In this essay I shall look at ideas of creativity, referencing speakers and debate at recent conferences on Artificial Intelligence and Cognition at the Leverhulme Centre for Future Intelligence, Cambridge, and some suggested new models of value judgement for creative outputs. I will also explore the ways in which breaking boundaries and rules is a fundamental human expression of creativity, as expressed in my final sculptural piece entitled “Heavens Embroidered Cloths”, wondering if this rule breaking concept could also be the key to machine creativity.
Notions of creativity and what is or is not creative is hotly debated in Artificial Intelligence (A.I.) and philosophical circles. Can animals or machines be creative? Is it just programming, genetics or learned behaviour? Perhaps this obsession with what is or is not creative, is because as humans, our creativity is often seen as a magical ingredient that cannot be replicated by a machine. When perceived as divine inspiration or something unique and special, it gives us a sense of superiority within the world, and without it our sense of self might crumble.
The engineering of a specific A.I. co-creator for this project was not possible for a variety of factors, (including skillset, time and financial limitations) so I decided to use the open source algorithm Deep Art. DeepArt (2017) is an algorithm developed by Leon Gatys, Alexander Ecker and Matthias Bethge and was developed as a response to the ability of humans to “create unique visual experiences through composing a complex interplay between the content and style of an image” (Gatys, Ecker and Bethge 2015). At the time of Deep Arts programming, the algorithmic basis of this interplay process was unknown and there existed no artificial system with similar capabilities. DeepArts algorithm uses two pictures as its creative data input, extrapolating Style from one and blending it via a Deep Neural Networking visual imaging system with the Content from another.
No claims of autonomous creativity are made by DeepArts designers, yet it still makes decisions that were not directly programmed into it and offers output that is both visually appealing and surprising to the human eye. I had no idea what image I might receive back from DeepArt, in contrast to using a tool or filter in Adobe Photoshop or Instagram for example.
Neural networking programs which use style as a building block of creativity are used in many disciplines of computational creativity, particularly in music, as described by Colton and Wiggins (2012) but also increasingly in the visual arts. Boden (1992) states that random processes generally only produce curiosities, not radical surprises. An algorithm that produces work solely in a recognisable style would appear to bear this out, creating work that is new, but also very similar to that which has gone before, rather than a completely innovative genre.
The images made by the Deep Art algorithm have been a fundamental part of my projects creative process, but does the production of this image mean DeepArt itself is creative? Is the 2D sketched design the demonstration of an ultimate creativity or does the creativity reside in the idea of combining a flat pattern with an illustration to feed into Deepart in the first place? Maybe the creativity resides within the skills needed to translate the image into a 3D garment?
We already have many examples of creativity which are not easily compared like for like, Prosecco (2014) points out this makes creativity very difficult to define in formal terms, so computational creativity researchers tend to focus on aspects of behaviour in humans and computers that produce new or surprising outputs, yielding unexpected value. Obviously in order to decide if an output has value, we also need to define what we will find valuable in that context. Artistic value is notoriously subjective, as the vagaries of the art market will attest.
Value of different levels seems also often to be assigned to types of creativity. At the CFI conference for Art, Science and Mind, Shevlin (2018) suggested separating conscious and unconscious creativity into two areas, borrowing business terms to describe a top-down approach - which involved an effortful search for solutions using inventiveness and ingenuity, versus a bottom-up solution - in which a flash of inspiration appears to solve an issue, akin to the concept of a muse.
Generally, in fashion and costume design, the idea of the creative “genius” whose flashes of “inspiration” leads immediately to solo production of an exquisite work of art is a fabrication. It takes an army of creative people to breathe life into a design, each searching for creative solutions in their particular area of expertise. What best thread to use to attach that particular crystal in such a way that it catches the light to maximum effect? How best to sculpt the collar using patterns and interfacing, what is the tacit knowledge and creative new methods used to achieve a particular shape? These are the creative skills of crafts people, and yet the Designer is often perceived as having something extra, a special insight that has more Value (both creatively and financially) than the others in the team. Therefore, is pattern cutting actually creative, or is it a tool, in the same way that grammar is a tool for literature, as suggested by Boden (1992), simply a technique to release the designer’s bigger idea? Alongside the physical problem-solving creativity of the construction team, the Designer has been incubating ideas collated from other places, thinking and exploring concepts within their mind, pulling memories, emotionally analysing and adjusting ideas consciously and unconsciously before ever putting creative pen to paper in a flash of “inspiration”, but is this creativity really different or more valuable than the steady creative learning progress and problem solving of the craftsmen?
Junaidy, Nagai and Ihsan (2013) believe there something different going on in the mind of a designer to that of a craftsman, they conducted experiments using a think-aloud protocol by analysing verbalized thoughts through a method based on associative concept analysis. Interestingly, craftsmen and designers did appear to activate different cognitive processes, explaining their different concerns with aspects of creative production, designers with intangible issues, such as culture and context and craftsmen with tangible ones such as proportion and shape. But this does not prove a designer’s creativity is rarer or more valuable than that of a craftsman.
This notion of Value can be seen in attitudes towards the monetary value of artistic works, Anscomb (2018) noted the huge descent in price of a painting when it is discovered not to be by the original High Value Author, for example the detection of mis-attribution of an old master or discovery of a forgery. Obviously, the painting itself has in no way changed, it is our attitudes towards the new information that reveal our value prejudices. It can be seen therefore that the origin of a piece, it’s story, it’s authorship, process of making and history gives an emotional connection that enhances its perceived creativity and adds Value.
With regards to how this may affect attitudes towards the original artistic work of an A.I., the assumption that a machine cannot be creative, as it lacks intuition, emotion, memories or story; immediately calls into question the Value of a work. Some pieces may be initially coveted with high Value as novelties, the computer algorithm/painter Aaron (The Harold Cohen Trust 2018) possibly falls into this category, but potentially computational-creativity could be viewed by many purchasers of art as simply factory production.
Partly I believe this may be due to the speed at which a computer can work. Choosing which images to put through deep arts algorithms took me several days, but Deepart was able to process them through its neural network in minutes and deliver an original, new work, apparently without effort. Maybe then it is the idea of effort that is linked to creativity. The rumination, conscious and unconscious as a human mind develops a new idea. Effort is needed to understand parameters, rules or frameworks and subsequently more effort needs to be employed to consciously reject them.
In order to knowingly break the rules, we first need to know what those rules or algorithms are. These could be as banal as a schoolgirl raising the hem of their uniform skirt an inch to assert their individuality from the rules of the institution, and in doing so present a “new look” to the uniform of that place. Within this scenario we can look at what Boden (1992) describes P and H creativity. P describes a creative thought or action that is new only to the person who has it, Psychological Creativity. In our school scenario, this may be a rebellion against the rules that is as old as time, but if it occurs to the student as a new thought, with no historic reason for thinking that way, it is P creativity. It could be argued that much of teenage rebellion and chafing against rules is based around P creativity. However, if it is an idea that has never occurred to anyone in the whole history of the world, perhaps a student demonstrated a new holographic device that enabled them to appear to be wearing a long skirt whilst actually wearing a mini, through a new process of light projection they alone had invented in the school labs, that could be considered H creativity, Historic Creativity.
If moving beyond rules is a good way to measure the creative process, as has been suggested by Boden (1992) is it possible for a machine to do this? Surely if it breaks rules in order to be more creative it will only be because it has been programmed to do so and is therefore not breaking them at all. Or should it be considered that breaking rules isn’t actually an assistance to creativity for a machine in the way that it can be for a human, “The program IS the rules, and regardless of where they came from, it is those rules – that is, the program – that generates material I could never have imagined or generated myself.” Cohen (2010)
Marta Halina (2018) describes how in its revolutionary winning Go game against world champion Lee Sedol in 2016, the A.I. called Alpha Go made moves that experts in the game repeatedly described as intuitive, irrational and creative. Alpha Go is one of the Google DeepMind A.I. programs, created to explore neural network learning algorithms and part of a family of narrowly focused A.I. designed to perform outstandingly in one particular area. After mastering Chess, a rules-based game, by drawing on a vast knowledge base of possible known moves and consequences, the team decided Go would provide a different challenge. It is a game often described as an artform, winning strategies and moves often cannot be described by logic but because they “feel right”. Halina(2018) argues that despite its surprising winning moves AlphaGo is not creative, she states that rather than drawing on a deeper understanding of the game, it’s success was due to the fact it had played the game against itself many more times while learning (due to computing speed), completing far more games than its human opponent could hope to in a lifetime and therefore had greater experience of the task. She suggests however that utilising this massive extra learning experience of an A.I. system could lead humans to greater creativity in problem solving.
While Art is not necessarily a problem to be solved, I found that the swift returns of DeepArts contribution to the work, and the ability to quickly work through variations of the input data advantaged my decision-making process, enabling me to make different final aesthetic decisions than I perhaps would have if I had been drawing and imagining the designs alone.
In my piece “heavens embroidered cloths” I was inspired by the life of Evelyn Boyd Granville, the second African American woman to receive a Ph.D. in mathematics (in 1949 from Yale). Evelyn worked on the Apollo space program in the early days of digital computation and had first-hand experience of the rules, both unwritten social norms and the laws in place at the time, that attempted to exert control through prejudice of race and gender.
I took images from a vintage astronomy book as a starting point for my inspirations of Style to share with the Deep Art algorithm, as Evelyn was fascinated by astronomy from a young age and it also reflected her later work with NASA. The concept of unexplored space, other worlds, virgin territory without earthly boundaries or regulations is infinitely fascinating in contemporary popular culture. The recent film Annihilation, directed by Alex Garland (2018) and based on the 2014 book of the same name by Jeff VanderMeer, features a female cast playing a four strong team (a biologist, an anthropologist, a psychologist, and a surveyor) who set out into an area known as Area X. This story explores similar territory to “Roadside Picnic”, a short story of 1972 by brothers Arkady and Boris Strugatsky, (which was also made into a film many decades earlier by Andrei Tarkovsky in 1971 entitled “Stalker”). All these stories focus on an “unnatural” place where our established laws of physics and chemistry might not apply. Physical logic and reason are out of kilter, and the sending of an all-female team into an unsafe, unknown area in “Annihilation” immediately contravenes traditional narratives.
I wanted to explore the illogical parts of DeepArts designs as relating to traditional dressmaking, seeing if the asymmetry, unrealistic fabric folds and non-Euclidean geometry could be translated accurately into the finished sculpture.
To establish the Content of the image I returned to the vintage pattern illustrations from the archive at Wikia (2018) narrowing my choice to those that typified the fashionable silhouette of the time. Pictures of smiling young women in Ebony (2018) magazine as well as the iconic image of Elizabeth Eckford , one of the "Little Rock nine", walking with head held high - through a barrage of verbal abuse after the rules were changed to allow non-white students into the Arkansas High School, affected me deeply. These girls wore simple but smart dresses of fitted bodices and full flowing skirts, they were obviously influenced by the Parisian Haute Couture “New Look” collections of Dior, the juxtaposition of expensive couture modelled by white women an d these home-sewn dresses provided an evocative starting point for the final piece.
This New Look Silhouette was a celebration of the relaxation of the strict and sensible rules of rationing imposed during the second world war. As Cawthorne (1996) explains, whereas the quantity of fabric allowed in a dress in the UK had been severely restricted under the government utility program during the war (and indeed for many years afterwards as trade and the economy recovered), the New Look decadently and exuberantly threw those rules aside to embrace a romantic extravagant silhouette turning women from essential workers into fairytale princesses in acres of fabric and ruffled petticoats, but by doing so trapped them in another set of rules, societal expectations.
Threats to autonomous individual creativity will always exist, the pressure to publish an academic work, the financial need for an artist to sell a painting or for an author to get an agent, were among the influencing factors that many creative people and works are corralled in, as described by Kieran(2018). Even art movements, often initially bursting into being from a collective desire to break free, can start to fester once strict rules for belonging emerge, fracturing the group into individuals once more.
After iterative trials alongside Deep Art with several different images from the book, I decided to use the same technique employed in my previous work “Fractured Bustle Gown” and filled a technical line drawing of the orbits of the inner planets with areas of solid and graduated colour. Combining the mechanical diagrams with this hyper-feminine silhouette, Deep Art produced a very interesting work. While previous images from the astronomy book had produced unusual and pretty 2D images, After I had re-processed the fashion illustration into monochrome and adjusted the levels to give a more defined silhouette, DeepArts algorithm continued to lock on to the very graphic coloured shapes of the technical design best, producing an illustration that could feasibly be pattern cut and remade in 3D.
In conclusion, did Deepart break any rules? No, it created work within set parameters established by its programmers. Did I break any rules? No, my work is still within the established boundaries of the material and the form of pattern cutting. However, by investigating the parameters and algorithms governing our individual artistic output, a better understanding of those was formed in order that DeepArt and I could combine the rules of two different disciplines, establishing a new creative work and way of collaborative working.
References
Anscomb, C. (2018, July). Creative Agency and Co-Working in Artistic Practice. Paper presented at the meeting of CFI Creativity in Art, Science and Mind, Cambridge.
Boden, M. (1992). The Creative Mind: London: Abacus.
Cawthorne, N. (1996). The New Look - the Dior revolution. London: Hamlyn.
Cohen, H. (2010). “Driving the creative machine”. [Seminar Lecture transcript]. Retrieved from
Colton, S. & Wiggins,G.A. (2012) Computational Creativity: The Final Frontier. Amsterdam: IOS Press.
DeepArt. (2017). Latest Artworks. Retrieved from https://deepart.io/.
Ebony/Google. (2018). Ebony magazine archive. Retrieved from https://www.ebony.com/archives.
Garland, A. (Director). (2018). Annihilation. [DVD]. United Kingdom: Paramount Pictures.
Gatys, L.A., Ecker, A.S., & Bethge, M. (2015). A Neural Algorithm of Artistic Style (Paper). Retrieved from https://arxiv.org/abs/1508.06576
Halina, M. (2018, July). Insightful AI. Paper presented at the meeting of CFI Creativity in Art, Science and Mind, Cambridge.
Junaidy D.W., Nagai Y., Ihsan M. (2013) Craftsmen Versus Designers: The Difference of In-Depth Cognitive Levels at the Early Stage of Idea Generation. In: Chakrabarti A., Prakash R. (eds) ICoRD'13. Lecture Notes in Mechanical Engineering. Springer, India
Kieran, M. (2018, July). The fountain of Creativity. Paper presented at the meeting of CFI Creativity in Art, Science and Mind, Cambridge
Prosecco. (2014). Computational Creativity as a Topic and a Discipline. Retrieved from http://prosecco-network.eu/node/242.
Shevlin, H. (2018, July). Creativity and Insight in Animals and Humans. Paper presented at the meeting of CFI Creativity in Art, Science and Mind, Cambridge
Strugatsky, A. & Strugatsky, B. (1972). Roadside Picnic. [Kindle Edition]. Retrieved from http://www.amazon.co.uk
Tarkovski, A. (Director). (1971). Stalker. [DVD]. Soviet Union: Artificial Eye.
The Harold Cohen Trust. (2018). Aaron Draws. Retrieved from http://www.aaronshome.com/aaron/index.html.
Van DeMeer, J. (2014). Annihilation. [Kindle Edition]. Retrieved from http://www.amazon.co.uk
Wikia. (2018). Vintage Patterns. Retrieved from http://vintagepatterns.wikia.com/wiki/Main_Page.
Illustrations
Fig 1 Design for Evelyn Boyd Granville (DeepArt 2018)
Fig 2 Sketchbook Page of data input and designs generated for "Heavens Embroidered Cloths" (J.Hewitt 2018)
Fig 3 Sketchbook Page of Designer Dior and home sewing patterns (J.Hewitt 2018)
Fig4 Early design for Heavens Embroidered Cloths (DeepArt 2018)
Fig 5 Detail: The first colour image created by AARON at the Computer Museum, Boston, MA. Collection of the Computer History Museum (Harold Cohen 1995)
Fig 6 Go champion Lee Sedol, at right, studies the game board during a match against the AlphaGo AI program. Google DeepMind researcher Aja Huang, at left, made AlphaGo’s moves on the board. (Google Deep Mind 2016)
Fig7 Early Colour experiments with Vintage astronomy diagrams (J.Hewitt 2018)
Fig 8 Elizabeth eckford walks to school (J.Jenkins 1957)
Fig 9 The Circular Skirt pattern reflects a vintage Star Map (J.Hewitt 2018)
Fig 10 Iterative draping process of heavens embroidered cloths (J.Hewitt 2018)
Fig 11 Heavens Embroidered Cloths (J.Hewitt 2018)