The impossible task of making creative A.I.

Artificial Intelligence is rapidly creeping into every fold of our daily lives, from helping us with music recommendations to preventing suicides through early signal detection. It is also treading precariously into the creative fields, which undoubtedly represent its biggest challenge to date. As we are already discovering, A.I. is both incredibly adept at producing creative work, whilst simultaneously making it impossible to love.

When Hello Games, an independent gaming studio in sleepy Surrey announced to the world that they were building a space exploration video game which would feature more than 18 quintillion unique planets to explore, the gaming press exploded with hype. But when the game was eventually released, it was met with a tsunami of disappointment and criticism.

The Grid, an A.I. startup that promised to deliver a fully automated website builder (and a respected rival of my own business, Firedrop), fell into a similar trap as its heady balloon of voracious hype eventually burst as the end product failed to match consumer expectations.

Neither product was bad, however (nor are they dead, both have done quite well commercially as far as we know). They were both victims of excessive hype which, as we all know, is a predictable prelude to mass disappointment. But neither company should be criticised for their part in fanning the flames; I challenge anybody with a pulse to stand up in the middle of a crowd baying for your product (with money in their outstretched hands) and tell them to stop being so excited. These are businesses, it would have been almost irresponsible of them to do so. But the reason for the hype becoming so inflated so quickly is less talked about: creative products are consumed in a very unique way that is totally beyond the control of the creator.

The death of the creator

In 1967, French literary critic Roland Barthes published a controversial essay entitled “The Death of the Author”, in which he argued against the practice of assessing an author’s background and motivations in any interpretation of a piece of literary work, instead arguing that text and author are separate entities entirely. The text, once created, lives separately from its creator, and its meaning is assigned by the reader rather than the author. At university, where I studied English Literature, this was the first essay we were assigned to read, and it changed everything. Its power is fully evident when you extend its domain beyond literature, to all the creative fields. Art’s meaning is endowed by the audience, not the artist. Beauty is in the eye of the beholder.

So if the audience owns the creation, who sets the value? Wikipedia hints at it with its definition of creativity:

Creativity is a phenomenon whereby something new and somehow valuable is formed.

“Somehow valuable” is a deliberately nebulous description, deliberate because the issue itself is both intangible and entirely subjective. If a five-year old child scribbles a few lines onto a piece of paper, does that piece of work have value? To the mother who receives it as a gift and then pins it to the refrigerator door, very much so. To anyone else? Unknown, unless the picture is shown to a wider audience. Take the story of Autumn de Forest, a child painter who, at age seven, became a professional artist when one of her works was sold for over $1000. At age eight, a batch of her paintings sold at auction for $100,000, taking just 16 minutes to do so.

“Desert” by Autumn de Forest, aged five

In art, the value is often set clearly at the point at which the hammer hits the block. It is also largely a transaction of perfect information, insomuch as the entirety of the object being purchased is visible at the point of purchase. It very literally belongs no more to the artist, but to the art buyer, who has quantified its value numerically via the price paid.

In the digital realm, the creative product itself is not fixed. Whereas the experience of consuming a painting, piece of music or novel is generally consistent — the creation does not change in structure once purchased — with digital products the experience is influenced by the consumer’s interaction with it. A video game’s qualities are tied to the player’s skill and specific approach; a website builder’s output is based on a user’s own content and eye for design. The experience is both owned and moulded by the consumer, and the creator has very little way of accurately predicting exactly what that experience will be like.

A software release’s value is, therefore, largely linked to the longevity of the experience, or at least its replay value. But price tends to correlate to usage time, and No Man’s Sky entry into the market at the premium price point of £60 meant that people expected a big, immersive game. When it turned out that there wasn’t a great deal to do, disappointment set in quickly and the rest of the criticism followed. Similarly, when The Grid pinned a $96 ticket price to its pre-launch beta invitation, people’s expectations immediately peaked, so when an unfinished beta product was eventually released, disappointment was inevitable. Although I cannot fault either company for setting these prices for commercial reasons, the reaction to each release underlines the point that the value in creative products is ultimately determined by the audience, not the creator.

Expecting the unimaginable

Many products launch with a high price tag, of course, as the producers seek to recoup some of the early production costs from early adopters who will pay more just to be first to own. Creative products are no different: books start in hardback at comparatively high prices and then progress to cheaper paperback formats; movies launch in premium-priced cinemas before ending up on a rental service like Amazon Prime. Hype and demand for a particular object or gadget is only partially tempered by high prices, as those who are put off initially tend to wait until the price drops, delaying their gratification rather than dismissing the product entirely.

But with A.I., there is an additional factor that multiplies the hype and intensifies the disappointment when expectation is missed: the unknown. There is no set definition for what A.I. is, which inevitably means that people fill the gaps themselves with whatever they think or would like to believe. Often, the knowledge gaps are so pronounced that people don’t even know what imagined reality to fill them with, leaving them hanging in expectation of something literally unimaginable.

This is not helped by the hyperbolic reporting from various mainstream press and digital media outlets on the various advancements — and potential future advancements — of artificial intelligence in the real world, which ranges from speculation about its eclipsing of human intelligence levels within 30 years to sinister predictions of humanity’s impending doom at the hands of killer robots.

Not the friendliest character

It is seductive to imagine a piece of art painted by an A.I. artist with an I.Q. of 2000, but in reality most applications of A.I. currently are about as sexy as a statistics textbook. The unglamorous truth is that artificial intelligence, for all its hype, is basically just maths.

One of the criticisms unfairly levelled at No Man’s Sky was how similar the planets seemed after a few had been visited. To explain, the planets in the game are procedurally generated from a set of variables with a 64-bit seed, which in plain English means that if you generate every possible combination of every single variable within the set, you end up with some 18 quintillion different combinations. It’s a purely mathematical process, but the numbers are big enough to give a pretty broad scope of variety. And yet, some people were disappointed in the results. When asked to quantify their expectations, the general response has been “more variety”. In other words, they don’t really know. It is impossible to meet such expectations.

There is another, slightly more unpleasant attitude that impacts people’s reactions towards A.I.: mistrust. In simple terms, people don’t want computers to be better than them, at anything. In the creative fields this is particularly pronounced, because the barrier for entry for activities such as painting or writing are comparatively low. Creative A.I. projects such as AARON, an automated painting robot originally built in 1973 that has produced work which has been sold for thousands, are continuously facing accusations of not being truly creative. Poetry produced by Google’s A.I. is being defined as “accidental” or random. Such creative effort is received with amused interest rather than credible acceptance.

A piece of art by AARON, the painting robot

And yet there is acceptance that AARON’s work has value, as indicated by the fact that some of it has been sold at auction. Similarly, Google’s A.I. poetry has been labelled with emotive descriptors such as “mournful” which seem to be at odds with its apparently cold, mathematical creation process. Computers are unfeeling and calculating, humans are emotional and reflective.

There is an underlying belief that art, in any form, is a uniquely human experience. While we can appreciate the creative results of computers such as AARON, we cannot accept it as anything more than clever imitation or curious random assembly. We would accept a five-year-old girl’s ambiguous paint strokes as true art than a 35-year-old robot’s work.

Consumer takes all

The philosophical ins and outs of the debate around whether a computer can really be creative are outside of the scope of this article. The fact is that most definitions of creativity validate AARON’s paintings, No Man’s Sky’s planet designs, and Google’s poetry. Roland Barthes, in his essay, described a literary text as “a tissue of quotations drawn from the innumerable centres of culture”. A.I. machines crunch through millions of lines of published literature (“centres of culture”) to generate lines of poetry (“tissue of quotations”).

Talent, skill and effort are the processes by which great creative works are produced. In the realm of A.I., skills are learned programmes that computers are becoming increasingly proficient at acquiring, while effort is a function of exponentially expanding processing power. Talent, in a world where an A.I. computer can beat the human world champion at a game as complex as Go, is a term that has to be allowed to be used to describe machines as well. The human domains of emotion, reflection and appreciation are not fully necessary on the side of the creator, when they can be imbued by the consumer. A creative work may be born in a studio, but it lives inside us.

Creative A.I. is one of the most exciting areas of innovation in both the technological and creative fields. In order for it to thrive, creators need to understand that value is determined by their audience, not just the work itself. Consumers, meanwhile, need to stop worrying so much about what computers can and can’t do. Just appreciate what they have actually done.

CEO & Founder of Firedrop.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store