Brain Candy #42 - High Quality Weirdness

I just found the following site a few hours ago. I think it's really good, but it might just be really weird. See what you think: go to the Reciprocality Project at www.reciprocality.org/Reciprocality/index.html and have a look.

The very first sentence of the introduction grabbed me, in particular the part about "how most people in most human societies have a consistently distorted view of everything" (everything is in italics.) The whole opening statement is full of provocative things. I've read through the entire synopsis - the thesis seems to be that most of humanity is profoundly dysfunctional. Most people in most cultures and most time periods, have become addicted to the biochemical byproducts of their own boredom. It gets progressively weirder from here, but I think there's something to it all. Many of the statements, while not immediately verifiable, seem to make sense. One thing that suggests credibility to me is that the links and references are of high quality; this site is worth visiting just for the quality of links and references. Even if you think it's a bunch of hooey, I believe that you will find this site thought-provoking.

There are now a number of screen saver programs that do useful work. This concept is called distributed computing and it is just beginning to be exploited. SETI@home is by far the most famous example; I mentioned it in Brain Candy #26. It is designed to look for evidence of extraterrestrial intelligence in radio signals received from space. Volunteers download small chunks of data, analyze it during idle time on their computers, and return the results to the SETI site. Just last month, on December 16, 2000 to be exact, they had accumulated over one-half million years of CPU time on the project. If you haven't looked at it recently, check out setiathome.ssl.berkeley.edu; it's matured quite a bit since I looked at it last.

But SETI@Home isn't the only distributed project; there are other projects underway. One distributed computing application helps chemists model complex protein structures. Another looks for huge prime numbers. The one I'm going to talk about looks quite weird. It involves an artificial intelligence experiment that creates "creatures" that evolve on your computer and move around on the Web. Visit The Golem Project at golem03.cs-i.brandeis.edu and you can find out more about Genetically Organized Lifelike Electro Mechanics or Golems.

A Golum is built from simulated mechanical parts and actuators. The goal of a successful Golum seems to be to be able to move around effectively, using minimal energy. Although I haven't been able to study the site in detail, the program seems to be doing some variation of genetic programming. This is an artificial intelligence technique that pits a number of candidate "things" against each other. The things can be artificial creatures, mathematical models or anything else that is to be optimized. There must also be some concrete definition of what a "successful thing" is. A competition is held based on the definition of success and some number of the best candidates are kept. The program then begins to create a new generation of things. It keeps some of the previous winners intact, while others are hybridized with other successful things. Some pure randomization is also introduced, to add variety to the "gene pool". The competition is then held again. Over many generations, great improvements can be seen and truly practical things can often be observed to evolve.

I downloaded the screen saver and let it work for a few minutes; nothing good evolved. This wasn't surprising, the web site says that it may take a few hours to a few days for really good Golems to start appearing in a new environment. The creators of Golem@Home aren't running the project like SETI@Home. You don't directly pass data back to them - you pass successful Golems to other Golem@Home hosts (but only if you allow it), and only rarely. Your best beasties then compete in the new environment. After a time, the creators feel that the really good survivors will migrate to them, but they encourage volunteers who observe unique Golums to e-mail their specs to The Golem Project.

This whole project strikes me as weird, but also cool. I'll be watching my Golems for a few weeks to see how they evolve. Genetic programming is a very exciting field; you too may want to start building Golums in your PC's spare time.