Curating the machine

On 15th October I caught the last day of Fake! The Great Reveal at Cheltenham Museum. The exhibition follows on from a series broadcast on Sky Arts about forgeries, plagiarism and fakes within the art world.

When talking about copyright and recent attempts to make it more restrictive I've often said that stricter laws will not stop perceived or real theft within the arts. As this exhibition shows, it has been happening for centuries, and whilst technology does enable theft to happen at a larger scale and faster rate, technology in itself is not the problem.

What this exhibition has also demonstrated to me is that there is a definite skill in making "fake" artworks, with some of the artists in the exhibition having a career producing these. The artists would carefully try to reconstruct the painting, even trying to mimic the cracks within the paint.

As well as being an insight into techniques we can use to spot originals and fakes in painting, this exhibition has made me think about how we can go about identifying real and fake art in the age of AI and computer generated art. In addition to the question on whether a computer can have authorship, are the clues that can give away when an artwork has been made by a computer or AI? Should we look for lines that are too neat or pieces that lack blemishes? Should we look out for work that is too resolved? What makes a human painting human?

Is it just the aesthetics or is it the aesthetics + the backstory of the artist. Would Jean Baptiste’s work be as prolific if we didn’t understand his history? One could argue that understanding why the art stands before us as it does adds to its appeal.

For AI artists there isn’t personal history or a muse in the traditional sense. AI artist are fed art images, thousands of images, and are taught to recognize the difference in styles. Through observation, they then generate their own images.

With the support of Near Now I went to Warsaw, Poland to take part in CopyCamp. Now in its sixth year, CopyCamp is an annual gathering of people interested in copyright.

The primary goal of CopyCamp 2017 is to broaden the scope of the debate about exclusive rights. We want to talk about culture, science, and education, but we also want to discuss how exclusive rights are used in, for example, health and food.

I was there to talk about my exhibition, No Copyright Infringement Intended, and how it gives insights into how creative digital practices are conflicting with copyright laws. At first I was worried that my presentation wouldn't be scientific enough but the response from the audience was very positive, and additionally it can be seen from the lineup that the organisers made a real effort to have a range of backgrounds represented i.e. not just lawyers!

Here's my presentation:

Aside from presenting, I was there on a research mission to find out what current and proposed laws have to say about, amongst other things, automation and copyright. A lot has been written about how Artificial Intelligence (AI) is set to change how we create a and receive art. Below are a few of the highlights from CopyCamp that address this issue in some way

Jarosław Lipszyc - CopyCamp 2017

One of first presentations of the conference gave a really good introduction into the complexities of the free culture movement. Copyright in itself is a difficult topic, and the many different licences for free culture do not change this.

Lipszyc shows some of the many licences and how some are incompatible with others. How can we increase engagement in this area if it's so difficult to grasp?

Martyna Czapska - Do Androids Dream of Copyright?

Is authorship only reserved for human beings? Czapska at the beginning of her presentation makes a very useful distinction between computer assisted works and computer generated works. There could be some confusion between the two, and it appears to revolve around how much the computer had in the creation of the artwork.

The UK law states that copyright is assigned to the person that created the program or script. In talking about how Polish law currently doesn't have the same provisions, Czapska asks should the copyright stay with the programmer or the user. After all, a program is dormant until someone activates it. Should the user who activated it or gave it data have copyright over the thing it created?

Ariele Elia - Appropriation in Fashion: Where to Draw the Line

Although not directly related to copyright, Elia's presentation was a fantastic and well thought-out look at the problems with appropriation in fashion. Elia highlights how fashion designers have been borrowing from cultures other than their own and how it can be problematic when those designers aren't sensitive towards those that they're borrowing from, especially with regard to how their new creations are displayed in a new context.

In thinking about AI I wonder if we need to train machines and programmers to be culturally aware when they are creating things through algorithms.

Piotr Płucienniczak - Rozdzielczość Chleba: art, culture and publishing beyond property

One of the last presentations of the conference was a very good insight into how artists engage - or do not engage - with copyright. He talks about in the start of his collective he would publish all of their works under creative commons licences.

I had the pleasure of performing at the Algorave for Blueudot Festival 2017.

The festival was a really great mix of science and music and saw both having equal importance on the lineup. After indulging in music acts like Soulwax (three drummers FTW!) and the BBC Radiophonic Workshop I attended a presentation by Robert E. Smith called Banality of Artificial Intelligence. A bit about Smith, taken from his website:

My interests are complex-systems-based artificial intelligence, including evolutionary algorithms, machine innovation, knowledge representation, evolving agents, and cooperative computation. I’ve authored over 20 journal articles, 10 invited book chapters, and over 50 conference papers on these subjects. I’ve conducted research projects for the U.S. Army Strategic Defense Command, The Center for Nonlinear Studies, Los Alamos National Laboratory, Oak Ridge National Laboratories, NASA, Boeing, NSF, EPSRC, The European Union, DERA, British Aerospace, Airbus, and British Telecom. I am a former Associate Editor of The IEEE Transactions on Evolutionary Computation, and the journal Evolutionary Computation. I am proud to be a fellow of the RSA: The Royal Society for the encouragement of Arts, Manufactures and Commerce.

Currently, I work as Chief Technology Office for BOXARR Ltd., a company I helped to found. I also work part time as a Senior Research Fellow of Computer Science at University College London.

After discussing the common misunderstandings between Machine Learning (ML) and Artificial Intelligence (AI) he went on to reference a paper published by Carl Benedikt Frey and Michael Osborne titled The future of employment: How susceptible are jobs to computerisation?. The paper Goes into a lot of detail about what attributes make each job unique and which of those can be automated by a robot.

The paper has been talked about at length on many websites, in particular highlighting the above graph.

Smith highlighted several problems with this study. One such problem is which unique attributes were being used to measure. See below the table from page 35 of the paper.

Are these the only attributes which make each job unique?

After getting an e-mail from Samuel Fry I stumbled across this article about Artificial Intelligence and Machine Learning, through which I found out about AI-CD β. The project aimed to create the first Artificial  Creative Director for Japanese ad agency H.K McCann.

Much has already been written about this and the challenge it had to create an advert for Clorets. Whilst the results are really interesting I'd be really eager to know what data was used. For example, it was stated that the team tagged lots of successful adverts with different descriptions. Were these descriptions automatically generated? Who decided which adverts were successful? How many adverts were tagged?

I doubt this will ever be revealed as that would reveal the secret to creating a successful advert!

I do wonder how this approach can be used to creative arts. I think there could be more complications, as it's hard to say what makes a successful piece of art. Plus, art can come in so many different forms. Typically an advert will be on a screen (TV, film) or print (magazines, billboards). Visual art can be sculpture, performance, video, digital etc.

Still, this is my task through the residency.

I'll leave you with a meme.

I went to the Thresholds exhibition at Birmingham Museum and Art Gallery on 16th July and had a few thoughts about it. First, a description of the exhibition

Using the latest in VR technology, Thresholds will restage one of the earliest exhibitions of photography in 1839, when British scientist William Henry Fox Talbot first presented his photographic prints to the public at King Edward's School, Birmingham. The experience will be a fully immersive portal to the past; walk freely throughout a digitally reconstructed room, and touch the bespoke vitrines, fixtures and mouldings; even the heat from a coal fire will be recreated. A soundscape for Thresholds includes the sound of demonstrations of the Chartist protesters who rioted in 1839 on the streets of Birmingham, and who can be glimpsed through the digital windows.

I have little personal interest in creating VR work, but I am interested on the impact it will have on experiencing art. My main criticism is that it still needs to mature past the point of experimentation and.

The first thing that struck me about the exhibition was that inside the Waterhall Gallery had been placed another gallery! This VR experience takes place inside a constructed room. I imagine that this exhibition will tour and so make the assumption that this will be reconstructed elsewhere.

When you arrive you've given a back pack and VR headset. I later learned that this backpack contained a laptop which was powering the headset.

Once you put on the gear and enter the white room it is transformed into a 19th century building. The website says that this room is a "digitally reconstructed room [featuring] bespoke vitrines, fixtures and mouldings; even the heat from a coal fire will be recreated".

The main content of the exhibition was well thought out and, although I am no historian, I felt it took me through history.

However, what took away from the experience was the technology, particularly when small bugs (some of them literal) started to appear.

The recreated environment felt very real and natural but the process of picking up the artfects in the VR gallery wasn't. You had to place your hand flat on it and wait for it to jump into your hand, after which you could use the usual pinching motions to make it bigger. This took away from the "realness" of the experience. Having paper as a physical prop which is augmented in VR would have kept me in the VR world more.

The sheer amount of technology and preparation involved also took me out of the experience a bit. Having to carry around a laptop on your back, VR headset, and having limits placed on how your interact in the "real" space made me think too much about what I was doing and not what I was seeing.

This, of course, is no fault of the artists. The technology is still very new and so in however many years it could all be contained within a phone, which would eliminate many problems.

At the end I was given a survey to fill out.

This article caught my interest for several reasons. I believe that robots themselves shouldn't be given copyright over the art that they create. Rather, as the article highlights, we should follow the example of the UK where the copyright belongs to "the person by whom the arrangements necessary for the creation of the work are undertaken". Defining who this "person" is can become tricky, as we could say that the person(s) who wrote the program made "the arrangements", not the person executing the program. But still, I think it's the right direction, and the reason for this is that robots cannot (and I believe will not ) be able to make aesthetic decisions.

When we program an AI or a neural network we program it to look for and analyse certain characteristics. Referencing The Next Rembrandt, the article highlights what characteristics were analysed to aid its creation:

The computer used something called machine learning to analyse and reproduce technical and aesthetic elements in Rembrandt’s works, including lighting, colour, brush-strokes and geometric patterns.

These elements only scratch the surface. Why were certain lighting conditions chosen, what meaning lies behind geometric patterns, why was that subject chosen, how does it fit in with the rest of Rembrandts work?

Of course, the article does say that it is only analysing technical and aesthetic elements, but can a technique really be given a copyright status? As an analogy, can the way a person plays drums be copyrighted or just the music that they create?

After quite a delay I finally watched the video of the talk Posthuman Curating. I really wanted to go to the talk back in April but I think it was during the opening of my exhibition, No Copyright Infringement Intended

I really liked the talk form Gaia Tedone. She talked about her role as a curator in the #exstrange project in which ebay was used as a platform for curating and selling art (I'm not doing the project service, check it out here http://exstrange.com/ ).

Part of her talk that I found interesting was her talking about how ebay categorises and chooses how and where to display art. These algorithms will dictate what we see and what we discover and we give them so much power over our experiences.

I tried doing an image search for curating. It's quite revealing what turns up.

I'm aware that I'm in a filter bubble and a lot of what I now see on Google is dictated by my previous search history. Still the results are a nice mix of what we might think of curating and curating as in content on the internet.

In May I started a fellowship with Near Now.