AI-Generated Painting Sells For $432,500 - A Deep Misunderstanding
Written by Mike James   
Saturday, 27 October 2018

We all know that the art world is illogical in its approach to the value of things and there are many examples. This time, however, things are a little different. To pay just short of half a million dollars for a painting created by a deep neural network displays a deep misunderstanding.

Technology has changed the what we consider art. No longer is it just a matter of paint on canvas. The recent stunt by the artist known as Banksy illustrates the point. Take a limited edition of a well known public art work and auction it off. Just as the final price is settled, the frame whirrs into action and shreds the print. What! Surely this must reduce the value to zero? No it doesn't and what is more the owner looks to make a profit as this unique art work appreciates. Some owners of non-shredded prints were inspired to attempt to increase the value of their works by shredding them. Guess what, their value went to zero rather than increased.

bsnkdy

The reason is that when Banksy conceives of a work of art, even if it involves shredding a print, the value goes up because the value is in the creativity and the event. Any one can shred a print, but not everyone is Banksy who can stage an auction and subvert the whole affair to create shock and awe.

What has all this got to do with AI and art?

The key point is that AI creates art without any intent. There is no aim to shock or please or even subvert. It is as automatic as an algorithm. The Ai-generated painting that has just been bought is like a shredded Banksy print - but not shredded by Banksy. The neural network can create another any time its keepers ask it to. What is more, I could clone the network and create yet another painting.

The end product has no value because it incorporates no intent and it doesn't even acquire value because of supply and demand.

art1

The painting in question Portrait of Edmond Belamy sold this week at Christies in New York for $432,500, over forty times its estimate. As reported in Christies To Auction AI-Generated Artwork, it was created by the efforts of three French art students working as "Obvious". They used a GAN (Generative Adversarial Network) trained using thousands of examples of art. However, to do the job they seem to have simply used an open source implementation by another artist, Robbie Barrat. He has now amended his licence by adding:

EXTRA: NO OUTPUTS OF THE PRE-TRAINED MODELS MAY BE SOLD OR USED FOR-PROFIT OTHERWISE.

If you visit his GitHub page you will find examples of what he has been doing with the same network and, guess what, the results are often very pleasing and interesting:

barrat

As you look at the gallery with the recent auction in mind, it is difficult not to think of the process as deciding which photos to select from a set. The creativity is applied by the human in the detection of something that is pleasing or special in some way. The human is part of the network structure as a sort of super adversary judging the output of the GAN.

So where was the intent in any of this?

The intent to make some money seems to be the only creative part of the enterprise - the actual artwork not so much.

ganart

It may look like a painting in a human style but the deep misunderstanding in this case is about the way that it was created. If the buyers had understood they could not possibly have put so high a value on something churned out by a GAN. There really is no creativity in this particular machine.

As Andy Warhol put it,   

"Art is what you can get away with". 

More Information

robbiebarrat

http://obvious-art.com/

 

ganarticon

 

Related Articles

Christies To Auction AI-Generated Artwork

Robot Art 2018 $100,000 To Win

Can Robots Create Art?

DevArt - Art Made With Code

Mathematical Art 2013

Van Gogh Meets Turing

A Robot With A Chainsaw!

More Efficient Style Transfer Algorithm

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.

 

Banner


AWS Lambda Upgraded To .NET8 Runtime
25/03/2024

An upgrade of AWS Lambda to the .NET version 8 runtime
brings major improvements to the platform.



Couchbase Adds Vector Search
07/03/2024

Couchbase is adding support for vector search across its entire product line including Capella, Enterprise Server, and Mobile. Support has also been added for retrieval-augmented generation (RAG) tech [ ... ]


More News

raspberry pi books

 

Comments




or email your comment to: comments@i-programmer.info

Last Updated ( Saturday, 27 October 2018 )