Among the many pictures of robots, news articles, random scraps and doodles posted on the walls of the Evolutionary and Adaptive Systems laboratory at the University of Sussex, there is a comic strip that has been printed off from the archives of PhD Comics. The strip overall is pretty underwhelming, as there really isn’t that much comedy to be had in post-graduate work, and recognition of similarities between your life and a situation presented in the media doesn’t make for laughs on its own. But one particular strip, I do laugh at, because it is utterly faithful to a conversation I’ve had dozens of time:

When you live and work with other students, asking what a person studies becomes the default icebreaker, as being in school is the one thing you’re guaranteed to have in common. But even among other post-grads, the more technical I got with my topic the less able I was to really communicate it. Depending on who I was talking to, I studied any of (from least to most specific) computer science, informatics, artificial intelligence, artificial life, or Evolutionary and Adaptive Systems, the actual program I was enrolled in. And when summer came around, the question we got asked the most became the hardest one to answer:

“So what are you writing your dissertation on?”

Ugh. I understand why people want to know; many of my friends had serious difficulty deciding on a topic they wanted to spend four months studying, and misery loves company. But that doesn’t make it any less awkward. How do I effectively communicate what I study, and what I wrote my dissertation on, to an audience that hasn’t studied it, and may be asking over drinks in a pub where it would be silly to go as in depth as my interests allow? How do you elevator pitch a research project?

Like this.

Artificial Life

First, a bit of background. Artificial life is a field of research dealing with the simulation of life, natural processes and evolution through computer models and robotics. The hope is that by studying controlled models that emulate natural systems, we can understand more about how life and evolution work, and can hopefully find inspiration for better research into engineering. It’s still a very new field of research, with most of the work being done at various institutions in the UK and Europe over the past 20 years.

The catch, of course, is that using simulations to study natural systems is a very vague description, and the work involved in it can take any number of forms. Studying the optics and neural systems of insects to model how they navigate over large distances with such simple brains? Yup. Using evolution-inspired algorithms to “breed” microchip layouts instead of designing them by hand? Sure. Designing simple robots that have to work together to accomplish a task the same way many animals work and live? Of course. Air-traffic control inspired by the migration patterns of birds? Hell yes! In our small class of fifteen people over one year, we had people working in sensory substitution, cell modeling, neurology, robotics, swarm behaviour, evolved communication, and a bunch of topics I still don’t entirely understand. It’s such a big field, in fact, that the first thing I had to do was figure out what parts actually interested me.

My Focus: Simulating Evolution & Animal Behaviour

To decide on a dissertation topic, I looked at the dozens of journal articles I’d read through to try and remember which ones I struggled to even finish, and which ones I immediately started to analyze and dissect and critique. It turns out there was a very obvious theme; taking theories about how specific animal traits and behaviours evolved, and creating simulations that tested those theories. One of the dilemmas of putting forward hypotheses about the nature and scope of evolution is that, until recently, all we could do to confirm or reject them is study the historical and archaeological records.

Simulations provide a much more thorough and flexible method. You can start with a blank population of random agents in an environment that you control, decide how the agents are going to breed and multiply, and determine exactly what it means for one of your agents to be “successful” in your experimental world. Then just hit the Start button, and your agents are off to the races, evolving and adapting and breeding and dying for thousands of generations in a matter of days. It may take a few tries to get right, but you’re rewarded with a complete and detailed record of every single step the evolutionary process took to get to your final, hopefully working population.

Suppose I want to know why so many animals have evolved overly exaggerated body parts and other traits, even when those traits make the animal less able to survive in its natural habitat. Peacock tails are a good example. They’re so big that they impede movement both on the ground and the air, and they’re too colourful to blend in with any kind of natural surroundings. Yet the male birds evolved them, so they must offer some kind of benefit. But what?

Magnus Enquist and Anthony Arak asked exactly that, and ran an experiment to find out. They created two populations of agents; males, with a body shape and size, and females, with a simple optic system for seeing the body of the males. The females were trained and evolved based on their ability to detect males of their own species vs. males of a different body type, and the males evolved based on how easily detectable they were. It turns out that those simple conditions were enough to lead to evolved bodies with highly exaggerated features, even if those features decreased the animal’s survivability. It doesn’t answer all of the questions about how exaggerated signals influence mate choice, but it does highlight how a simple simulation can shed light on a very complex topic.

My Dissertation: Sexual vs. Natural Selection

One thing I did notice, in all of my readings, was an assumption that influenced entire paradigm of evolutionary algorithms; animals that were able to survive were always assumed to be more likely to bear offspring. But as soon as you say that out loud, you realize it isn’t true at all. Charles Darwin himself theorized that, in addition to natural selection, there was a sexual selection process taking place at the same time. While natural selection focused on the competition between members of different species for the resources needed to survive, sexual selection focused on the competition between members of the same species (usually the males) for the resources needed to reproduce; namely, the females.

There are a lot of issues that factor into whether or not a given animal is able to find a mate, bear children, and ensure that those children survive long enough to have children of their own, and not all of those issues are related to survivability. There is always a balance to be struck. If an animal’s camouflage is good it will be able to escape from predators, but if it is too good, even members of its own species won’t be able to see it, and it will die of old age before having any offspring. Good for the self, but not so good for the species. I wanted to find out where that balance was.

I ended up using a robotics platform to simulate my environment and populations, mostly because I’d worked with it during my courses and it lended itself particularly well to my experiment. The idea was to have males and females try to find each other in a simple environment; the females would evolve to send out signals that showed the male where they were, and the males would attempt to move to that spot as fast as possible. But every move that a male or female made cost energy, and the less energy a robot had, the less able it was to survive. By letting my populations evolve several times under slightly different conditions, I was hoping to find out which ones led to the robots forgoing reproduction in favour of ensuring their own survival, and which ones made them throw caution to the wind and find each other no matter the cost to themselves.

So did it work?

Kind of. They evolved alright, but due to a bit of oversight in my original model, both the males and females quickly developed an optimal strategy; sit there and do nothing. Even when the importance of reproduction was high and movement was free, nothing I could do made them evolve to any kind of interesting behaviour. But I still think there’s something there, and I got enough data about how they evolved to determine exactly what the problem was, so I plan to go back and fix it one day. If anything, my experiment highlights how important finding that balance is in real life. I did manage to get a lot of cool graphs out of it, though.

That’s probably too long for an elevator pitch, but it’s a hell of a lot shorter than my dissertation, and includes a lot more background info as well. For those of you who I told “I study robots,” I’m sorry for slightly misleading you, but now you know why I didn’t want to get into the details of what I actually studied in the middle of a party. And for the rest of you, if you ever want to see me turn into a big nerd (no, I mean way WAY bigger than I already am), just ask me about this stuff in person. I dare you.

artificial life, school

Welcome! Come in, I love having visitors. Hungry? Need anything to drink? Pardon the mess, I just moved in myself and I’ve been too busy to properly unpack and set up. Give it a few weeks, and you’ll think I’ve been here forever.

My old site was created to let my family and friends keep track of where I was in the world without my having to e-mail people directly. Since I’m not traveling as much as I used to, my writing is simultaneously going in three different directions:

• Research: My time with the EASy group was an utterly fascinating experience, and I’d like to think it has helped me determine the kind of work I want to do in the future. But whenever I try to describe what I studied, I am immediately asked about whether the robots I build will ever take over the world (they won’t). The first dozen or so times were cute, but now I feel like I’m just not explaining it well enough. So this site will become a dumping ground for any fascinating links and research I come across that help show what exactly I do, as well as demonstrations of a few papers I’m going to try and replicate.
• Development: I noticed while job-hunting recently that, for a web developer, I don’t actually have a lot of work that I can show off online. Many of the projects I’ve worked on are down for renovation or otherwise unavailable. And frankly, even if I did have access to the actual code I wrote, I wouldn’t feel comfortable making it open to anyone other than the company that paid me to write it. So this site will also be a place to highlight the tips and tricks I’ve picked up in my years of web development and programming, and all code will be made available for other developers to ridicule see.
• Life: I’ve been told that I’m not a typical computer nerd, and in my eyes that’s a very very good thing. I cook, I play capoeira, I’m socially and environmentally conscious, I love cinema, I bike, and I have strong opinions on topics I know nothing about, although I suppose everyone does. The last thing this site will do is let me get all the random thoughts in my head out into the open, where I hope the Internet will pick them apart like vultures so that I can rebuild them better and stronger, just like Steve Austin in that show I haven’t seen.

So why not just do that on the old site, you ask? Because I have a serious issue with “stuff” in my life. Whether it’s physical possessions, receipts, ongoing projects, daily reminders, or whatever, stuff tends to pile up faster than I have the ability to deal with it. I’ve been gradually developing the two skills I need in order to prevent this from happening; the ability to deal with stuff faster, and the ability to avoid new stuff, or at least to decide where it goes before it turns into the generic “what do I do with this” kind of stuff. While everyone has a different approach to this, one of my favourite techniques is rebooting. Instead of trying to fix a big pile of stuff, simply make a new empty pile, make brutal and harsh decisions about really needs to go from the old pile to the new, and trash the rest. Chances are, if you’ve had something sitting in that old pile waiting for your attention for a month, you’ll never get around to it anyways. And once you have a nice clean space to work in, you suddenly find yourself motivated to keep it that way.

Hence, the new site. Instead of re-designing and re-editing my old website, I’m starting from scratch and bringing in the posts from other locations that I want to be here. That includes both my old Japan travel journals and the stuff I’ve posted on Facebook lately. Wordpress was good enough to bring in all my old posts, complete with the original posting dates. Unfortunately, it wasn’t good enough to bring over all of the embedded images in those posts, nor fix the links to sites that have broken since the time of writing. I’ll be touching up all of the old posts and bringing them into the new Web 2.0 format soon, so if you want to read them in the meantime, now you know why they look so screwed up.

That explains the new design, or lack of. But why did I get a new domain name when I already had a perfectly good one? Frankly, because I didn’t think it was a good one. When I was posting about living in Japan, I wanted a domain that reflected that, so a play on my name using the Japanese alphabet actually made sense. Now that I’m writing about whatever happens to be in my head when I put finger to keyboard, I wanted something a bit more general and personal. Not gonna lie, I spent a lot of time trying to think of a catchy new name; I was even reading blog marketing tips, for some reason. But nothing seemed to stick, and in retrospect, of course it didn’t, because I’m not trying to market my site and become a (barf) professional blogger. And I was resistant to using my own name for a while, mostly because of my slight paranoia regarding the potential for things I say online coming back to me in real life.

As you can see from the URL bar, I got over it. I figure that, even if I never mention my name in my writings, I’d inadvertently give up enough details for a motivated individual to put the pieces together anyways. I’ve Googled my own name (you know you’ve done it too) and there’s nothing out there that I feel the need to hide; it’s mostly just public profiles and technical posts related to my previous jobs and research. Hell, most of the hits are for a misspelling of a Latvian hockey player. It’s easy to remember for anyone who already knows me, and because my name isn’t exactly common, it hadn’t been taken yet. And once this site is properly up and running, the only thing all my writings will have in common is the fact that I wrote them. So davezoltok.com it is.

Now, sorry to kick you out the door, but I have to get this place cleaned up, and it’s likely to look worse before it gets better. Stop by anytime, coffee’s always on.

No tags