Jer Thorp is a software artist and educator based in Vancouver with a wonderful knack for creating work that draws on (and questions) a variety of disciplines. Genetics, biological growth, questions of representation and the emergent logic and aesthetics of large, collective pools of content are all topics of exploration in his projects. Jer teaches at Langara College’s Electronic Media Design Program, is quite active on the international new media lecture circuit and will be running a creative coding workshop this summer. Jer and I started chatting about a post of his on data, conspiracy and the work of Mark Lombardi and that conversation evolved into the following interview.

[Jer Thorp / Glocal Image Breeder / 2008]

Greg J. Smith: You recently spent a year in residence at the Surrey TechLab working on The Glocal  Project, a collaborative venture dedicated to examining contemporary image production and management. As part of this research, you developed an application called the Image Breeder which allows users to create "offspring" from sets of "parent" images. Given your background as a geneticist, evolutionary process seems to be a frequent point of reference in your work. Could you talk about genetics and evolution within the context of the Image Breeder and your work on a whole?

Jer Thorp: A few years ago, I was working on a project in which I was using neural networks to make little percussive digital instruments. I liked the sounds that they were producing, but couldn't seem to tweak the hundreds of possible parameters & combinations to get a good result - which I felt must be hidden in there somewhere. I wondered if it would be possible to evolve a better instrument, and a lot of research into evolutionary computing and genetic algorithms eventually led me to a solution that I was quite happy with.

Genetic algorithms are good at solving exactly this kind of problem; problems that have a lot of possible solutions, and in which we don't know exactly what the best answer looks like. GAs have been applied to all kinds of different practical problems - everything from shaping satellite antennae to spacecraft flight planning to circuitry design. What's interesting in a lot of these applications is that often solutions are evolved that couldn't have been predicted, or that seemed impossible before the process started. For example, adding circuits have been evolved that out-perform conventional adding circuits - even though engineers aren't exactly certain how or why. I am interested in how evolutionary computing techniques can be applied to creative 'problems', and am looking forward to seeing what kind of unexpected & unpredicted solutions might emerge.

A role for evolution in the creative process has some scientific grounding. Research by neuroscientists has suggested that evolutionary processes may be at play in our brains when we come up with new thoughts or ideas. Theoretical neurophysicist William H. Calvin from the University of Seattle suggests that creativity may depend on rapid, 'micro-evolutionary' processes inside of our brains that sort through a myriad of creative possibilities every microsecond - we eventually become aware of the best or 'fittest' ones. I think a lot of my work over the last few years has been influenced by this, and in some ways represents repeated attempts to bring form to this abstract idea.

In The Glocal Project, we worked closely with the local community in Surrey, as well as with a global group of participants, to build a very large pool of contributed images (about 20,000 in total). Part of our research involved finding ways to make sense of this huge image base. One of the ways that we approached this was to try to store as much information we could about each individual image: everything from date & location to colour & composition. For compositional analysis, we used a C++ library (libpuzzle ) which converted each image into a string of characters that defined the image's compositional make-up. I was immediately struck that these strings could be thought of as an image's 'compositional DNA', and wondered what would happen if we used a very simple GA to cross-breed them. The result is a tool which allows people to select any two images from the pool and breed them. The 'children' from this combination are other images in the Glocal pool that the computer thinks could have been produced from this hybridization. Through this process, 'phylogenies' are created which show imagined relationships between the images in the pool. The Image Breeder can produce some fairly attractive and interesting visual compositions, and at the same time it's a novel way to browse through a huge database of images in a non-linear fashion.

So would you say that your interest in genetic algorithms, natural selection and biological systems is process focused? From your comments above it sounds like that is the case but on first glance some of your older work (i.e. tree.growth - pictured above) appears to engage in Biomimicry. The thing about this experiment, in which you used L-Systems to develop a "growth grammar" for trees rather than model the way various species appear, is that in your words "you were trying to capture their inherent treeness" rather than produce realistic representations - it sounds like a metaphysics project! Could you talk about this piece and Plumage? These projects seem to thrive in a middle ground between naturalism and algorithmic art.

My work certainly began solidly placed in the arena of biomimicry. The very first thing that I ever built with code was a tree, using a very simple recursive algorithm and mired by the painful constraints of Flash 4. I'd say that the first few years of my exploration into algorithmic art and generative techniques were focused on trying to recreate natural systems in one way or another. I worked at the Vancouver Aquarium for 4 years after University, and spent every morning cleaning the windows of hundreds of exhibits containing fish, plants, and invertebrates of various species. So it's not surprising that, given a new set of tools, I set out to try to replicate some of the forms that I'd been seeing over and over again.

tree.growth was about 5 years after my original tree explorations and I suppose in some ways was an attempt to 'get it right' the second time around. Early in the process, I was intent on making realistic tree forms, but these just ended up looking like bad CGI. So I simplified, working less with the end products, and more with the algorithms that produced them. I introduced some stochastic elements and built a system in Processing that allowed me to very quickly generate a wide array of tree forms. I found that the best tree forms - the ones with the most 'inherent treeness' weren't the most complex ones but, in fact, ended up being some of the most simple ones. tree.growth certainly didn't start off as a metaphysics project, but I did find myself thinking along some of those lines as the project progressed.

The results of tree.growth were very representative and were presented traditionally - as large scale prints on canvas. But the project turned me to a more process-focused path that I've been doing following over the last 3 or 4 years. In some ways I still consider it an unfinished project, and I suspect I'll return to trees and L-Systems at some point in the future.

While tree.growth had elements of biomimicry from the beginning, the feather forms in Plumage emerged more or less by accident. That project started with a very practical goal - I was interested in creating a tool that would generate colour palettes from flickr search terms. The original idea was that you could search for a word like 'anger' or 'toast' or 'Budapest' and the tool would give you a set of colours that were contained in Flickr images tagged with those words. Somewhere along the line, the arrays of colours that were being generated reminded me of feathers, and I adjusted the code to make this more clear. The whole project took less than a day to produce. A small project, for sure, but one that (again) acted as a bridge to more involved undertakings.

I think Plumage was most successful in decontextualizing colour. When colour information from an image is extracted and presenting it in a totally unrelated form, we're able to look at the colours totally independent of their images of origin. I don't think you get the same effect when looking at a grid of colours sampled from an image - but when that grid is rebuilt into a feather, for some reason our minds can remove the locations, forms, etc. that we had initially associated to the colours.

After building Plumage, I became a bit obsessed with colour, and in particular its representation to us in pixel form. In a moment of abstraction, I wondered what would happen in a system where pixels had autonomy - what would a pixel want? This question led to the beginnings of The Colour Economy, a project I've been working on now for about 2 years.

[Jer Thorp / NYTimes: 365/360 - 2008]

Over the last few months, you've done quite a bit of experimenting with the NYTimes Article Search API and, more recently, The Guardian Open Platform. Thus far you've created news maps for specific years and executed a number of Term A vs. Term B comparisons to chart correlation and ideological shifts. How would you describe your interest in these data repositories?

So far, most of the perceived value of the semantic web seems to be counted in dollars. Huge amounts of data are becoming available, but it seems that only a small percentage of people have the know-how to access and make sense of it. With my NYTimes & Guardian work, I'm most interested in building simple tools to allow people with limited programming knowledge to tap into the massive pile of data that is available. With this in mind, I've published a few tutorials on my site to get people started, and will be releasing a pair of Processing libraries shortly to make it simple to connect to the NYT & Guardian APIs. I think it's very important to try to foster this kind of data literacy. As an educator, I'm excited about the possibility of my students having easy access to data, and am looking forward to seeing what new projects and ideas will emerge. I think that we are going to see a lot of work over the next few years using data as a medium.

At the same time, I've enjoyed building various visualizations and creating some hacked hardware objects that connect to the news APIs. Data visualization is often a very serious business, with assorted constraints and restrictions that typically apply to scientific pursuits. As an artist, I've felt that I can leave some of this objectivity behind and create work that has less to do with legibility and communication and more to do with aesthetics and concept. I look at the work that I've done so far in this direction as sketching. I hope that with some more time I can uncover some more robust and more interesting ideas and directions. Specifically, I'm interested in using these data sources as substrates for more abstract visualizations - systems that don't so much embody the actual data as the patterns and trends that might exist, hidden, within them. I'm not entirely sure how this will work, but I'm excited to find out!

Speaking of abstraction and hacked hardware, your NewsAlarm (pictured above) builds on the curiosities evident within the NYT visualizations. For those unfamiliar with the project, the piece repurposes a generic smoke detector as an audible alert that is activated when a certain density of user selected keywords comes through the NYT Newswire API. Could you comment on your decision to choose this device as the audio "voice" of this piece? Was it a play on words in response to the NYT R&D team calling the Newswire a "firehose", or is their an added layer of commentary on the alarmist tone of post cable news journalism?

The idea to use a smoke detector for this piece certainly came from consideration of the alarmist tone that tends to be present in news media, specifically in 'breaking news'. Smoke alarms are supposed to warn you of very urgent news - that your house may be burning down! Compared to this, I'd say 99.99% of the news that comes over the NYT Newswire is trivial.

As an object, I find that the smoke alarm is very anxiety-producing. Even without the battery, I found holding the alarm in my hand akin to holding a wound-up jack-in-the-box. I couldn't get the feeling out of my head that it was going to sound at any second. I get a similar feeling waiting for a news website to load, or listening to the first minute of a radio news broadcast - there's always the possibility that something truly terrible may have happened. I think the news media targets this anxiety of (and perhaps the desire for) disaster, and NewsAlarm is an attempt to speak to that in a somewhat ridiculous way.

Despite my documented use of the alarm to listen for the occurrence of the word 'aliens' in NYT headlines, a lot of people have mistaken NewsAlarm for an attempt at making a useful device, which is amusing and at least somewhat ironic. I've had several e-mails and comments from well-meaning people concerned about my hearing loss, or exposure to the radioactive materials which are contained inside the alarm.

I'm happy to report that the Handbook of Research on Computational Arts and Creative Informatics will be released sometime in the next several weeks. I contributed a chapter entitled "Information Visualization and Interface Culture" to this book and in the space of about 10,000 words I outline a few backstories related to visualization. These include: a survey of early radar technology, Vannevar Bush's Memex, the Head-Up Display (HUD), an excavation of the Graphical User Interface (GUI), the obligatory salute to Edward Tufte and an examination of John Maeda's Aesthetics + Computation Group at MIT. The text isn't entirely about timelines and technology as I refer to the work of Lev Manovich and Alan Liu to consider the cultural implications of pervasive interface culture. The chapter also includes some exemplary work by Ben Fry, Stamen Design, Burak Arikan, lab/RAD and a team comprised of José Luis de Vicente, Irma Vilà and Bestiario.

You can get more info about the Handbook of Research on Computational Arts and Creative Informatics here. If you wish to order it I'd warn you that it is approximately as expensive as a black market kidney. I'm not familiar with that many of the authors but I did notice Ben Bogart and Ethan Ham amongst the list of artists and scholars that have contributed chapters.

This is a nice moment for Serial Consign as the majority of the material in this chapter telescoped out of 2007 and 2008 blog posts and the Interactive Digital Culture for Software Engineering course that I taught at McMaster. I'm still trying to wrap my head around the fact it took almost a year for this text to be proofed, edited and published but perhaps the immediacy of blogging has coloured my perception of workflow. I'm currently working on two nascent writing projects and at least one of these might yield some kind of "book thing" - perhaps I'll have some news to share in the fall of 2010.

Edit: I have documented this writing project more formally in my portfolio and posted a PDF version of the chapter for download.