Penny Webb

Misplacing Values in the Age of AI

Could artificial intelligence replace human creativity entirely? Designer Penny Webb muses on the current status of artificial intelligence as part of the creative process.

Can we appreciate art without bias towards its creator? Daily experiences of choice are undoubtedly influenced by personal taste, a feeling towards a style, an individual, a political opinion and so on. In an era where artificial intelligence (AI) has the capabilities to create based on trained influences, are we misplacing our efforts by allowing algorithms to permeate our emotional experience?

Within the contemporary art world, artistic endeavors such as Creative Adversarial Networks (CAN)1 a successor of Generative Adversarial Networks (GAN)2, can ‘generate art by looking at art’, subsequently learning about style. Creativity becomes replication, and originality simply a process of ‘increasing the arousal potential of the generated art by deviating from the learned styles’3. Paradoxically, to appreciate art of this nature, the same principles are used to form judgement that may once have been based on a backlog of preformed opinions of an artists context. That ‘context’ however, is now formed from pre-designed influences, and any ‘quirks of character’ are merely the linear progression of an algorithm.

Take, for example, a painting by a favorite artist. Usually, we are prompted to think not only about the piece in question, but about the creator behind it; their history, their context, their choices, their aesthetic. To appreciate a painting created by AI in the same way is a pointless task. AI has no history, it is an algorithm that has never had to face the perils of life. It is, to all intent and purposes, a perfect child, following the rules and learning what it is told, and who can honestly identify with perfection?

Works of art are not the only, in my opinion, worrisome area of culture being targeted by AI. Our political and economic environment is very much privy to algorithmic intervention. Sophisticated recommendation engines are able to discern a user’s taste, and in doing so, able to reinforce those affinities by a bombardment of related recommendations. Think about all the targeted advertising consumed minute by minute on social media, influencing emotional choices buying habits and so on. On a political level, reinforcing an opinion and generating filter bubbles4 has the dangerous effect of creating ever more partisan views.

When forming an opinion about the creative output designed by AI, one can think about the creator behind the algorithm. What was their influence, why did they create it to behave in a certain way? The subjectivity of taste – be it for fashion, art, design, politics and so on – is a construct of a multitude of influences imposed on us by subconscious affinity for aesthetic and habit. As the revered ‘father of AI’, Marvin Minsky, writes5:

1 Cascone, Sarah. AI-Generated Art Now Looks More Convincingly Human Than Work at Art Basel, Study Says. 11 July, 2017, Artnet 

2 Wikipedia contributors. 12 September 2018. Generative adversarial network. In Wikipedia, The Free Encyclopedia. Retrieved 12:00, October 08, 2018 

3 Eingchen, L; Elhoseiny, M; Elgammal, A; Mazzone, M. 21 June 2017. CAN: Creative Adversarial Networks, Generating “Art” by Learning About Styles and Deviating from Style Norms. USA: International Conference on Computational Creativity (ICCC) 

4 Wikipedia contributors. 6 October 2018. Filter bubble. In Wikipedia, The Free Encyclopedia 

5 Wikipedia contributors. 3 September 2018. Society of Mind. In Wikipedia, The Free Encyclopedia 

 

‘When should we quit reasoning and take recourse in rules of style?’

A image generated by pix2pix software. The pix2pix model works by training on pairs of images such as building facade labels to building facades, and then attempts to generate the corresponding output image from any input image you give it. Source: Isola P; Zhu, J; Zhou, T; Efros, A. 2017. Image-to-Image Translation with Conditional Adversarial Networks. United States: Cornell University Library.

Human creative output will always have a certain ‘je ne sais quoi’, that touch of ‘something’ non-prescribed, that you can’t quite put your finger on but resonates somewhere within. To give up this magic, this extra effort of thought in favor of stylistic reproduction or habitual evolution, feels like somewhat of a withdrawal. Until the creation of an algorithm that carries as much history, joy, hardship and context as one experiences throughout a lifetime of interactions, it will be hard to see the artistic output of AI beyond anything but a pretense.

There are advantages to using AI within the creative process however, in that AI does generate fantastic apprentices. Creators (coders, musicians, artists, designers) can impart their wisdom through rigorous algorithmic training, and thus virtuosity is now achievable at the touch of the fingertips. Who said it would take 10,000 hours to master a skill? In this sense, automation has its place in the creative process, as a tool for reproduction.

With regards to ‘design’, a word that permeates all the way from hyper functionality to expressive form, the concept of automated production, and re-production, is somewhat synonymous. Within design, there is an essence to improve, be that through adorning a home or public space, creating useful or pretty objects, interfaces, and other forms media. In this case, where context is something the user creates, and designers create for, could stylistic automation, or algorithmic generation of functionality and form be a benefit? Think about a scenario where the use case is comfort; if an AI could learn what we like and watch how we sit to generate a design that suits our form, would the automation of this output really be much worse than what a designer studying a user could generate?6

When thinking about the evolution of self-referential algorithms, its easy to imaging a hypothetical future where creative industries have been taken over by neural networks. In this world, would this perpetual feedback loop of references eventually merge into one homogenous style? Would we just be left with the color white, listening to a sine wave on the radio? As a designer, I can’t say I have the answer for this, but as with the case of art, losing that certain touch of ‘something’ is an irksome prospect. The solution; diversify reference and take charge of our creation, the greatest power we have is freedom of thought, accept the benefits of automation but don’t let the algorithms do the thinking for us__

6 “What if a CAD system could generate thousands of design options that all meet your specific goals? It’s no longer what if: it’s Project Dreamcatcher, the next generation of CAD. Dreamcatcher is a generative design system that enables designers to craft a definition of their design problem through goals and constraints. This information is used to synthesize alternative design solutions that meet the objectives. Designers are able to explore trade-offs between many alternative approaches and select design solutions for manufacture.” Autodesk Research: Project Dreamcatcher. Retrieved October 10, 2018 

Biography

Penny Webb is an interaction designer specialising in the integration of hardware and digital interfaces within product design. Penny works within the Philips HealthTech Innovation Lab, designing and building digital solutions for medical devices. She is a graduate of the Massachusetts Institute of Technology, where she worked as a research assistant in the Tangible Media Group at the MIT Media Lab, focusing on Human-Computer Interaction (HCI) and material advancements. Penny’s work has been exhibited internationally, and nominated for awards including the Design Museum London’s Designs of the Year.