Author Topic: A Hello, and a Question  (Read 3650 times)

Offline JossiRossi

  • Bot Neophyte
  • *
  • Posts: 31
    • View Profile
A Hello, and a Question
« on: March 20, 2007, 01:52:02 PM »
This is my first post, and just saying hi. I tried out Darwinbots what may have been even a year or two ago and I thought it was a really neat and cool program. I could never quite get the hang of it though and all my bots ever seemed to do was slow my computer down while being pretty much the same.

Now my computer is a bit better, I'm a bit smarter, and the program has really advanced. It's probably the most entertaining artificial life simulator out there. Glad it's still around and actively worked on =]

This is probably a bit of an advanced question, but I've been trying to make my bots become more advanced, but I can't seem to figure out what kind of settings to put things. For better or worse, I tend to restrict energy, and keep certain costs high. Right now I got bored of the bots spinning in circles, so I gave rotation a relatively high cost.

The bots do evolve to cope with the new conditions even if it takes a few days, but I can't seem to get beyond the "run in a line until you hit something and eat it". I assume a multibot bot (?) is something that's difficult to just evolve, and who knows with my settings perhaps they DID evolve and I just killed them off.

Not sure, I keep tinkering and slowly changing the environment to get different reactions. Regardless it's a hell of a lot of fun, just not sure if there is some tried and true method that I missed somewhere.

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
A Hello, and a Question
« Reply #1 on: March 20, 2007, 10:58:28 PM »
What's your starting point?  If you start with an animal that does a bit more than just blindly move forward eating, you might get more interesting results.  Fact of the matter is, we're all still working trying to figure out how to evolve a proper multibot, or a bot that demonstrates even the least degree of intelligent behavior.

Evolution's a little spunkier than most people have been led to believe.  It's not anywhere near the ideal of every generation getting better and better.  It's closer to an explosion: very chaotic and usually unproductive.  IMO, the most "successful" sims are the relatively new ex nihilo that start with an all 0 genome, and build up bots from that.  Still, the bots tend to evolve towards the blindly-move-forward-eating-anything-you-find direction.  Again, no real intelligent behavior.

Offline JossiRossi

  • Bot Neophyte
  • *
  • Posts: 31
    • View Profile
A Hello, and a Question
« Reply #2 on: March 21, 2007, 12:42:21 AM »
My at least sideways attempts have been to try and make it so that the conditions require a more advanced system each time I progress it. I think all I have been accomplishing is making them more efficient. I think it would help if my computer was a bit better. I might just be impatient though, I like it when things can run 100 cycles a second or more, so the population size I try to keep down. I guess this isn't so much advancement through diversity as it is trying to make it so a single really beneficial trait becomes of immediate benefit.

To that extent things have worked well. The bots adapted to all the pressures I have put them through but there are some limitations. I think the bots themselves are very well done, they can do a large number of amazing things when programmed, I'd love to see what can happen when it's by chance.

One of the bots I saw was the Ant one, that is really awesome and shows just what a designer can do, and I know that in time evolution could potentially create the same thing, but what it comes down to is what kind of pressures would be needed to create that?

I don't know, and naturally I surmise that no one really has a good answer, but maybe that's something I need to figure out, is what types of pressure encourage wild deviations, while what pressures force the bot down to it's most effective, yet most base systems.

I'm sure there's a thread better suited for such ideas, but like a real ecology, maybe the food system should be revamped in some ways? At the moment a bot seems to be able to eat anything it feels like all it needs is the impetus to do so. What if it needed more, the right genes to digest, not the right genes to simple attack? Least then you'd get bots that have mutated to where they are ineffective in one setting yet widly successful in another.

At least in my sim which has run 28,000,000 cycles, I know I have lost many new things just because Being strange is almost ALWAYS deadly, and there's little in the way of benign mutations.

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
A Hello, and a Question
« Reply #3 on: March 21, 2007, 12:44:46 PM »
That's quite a long sim you've been running. Have you seen any interesting behavior yet? My sim is at 15 M cycles and the best I've seen so far is a kind of bot I call the TimerVeg, because it reproduces every 2000 cycles and a few cycles after that it shoots a virus. Other than that my bots have become very hardy with CostX up to 26 X the starting costs. I'm pretty sure any of the default bots would get killed in a moment if I put them in the sim. I'm thinking about running an experiment tonight putting a few of every default bot together with my evobots to see what happens. ..in another sim of course, so not to mess up my current one. If an autotroph evobot species can exist together with a few of the default bots than it's pretty much a fully evolved veggie in my opinion, and the next step would be to get a successful heterotroph.

Always good to have new people running these sims. The more the bigger the chance someone evolves an interesting bot. ..maybe.  
« Last Edit: March 21, 2007, 12:48:04 PM by Testlund »
The internet is corrupt and controlled by criminally minded people.

Offline JossiRossi

  • Bot Neophyte
  • *
  • Posts: 31
    • View Profile
A Hello, and a Question
« Reply #4 on: March 21, 2007, 10:24:29 PM »
I was actually just "served" by my own bots. (My girlfriend's words)

I set it up so that when the population dips under 50 the costs turn off (safeguard from extinction before the "goal" reached), then the population is safe until it hits a population of 90. I though that maybe this system of bursts and then whittleing away of the least viable bots would yield some results that I wanted. My goal was to stop them from spinning so damn much, crazy bots them.

So this morning I wake up and check the progress, and Hooray! The bots stopped turning.

Then I looked closer, they weren't eating either.

Nor were they reproducing.

Nor were they reacting to each other.

DAMN! The bots basically shut themselves down, but the trick was they managed to get their to dip below 50, then rise to 70, but never to reach up to 90. The population basically found that bubble of stability and stayed there, they could have lived forever.

Until I turned the costs back on to test my theory, and out of spite of their laziness =p

I learned a few tricks though, I might gain some patience and keep the population a bit higher like 100-200, but even with lower numbers I can do better.

Also keep in mind this nearly 30 million cylce sim was all done in a week, constantly running. That's what I like about low population =]

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
A Hello, and a Question
« Reply #5 on: March 22, 2007, 01:05:45 AM »
That's pretty much what I learned.  They'll adapt, just never in the way you want them to.