Design & AI

March 15, 2021

In the context of design, AI comes into play via two distinct phenomena: artificial intelligence enhancing the work of designers and artificial intelligence all on its own as an artistic movement.

When we talk about AI design boosting what designers create, we speak of how technology empowers designers to take their artworks to a whole new level. When we refer to this design trend as a creative movement, we’re touching on the aesthetics that operate on the continuous human-machine feedback loop. This is when a designer interacts with artificial intelligence in a collaboration to produce mesmerizing visuals.


We have to understand the actual history of machine intelligence to understand the history of neural network art. If we go back far enough, we can see that, even since antiquity, AI has been cited several times. Think of mythology with reference to artificial intelligence in all cultures and their cognitive explorations.

In literature, AI has also been highlighted, such as, notably, the gothic novel Frankenstein by Mary Shelley, which some arguably identify as an origin point for steampunk style. AI science as a recognized area of study debuted at Dartmouth College in New Hampshire in 1956. Artificial intelligence, coined by the late cognitive and computer scientist John McCarthy, expanded into computers immediately:

  • Playing Checkers
  • Tackling Algebraic Equations
  • Proving Logical Theorems
  • Speaking English

By the 1960's, the U.S. Study into machine learning has also been greatly supported by the Department Of Defence. However, funding and interest in this area of research dramatically dried up in the 1970s and early 1980s, contributing to what was colloquially referred to as the AI winter. Interest in AI and AI architecture has seen a revival in the late 1990s and into the 21st century after striking a bottom in the early 1990s.

In data processing, medical diagnosis, and logistics, artificial intelligence has been rapidly used. Perhaps the most significant development towards the end of the 20th century was when the then reigning world chess champion, Gary Kasparov, was beaten by Deep Blue, IBM's chess playing machine, in 1997.

By 2011, AI had advanced to the point in the 21st century where IBM, again, built a question answering machine called Watson to potentially appear on the game show, Jeopardy! In which it defeated the two most-winning champions in the history of the show.

Popular items that have recently been involved with AI include:

  • Microsoft’s The Kinect, a body-movement user interface for both Xbox 360 and Xbox One
  • AlphaGo, a computer program specifically made to play Go.
  • Microsoft’s skype-like system that translates languages.
  • Facebook’s system that describes pictures to the blind

And that takes us all the way to 2018, where the art of artificial intelligence began to emerge.

All this is due to something called, generative adversarial networks or GANs. 

These are artificial intelligence systems groups developed by a machine-learning researcher, Ian Goodfellow.

In short, GANs take a training set of data and then create new data based on the characteristics of the training set they get. In Common parlance, if they are "trained" in photography, GANs may produce images that look realistic to observers.

Tools from companies like Adobe, such as Spark, of course, enable artists to produce masterpieces of AI design. Adobe has also been flirting with making InDesign more AI-friendly, offering designers another tool for producing truly original, digital artworks to increase their penchant.

There are examples of AI software that make it simple to build AI designs, but there is also dedicated artificial intelligence art created by humans and machine learning themselves who use these tools and others.


Defining the features of neural network art is somewhat tricky because they are mostly based on simplistic realism. This style trend is not to say surreal, like simple Surrealist design, but more like a difficult-to-spot reproduction of real-life design – that is where the amusement starts.

There are two key things to keep in mind:

  • What looks like a picture or portrait of a realperson is in fact not real
  • What looks like it was rendered by the hand of a human artist is in fact fake

In other words, AI architecture suggests that machine learning has evolved these days to the point that it is no longer even possible to say what the human created and what the machine created.

Here are some visual idiosyncrasies in artificial intelligence art to try to distinguish an AI design from a human design:

Distortions – However slight, some artworks in this genre display, whether on purpose or not, fuzziness and jagged edges, making them a dead giveaway.

Too much perfection – While this is generally a very subjective term, when it comes to AI design, images (especially of people) will often look too symmetrical or too balanced

Angular features – This is a reference more so to illustrations rather than people’s faces; straight lines and geometrical shapes abound for the ultimate in abstraction.

Intricate patterning – Since machine-learning systems are capable of a lot of calculations, it’s no wonder that some visual artworks boast patterning so detailed and interesting that it’s easy to get lost trying to figure it all out.

Striking, vibrant colours – Not just the territory of flat design anymore, neural network art has shown that it, too, can compete with flat design and other modern trends in the area of bold, stirring colours.

Great attention to detail – Unlike design trends like naïve art, where the illustrations are purposefully more simplistic and rough around the edges, AI design means details galore, perhaps almost to a fault.


ALAgrApHY's Robotized Fembot

ALgrApHY, a Paris-based artist and scientist, takes AI and all its visual potential to new heights.  His machine-learning artwork, the designer of many notable works in this design revolution, shows the complex patterns that are possible with this technology and approach to art.

His newest piece is the self-fanning "fembot."  The presence of a multitude of "eyes" within the production is what instantly stands out for the viewer, creating a pattern that, far from distracting, adds to the overall composition due to the visual texture it offers.

This Person Does Not Exist

Human image synthesis is one feature of AI architecture that is likely to go unnoticed by a lot of people. This is when movies construct human beings, even though they are not actual individuals, by digitally composing them into a scene.

With this technology, the flaw is that it actually took a lot of human participation to pull it off, but that's a thing of the past. Nowadays, machine learning has progressed so far that artificial intelligence automatically generates these fake images without a human need to raise their finger.

That's where the image of This Person Does Not Exist comes into the story. This website ( demonstrates an endless collection of pictures of people that are all generated by AI. Note how lifelike these fake individuals look, but also how perfect they look too. All of them have beautiful skin without blemishes.

Beyond appreciating the creepiness of this application of AI, it’s also not lost on us how being able to easily fabricate realistic-looking images of all sorts of people is a recipe for fraud of all kinds.

Christie's Portrait of Edmond Belamy

You know that when even Christie's, the renowned auction house specializing in high-class art, has to get into the action of art created by machine learning, AI design becomes massive. This auction house recently successfully sold a painting in the middle of 2018 that was an artificial intelligence artwork, rather than a human hand. You heard us right: this artwork is the result of a computer in its entirety.

Titled Portrait of Edmond Belamy's , the digital piece derived from algorithms, but it still sold for an incredible $432,500. The idea that a machine-learning algorithm might create a painting worth almost half a million dollars is incredible, but art is often subjective, as are the prices attached to its works.

The actual painting is the work of a Paris-based collective of artists called Obvious. Their approach was simple: they entered a data collection of thousands of paintings from the 14th to 20th centuries into their scheme. In the method, one algorithm is called the Generator, while the Discriminator is the other one. The former generates new data set pictures, but the latter tries to distinguish differences between paintings made by humans and those generated by the Generator. If the second algorithm is fooled into believing that human beings render the newly generated images from the Generator, then a result is created (painting).

Note how the portrait features strong distortion and unclean lines, creating the impression of mystery (intentionally or not).


In all kinds of programming, machine learning continues to integrate with the human side. Today more than ever, in ways both impressive and thought-provoking, human imagination is complemented by artificial intelligence.

AI architecture is still in its infancy, to be certain. In this revolution, what you are seeing today is just its beginning, and its potential is practically infinite, the more technology progresses. In other words, if you were impressed by this movement's beautiful visuals in the artworks we covered, you haven'tseen anything yet.