Darwinbots Forum
Bots and Simulations => Evolution and Internet Sharing Sims => Internet Mode Commentary => Topic started by: EricL on December 03, 2007, 12:54:18 AM
-
Looks likes Flypaper 3.2 finally killed off Nasty Plant. There may still be a few left in offline sims, but NP now looks to be extinct in the currently on-line sims. Flypaper mutates, whereas Nasty Plant did not. It looks like after 300 or so mutations and 2500 generations, FP was able to adapt to the point where it could overcome the inherent advantage NP had in being an autotroph. Go evolution!
NP died out first in sims with day/night cycles. The nrg boost it got from being a veggy was mitigated at night and once FP was able to knock it's population down to the point where veggy repop could occur, NP slowly got edged out by native plant repopulation. Not being able to reproduce when you want is a large handicap, one which free nrg for nothing perhaps can't quite make up for, especially when you can't evolve.
One point of interest. A quick glance at the DNA of a representitive FP shows 3 instances of the new dropbool operator in a genome 178bp long. Pretty fast uptake.
Now I wonder what's going to be able to take out FP?
-
Any Lionfish 4 that crossed the line last night may have helped aswell, I added it two eyes so it stopped ignoring nasty plant.
I added communication to L4 -bots use out4 and out5 to announce the coordinates of food to other bots-, but as many algas are smaller than a full grown Lionfish and the feeding style of Lionfish is so aggresive, the biggest bot often ends eating unintentionally the smaller feeders.
-
I'm getting a bit tired of Flypaper 3.2. Check the numbers in this screenshot. Needless to say this sim didn't move many cycles through the hours it was running. Can't wait for CostX minus to be optional.
-
Ok, the FP rules the sims now for some time, last time I introduced a multiply it suddenly conquered everything, let now see if some of FP-population can be cut back.
-
I must admit it is rather surprising that a mutating bot is doing so well for so long. It is causing me to question some things I had long assumed.
I had long assumed that hand authorred DNA represented a steep peak in the fitness landscape with no gradual path leading "uphill". Said another way, I had long assumed that other than simple point mutations such as changing a shot type or similar, hand coded DNA would be limited in it's ability to adapt and would generally "devolve" radically, losing much of the hand authored capability after a few hundred generations. I had assumed that while a mutating hand authored bot might be able adapt to a specific specific threat, a specific hand authored non-mutating attack by changing its DNA in some small way, it would eventually become very vunerable to trivial attacks as selection constraints on things like guarding against various memory shots would relax in the absense of conserving selection pressure to maintain them and drift would destroy the carfeully hand crafted defenses. Perhaps this is indeed happening and we just don't realize that FP is now vunerable in some trivial ways....
-
.....I had assumed that while a mutating hand authored bot might be able adapt to a specific specific threat, a specific hand authored non-mutating attack by changing its DNA in some small way, it would eventually become very vunerable to trivial attacks as selection constraints on things like guarding against various memory shots would relax in the absense of conserving selection pressure to maintain them and drift would destroy the carfeully hand crafted defenses.Perhaps this is indeed happening and we just don't realize that FP is now vunerable in some trivial ways....
I am pretty sure it has weaknesses, I am not sure what opponents it gets, as I can't connect to IM somehow. On the other side if the diversity is large, there are probably bots left that still have an counter-defence if an new tactic of a bot turns up, still pretty amazing a mutated bot keeps his place. As in my opinion it has taken control a little long, I can't wait to put some dangerous venow, poison, virus-shooting and tie-feeding to get it out.
-
Can't wait for CostX minus to be optional.
2.43v.
-
I'm getting a bit tired of Flypaper 3.2. Check the numbers in this screenshot. Needless to say this sim didn't move many cycles through the hours it was running. Can't wait for CostX minus to be optional.
That's why I always run sims with fixed bot radii. And my Flypaper has most certainly little in common with yours.
-
That's why I always run sims with fixed bot radii.
Hmmm. Perhaps something else we should disable in internet mode....
-
That's why I always run sims with fixed bot radii.
Hmmm. Perhaps something else we should disable in internet mode....
Why as long it is just his sim, hmmm, it isn't a goodpoint everybody is talking about 'his' flypaper, it means it is everywhere in any sim, okay destroying bot has to come.
Edit: It would already have come if I could connect
-
Edit: It would already have come if I could connect
Try restarting DB. The VB Inet control gets into a funny state sometimes. I have yet to find a way to get it out of that state other then to tear down the process...
-
Wow. Last night before I went to bed, Lionfish 5 had taken over EricL3 and wiped Flypaper out completly in that sim. I thought for sure I would wake up this morning and LF would have colonized the others and own IM. But FP has retaken the sim and Lionfish is dead there and in my others. I wish we had the tools to see whether this was ther result of FP adapting or LF devolving (does it even mutate? I can't find one to check!) or something else...
-
Shape awareness.
I've been too lazy to number the LF5 subversions. Anemone also helped. Still, I'm totally amazed on how efficient is FlyPaper. And with mutations on.
LF evolves but it doesn't mutate. LF4 had 4 evolving variables and LF5 has 13 or so. They all are stored from 971 to 983. Two LF's can have quite different strategies one from another yet changes won't wreck them.
For LF5 I had to code it back to true conditional code as many genes were using too many conditions already- That's a limit of 1G coding, or at least from the human readability and debugging point of view.
I stole also the bodyfeeding gene -I didn't knew you could use -6 for .tieloc too-. When I saw stickerbush 'deflating' bots I thought it was injecting values in .fdbody.
LF5 is a bag of tricks, I put in everything that sounds cool, but it probably has so many overkills that it leaks tons of unoptimized energy. It has 31 genes and still some 1G code.
-
Woah, LF5 died again.
What kind of evil settings those sims have?
-
What kind of evil settings those sims have?
Pretty vanilla. Autocosts with F1 settings minus DNA charges and a few shapes.
-
Looks like LF5 has got it this time. It's on top now and owns 3 of my four sims.
For LF5 I had to code it back to true conditional code as many genes were using too many conditions already- That's a limit of 1G coding, or at least from the human readability and debugging point of view.
Does the new in-body conditional paradym help with this?
-
Uh oh... Looks like FP is retaking EricL2...
-
LF was down to 100 when I disconnected.
Add a lousy support for my wireless card on Linux and running DB on Wine...
So LF was totally on its own. I betcha it has gotten sloppy not too long after it dominated the sim. Without pressure it's fighting against itself and prolly spoiling its fitness.
(no, I haven't looked yet at the new thingies... I didn't even look at how shape vision worked till yesterday, yuck.)
-
Variable names don't show in the robot inspection window, so just in case you wanna look at what's going on inside a Lionfish racial memory, these are the variables:
971: turnfactor - angle of turning using all eyes.
972: reprothreshold - LF normally reproduces when .mass is over this value, and energy over mymaxenergy.
973: speedlimit - the highest speed it'll attempt to reach.
974: reproshare - the value before .repro.
975: mymaxvenom
976: mymaxpoison
977: mymaxslime
978: mymaxshell
979: mymaxenergy - energy to reproduce, body management attempts to keep energy at that value.
980: myfareyes - how wide will be eye1 and eye9.
981: mybodyfeed - how much will be placed on .tieval when bodyfeeding (negative).
982: myvirusshot - value before .vshoot.
I did a mistake and placed .imasettler outside racial memory, but it will be 983 for the next version. imasettler activates a secondary reproduction gene when .totalmyspecies is lower.