traditional viewpoint of art and then discuss the relationship between and development of artificial intelligence art and human art.
As with the issue of AI and traditional art, there is a great deal of debate about what exactly art is. Anselm Kiefer is an influential contemporary artist. In 2011, he gave one year lecture at the Institut de France, which started by discussing what art is. ‘What is Art? Nobody knows,’ he told a crowd of art experts, philosophers and writers. ‘And when you try to define it with words, it disappears.’ It was an unusually honest answer. From a material point of view, as Ernst Gombrich indicates in The story of art (1999), ‘There really is no such things as art. There are only artists’. He gave the following example. Someone likes a painting, not because of the art itself, but because of the artist. There is nothing wrong with people liking a painting only because it reminded them of their childhood. There is no wrong ‘like’, only wrong ‘do not like’. The ‘wrong disliking’ is letting prejudice get the best of the people and losing the opportunity to appreciate art. In this book, he also expresses an opinion on how to appreciate art. This view is expanded from his premise about ’what is art’. Since art is a kind of beauty, the appreciation of art should not focus on what 'art studies' or 'art history' are concerned with but there should be appreciation of a kind of beauty, and people should seek resonance from works of art. The concept of art is the arbitrary use of a frame of mind to cover all the works of all artists. If we assume that art belongs to the unknown and is a means to discover and to use discovery as a new achievement, and thus to make man more expansive, in short, to belong to the new world, then an artist should be identified as ‘a man who continually discovers new worlds with intelligence, intellect, and wisdom’. Artwork is a collection of the artist's spirit, intelligence, aesthetic feeling and wisdom, all of which produces an actual physical appearance. It is a direct result of the artist's thinking. The artist's neural connections dig deep into what people would not expect and have a new, immediate impact on the audience. Moreover, when the audience sees an excellent work of art, the impact is a rewiring of the brain's understanding of what is there. Therefore, if AI creates art that feels beautiful and resonates with it, does this mean that what AI creates is art?
In addition to reflecting reality, art can also stimulate certain emotions in the audience, such as excitement, sadness, meditation and depression. The key to creating real art instead of just ‘drawing a picture’ or ‘writing an article’ is that AI needs to understand the physical basis of these emotions. Emotion is a response to the release of chemicals in the body and the state of the nervous system. Therefore, if AI needs to create art, it needs to have a real understanding of the objective function of the art creation process and gather sufficient data. If a chip is implanted in the human body, it has the basis for collecting the corresponding endocrine and nervous systems’ state of emotion and then it can obtain such real data. Moreover, with that kind of real data, it might not be a problem at all to use machine learning to create more compelling work. Before discussing whether AI can create art, understanding the nature of AI is the foundation of computational arts. AI is a new technology science that is used to research and develop the theory, method, technology and application system for simulating and extending human intelligence. AI is a branch of computer science that seeks to understand the nature of intelligence and produce a new kind of intelligent machine that responds in a similar way to a human. The research in this field includes robot, language recognition, image recognition, natural language processing and expert systems. Since the birth of AI, theory and technology has become increasingly mature, and the field of application is also expanding, but there is no unified definition. AI is the simulation of the information processing of human consciousness and thinking. It is not human intelligence, but it can think like a human being, or it can surpass human intelligence. However, this kind of advanced, self-thinking AI still needs breakthroughs in science, theory and engineering. In the purest sense, the essence of AI is algorithms. 'Digitalisation changes how we know, and we do not know enough about it’. As Bunz (2014) states, ‘That algorithms have learnt how to write, for example, was not expected' (Bunz.2014). Even if we fully understand the nature of AI, it can achieve much more than what is socially expected.
In 2017 Gaetan Hadjeres and Francois Pachet of the Sony Computer Science Laboratory in Paris trained their AI system to reproduce of Johann Sebastian Bach. These choral songs were chosen because, as the MIT Tech Review (2017) states, ‘it was generated in a step-by-step, algorithmic way’. The AI system was trained to use 352 Bach songs, which were converted into different notes, for a total of 2,503 works. More than half of the 1,600 listeners (including professional musicians and music students) thought the work was Bach's. At a contest held by Dartmouth College(2016), the judges were given the task of reading 14 lines of verse with the required escort sentences. Humans wrote some; AI systems wrote others. All the judges could identify which were written by a human and which were written by an AI system. There was a time when machine learning enthusiasts used deep learning algorithms to train AI systems to write novels that were not very effective. Human beings are born with rational thinking and mathematical abilities; we still need education to realise these abilities fully. However, computer systems are entirely different. Their algorithms perform well on tasks, but they lack most of the abilities that humans are born with. They can do complicated mathematics in milliseconds, but nothing happens when someone tells a joke to a computer or plays a love song. Therefore, AI may have a long way to go before it can create its own art. However, on the bright side, AI can begin by understanding how humans think creatively through their creations first. In areas of AI not directly related to creativity, designers and artists are already doing well. Algorithms have completed some of the most challenging logic problems ever devised by humans. Alphago, for example, beat human players in chess. Language, in particular, is very logical, and so AI allows artistic pursuits, such as storytelling and poetry. Let us move on to the visual arts. Aaron first appeared in 1973 as an AI system created by artist Harold Cohen. In an interview with the BBC, Cohen reports ‘Aaron has become self-sufficient’, but he adds that Aaron is still a long way from being a genuinely human-like creation: ‘I don't deny that at some point in the future, a machine could make something more complex, something far more complicated than self-driving, but I don't expect that to happen in this century’.
At present, AI intervening in artistic creation has already become the new trend. However, important questions remain. How do we feel about art made by intelligent machines? Will we appreciate its creativity and design? Google's ‘DeepDream’ project, launched in 2015, allows a machine to demonstrate an interpretation of art. Autodraw is a stick figure drawing tool; a few strokes are drawn and an algorithm can identify how people want to draw the pattern, and it is optimised for the user. This is how humans learn the content and effects of machine learning. Autodraw uses deep-learning techniques to identify images and then presents the ones it interprets. However, for some reason, the pictures that the auto draw outputs are full of dogs’ face and eyes, and strange spiral patterns, with a kind of strange magic. Autodraw can achieve more than recognise what users are drawing. It can even help the user fill in unfinished doodles and correct them; if a user draws a three-eyed cat, Autodraw will remove an eye. It means that Autodraw already has what we call abstract thinking, which is not merely organising the lines of a drawing according to historical data, but rather the concept of ‘knowing’ the eye and knowing that a cat has only two eyes. Behind Autodraw is the AI system known as SketchRNN. SketchRNN will note the shape and order of each of our strokes and train a neural network for each particular object (cat or chair, etc). If we take the strokes of human graffiti as the input, it can encode them sequentially, and train the neural network in the way people paint. After completing this exercise, SketchRNN learns the ‘General Rules’ of drawing a pattern, such as having a round face, two pointed ears, two eyes and six whiskers for a cat. SketchRNN will know that a big circle with two small circles, six lines and two sharp corners add up to a cat.
Then, variables can be introduced into it so that SketchRNN can provide an output and accept randomness. For example, the round face of a cat can be less round; the cat's beard can be uneven in length. However, the bug is out of bounds for the variable float and will be fixed with SketchRNN. The reason SketchRNN can identify and even create graphics is that the team has trained a neural network for each pattern. However, with so many resources at its disposal, SketchRNN's application is limited to doodles and, if confined to the ‘stroke’ logic, it may be able to create something using calligraphy. Autodraw(2017) held an exhibition in which six pieces were snapped up by a single collector for up to $8,000 each. It may be the first piece of AI to make money from art. Do collectors collect them as works of art? Can the collectors feel the beauty in the artwork (picture)? Alternatively, is it merely that these works are rare, exotic, collectable and of economic value?
In October 2018, Alibaba's intelligent AI system, ‘Lu Ban’, was put on the job, making 8,000 posters per second, driving the art world, especially the design world, crazy. Some designers claimed that it is inevitable that AI will replace human designers; the first to be replaced will be the drafting process. The more technical a piece of work is, the more likely it is to be replaced; the more creative it is, the more likely it is to be replaced over a relatively long period. However, some in the industry believe that AI will not replace designers in the future, but will help designers solve the problem of repetitive work, and so creative designers will only become more valuable. In July 2018, Wrike and EdgePeak Consulting surveyed 1,552 ‘creative professionals’ who work in various fields, including design, art, film, video production and advertising. The researchers found that 85% of respondents believed that AI would have a significant impact on their creative output in the next two years. No other emerging technology was expected to have an impact this big. Some designers think the posters are not valuable or original, but just 8,000 permutations and combinations based on algorithms. Most of the posters are just variations in colour and format, not styles. However, for Alibaba, the posters satisfy their need to appeal to consumers. ‘Luban’ is the right combination of business and technology. First of all, in terms of technical depth, it has a set of high threshold systems and, in terms of business, it can really through ‘intelligent’ and ‘personalised’ in order to achieve maximum business value, a subversion of traditional ways. It is not art or design. Today, creativity is still a human activity. While many AI systems are trying to create works of art, they still do not seem to measure up to artists’ standards. A good designer or artist does not just sit in front of a computer and draw mechanically. He or she is good at observing and photographing whatever he or she takes for granted. They also do not allow thinking to stop as they leave work, and then inadvertently reap that Eureka moment on the road, in the shower or before bed. AI may never be able to achieve this.
In conclusion, art is a kind of creation and a kind of conspiracy, and it requires the participation and trust of the appreciator. According to the artistic intention, the creation is endowed with meaning. Furthermore, at the same time, as the receiver of art, people interact with works of art with a beautiful view formed by the internal and external environment, thus leading to an exchange of concepts, information and thoughts, which ensures the interaction of art creation and the effectiveness of the works of art. I think that, while AI may not be an independent artist for a while, it has become an increasingly integral part of people's lives, including but not limited to artistic creation and aesthetics. The AI art industry in the future will be very optimistic, and will bring about a massive disruption to the traditional art industry. I feel the future of the art world is AI art, and human artists will co-exist and work together in the world. With regard to my artwork, AI art and human art should be complementary to each other, and should learn from each other in order to grow. Therefore, my artwork is a piece of music that AI and humans are playing together, and I am going to use a computer keyboard to play pieces of mixed piano music, every time I play a piece, the machine leaning's system will generate new music based on what I am playing. Then I will post-process this into new music, which is a display of AI art and human art coexisting. I will be optimistic about the future of AI art development.
References
Bunz, M. (2014). The silent revolution. London: Palgrave Pivot, pp.1-24.
eMarketer. (2019). Will AI change how people approach creative work? [online] Available at: https://www.emarketer.com/content/will-ai-change-how-people-approach-creative-work?ecid=NL1007 [Accessed 1 Aug. 2019].
Gombrich, E. (1999). The story of art. Phaidon Press Ltd.
Meyer E, et al. (1999).Biochemistry. ACS Publications.
Angus, B. (n.d.). Intelligence and metadrama in the early modern theatre.
News.99ys.com. (2019). Artificial intelligence has made a breakthrough in the arts. [online] Available at:
http://news.99ys.com/news/2017/0627/10_209139_1.shtml [Accessed 2 Aug. 2019].
Knight, W. (2019). There’s a big problem with AI: even its creators can’t explain how it works. [online] MIT Technology Review. Available at:
https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/ [Accessed 2 Aug. 2019].
BBC News. (2019). Michael Cohen - BBC News. [online] Available at:
https://www.bbc.co.uk/news/topics/cyykd0lnl98t/michael-cohen [Accessed 2 Aug. 2019].
Schapiro, M., Schapiro, L. and Esterman, D. (2000). World view in painting——art and society. New York: George Braziller.
Figure 1
https://en.wikipedia.org/
https://upload.wikimedia.org/wikipedia/commons/c/c1/Edmond_de_Belamy.png?
Figure 2
https://gizmodo.com/someone-finally-turned-googles-deepdream-code-into-a-si-1719461004
Figure 3
https://ai.googleblog.com/2017/04/teaching-machines-to-draw.html
Figure 4
http://256wz.com/topics/17989.html