Polyworld

Check our this video on evolutionary design of artificially-intelligent neural-net computers.

This is an interesting google techtalk by Virgil Griffith. It's a little raw, but the ideas are exactly what I am interested in recently (there may be some parallels to the project I was sharing below). There are also some pretty amazing visuals.

When I was watching this I was asking a bunch of questions about the fundamental assumptions of the project. Perhaps this is just because I am an artist and want to use these ideas for my own work, but also I feel as though these ideas can be so much more than just a way to study artificial intelligence. First of all, this simulation is run in order to understand how neural-nets function. Of course this makes it possible to "tweak" the system and set the stage for efficient evolution, but it strikes me that the tweaking was not wide-ranging enough - that chance should be used in the tweaking of the environment as well (because intelligent-design is usually bad design). But also it seems that understanding these systems may be irrelevant to their final use. They aren't made to be understood in the way that traditional computing systems are understood. After all, nobody looks at traditional digital computers and looks at how the random errors in their code function to make them more fit (or at least nobody did until the development of neural-net computing). Then, I don't know why these kinds of algorithms need to be run on computers at all. What if we looked to Von Ahn's "wet ware" concept and tried to create detailed and self-correcting algorithms that used humans as the nodes or agents? How can we use this simulation as a plan for both using our brains more efficiently and structuring society more democratically? These concepts are not immediately useful to the development of artificially-intelligent computers, but it seems to me that they are already useful in some way.

Artistically, these areas have been explored to an extent (for examples, read here), but most of the work seems to be an aesthetically-enhanced version of the "life" screensaver. I am searching for ways to use the ideas from simulations like Polyworld, not to replicate its look or even function. The use-value may be entirely outside the realm of computer simulations. Perhaps they will be performative, perhaps the art will simply seek to explain the process, perhaps the work will be some kind of an adaption of algorithms taken from these experiements, but adapted to new circumstances.

Finally, it seems as though our fucked-up society (so well-spoken today...) is flailing around (like an unevolved virtual agent) for a new way of seeing itself. Seeing is the first step to understanding and taking action. Part of our perception of ourselves is the metaphor we use to define ourselves. I think this project and many other developments in evolutionary studies, cognitive science, and artificial intelligence research are leading to establishment of a new analogy by which to judge ourselves.

One of our goals as artists is to give people something new to see. We accomplish this not by creating an image that has never been seen before, but by leading the viewer down a visual path that sneaks by the locked gates of perception, allowing them to see the commonplace in a revolutionary way.