Experiment in Evolution.

After reading Philip’s blog, it refreshed some thoughts about evolutionary experiments I had been pondering some time ago. It would be so cool to run such experiments in SL, but is it possible?

First of all you would need some kind of base unit or cell. This unit would have to be coded to thrive or struggle, depending on various conditions. It would of course also need to be coded to have the ability to mutate upon replication, a vital ingredient.

We would also need to create conditions. In the real world we have evolutionary pressures such as – temperature, amount of sunlight, humidity, presence and nature of predators, availability and nature of water/nutrients, amount of/nature of the space it exists, the existence and nature of organisms that share that space (not just those above and below the food chain, but organisms that thrive in co-existence, each offering the other something advantageous e.g. Lichen). How would we simulate evolutionary selection pressures in SL? There are SO many more questions that this all raises…

How do we code in the mutations? In the RW, when a mutation occurs 99%+ of the time it’s disastrous for the organism, yet the rare beneficial mutations are vital.

What conditions can we create to simulate evolutionary pressures? How do we code these pressures into the ‘cell’? Also, we would need to find a way to have these conditions fluctuating from time to time, and different from area to area, to give diversity a chance and avoid an evolutionary dead end.

Could we code in major events? In the RW we have disastrous sudden events, and those that take place over geological time periods. The impact of meteors and major geological events, such as super volcanoes, has shaped evolution in a very short period of time geologically. Events that take place over longer periods have been equally important. Continental shift - offering isolation, the tilt of the Earth’s axis – altering the climate and inducing ice ages.

One of the most interesting steps for me is the interaction between different ‘species of cells’. How would you simulate predation and the food chain? More importantly, how would you code in the possibility of multi-cellular organisms arriving on the scene and thriving.

To make an experiment like this work, I think you would need a LOT of sims. I don’t think one sim would be enough space to offer a variety of conditions and space to grow, it would be too isolated.

So I ask the scripters here, is it possible? I would love to hear everyone’s ideas on this 

One thing that I am struggling with, if we are trying so desperately to create a code that allows the possibility of all the above to happen, how much of it is evolution, and how much is by design? A question not limited to SL 

Looking forward to reading your thoughts.

Hiro, it's all fascinating, and people who understand the coding could better understand you, and yes, it sounds like it would take lots of sims and being able to build up in the air high and down in the ground low, etc.

I would like to suggest that you could discuss this issue perhaps not quite so literally, i.e. with single-cell creatures that you create as some kind of AIs or NPCs or something in the game.

As you know, anyone who buys a 512 automatically enters himself in one of those amazing and fantastic Social Darwinist experiments of the 21st century. He faces all kinds of extraordinary pressures whether lag, or griefing, or bad builds. As he struggle to survive, he adapts in all kinds of extraordinary ways with building skyboxes, putting invisible tiles on this house or buying the next lot and linking it, etc. etc.

The effort various unrelated people make as they try to buy and govern pieces of sims, for me, supplies all the data you need ready-made to study Darwinism at work, and a lot of other isms too!

Already being done.....



I have fish that shoal together, they stay in the water (mostly), detect sims edges and stay way from them (mostly). Their behavior is a result of several generations of 'forced' evolution, ie. I make minor changes to several fish, add them to the population, make different changes to other fish, add them to the pouplation, then see which ones perform the best and select them.

I am currently working on DNA code to make these minor changes for me. Two fish breeding would produce offspring that contained parts of each parents DNA with a random number of params altered by a random amount (within design limits). DNA params would include swim speed, size, muscle power, sight distance, sight angle, anything else that affects the performance of the fish.

There are many aspects in developement which I don't have time to write about (I'm at work...) but it is both fun and frustrating all at once.


(Check out the current generation of fish in Aretias or Hypatia)

Jeez, kettle, black! from someone who outright refuses to summarize / shorten his posts, you DARE ask him to Layman's terms his thoughts?

I thought Hiro was wellspoken.

You know, you've totally misinterpreted that.

As for Hiro Queso's post, there was a guy (I don't remember his name now) doing something called Genprims that evolved. It was quite fun but I haven't heard or seen him for a long time so I'm not sure if he's still working on the project - your post has made me want to check that out again.

The whole subject fascinates me. I majored in Neural Networks for my degree and the idea of learning, teaching and evolution has always sparked an interest with me. I'd love to do something down this route in the future, RL permitting.

Yeah, this is definitely possible to do, and it's actually something I thought about trying. You'd want to introduce some kind of maximum distance moved and food source... From a technical standpoint, sensors and volumedetect would probably be the way to go.

You could have a random number to diverge from the parent's movement/energy usage/colour attributes... it'd be pretty cool, actually. You've got me looking at doing this now.

I've been thinking of doing something like this myself - I wont say in what specific way, but more than just experimentations. The scripts for the free-swimming fish and floating lilies (you can see them at Moss Beach in Hooper) uses a rudimentary fuzzy logic system to allow for random movement while still preventing them from drifting/swimming away.

In either case - perhaps this thread would be better served in the Scripting Tips forum? That's a generally more script-knowledgeable forum than here.


- Newfie

I thought we were discussing evolution within SL, not scripting tips...

Artificial life is a fun and fascinating subject and it's something I've been into for many years. Unfortunately the limited memory available to scripts and the heavy physics load that items of this kind tend to require are prohibitive in SL at this time. I'm really hoping that havoc 2 will be more generous with physics. Memory problems can be worked around by using many scripts, but when I see animals in SL containing up to 25 scripts scooting around bouncing off every wall, I must say it makes me worry a little for the sim they are in. I believe SL will eventually improve to the point where people like us can truly make some wonderful evolving alife.

Considering that the bulk of his post was to do in the scripting portion of it - and digital evolution is pretty much a 95% algorithm-based (aka 'scripting') process - then it still seems to fit within the scripting tips area.


- Newfie

It should, in theory at least, be possible to off-load some of the coding externally. To what extent and in what manner this could be done is obviously open to varied opinions, and there are obviously issues with doing this (such as delays etc.)

I'm also curious about the use of physics and whether that would truly be necessary. As you quite rightly say physics isn't a high point in SL's feature set at the moment, so maybe working without physics would be a better solution until physics was reliable and streamlined enough to be able to cope. It depends on what you're trying to achieve. For instance the fish I've seen (correct me on yours Surina, if I'm wrong) don't use physics at all.

Does anyone remember the old program called SimEarth? That was so fascinating. By doing some changes during the evolution you could make say ocean bound animals to end up the dominant species on earth.

This is probably off topic but it did bring SimEarth to mind

The tradeoff between bandwidth used to move non-physical items compared to the processing power required for physics ( at this time ) is worthy of a seperate thread maybe ! It's certainly something I'm interested in discussing.

I've been reading recently about Will Wright's latest "game" called Spore - this takes modelling evolution in PC Games to a whole new level and sounds very intriguing. There's a good article and screen shots of it here: http://www.gamespy.com/articles/595/595975p1.html

Yes, you're absolutely right. Maybe it would be an idea to setup an in-world group and get the Lindens to give it a seperate group forum on here, then we can discuss all the issues to our heart's content. What do other people think about that? Actually, that's presuming there isn't one already.

in theory the cell needs to respond to its environment. mutation in a sense is a response to its environment. response range from simple "chemical" responses (e.g. cell encounter "element" that disrupts its "cell membrane" and the cell "dies" to more complex reactions like searching for something edible.theoretically, you only need to put the "physical" attributes into the simulation, things like predators and symbiosis are suppose to evolve under the "environmental" conditions of the simulation.this is pretty straight forward, you create a stimulus that triggers a mutation.you could code them into the cell, but usually by limiting the resources, and making the environment changeable (usually in a cyclic way), evolutionary processes are supposed to come into play.sure. just write them as major environmental changes.if your evolution is fast enough... predators should evolve. (e.g. they "discover" that other cells are compact sources of live resources aka food).depends, but certainly you'd be limited by the number of prims unless you were just going to do simulations in code.yes, it's possible.

you might want to consider if you are dealing with single-cell organism cells or multi-cell organism type cells. and which conditions you start with (e.g. things like biofilms)

One thing that I am struggling with, if we are trying so desperately to create a code that allows the possibility of all the above to happen, how much of it is evolution, and how much is by design? A question not limited to SL 

Looking forward to reading your thoughts.[/QUOTE]

actually that sounds on topic to me... could do a model off the sim earth model

Ah btw, the guy behind Genprims was Rhu Altman but it looks like he hasn't been in-world since January unfortunately.

Yes I still have a copy of SimEarth.

That was back in the early days when Maxis and Will Wright were developing only on the Macintosh platform - oh yes folks, Sim city was Mac only. Hell thoes were the days when you could call Maxis and talk with Will on the phone.

Hehehe.

Toy you get the Distinguished Digital Classic Oldie Remembrance Award of the month!

The biggest problem my fish have is the physics. Rather than trying to write my own collision detection algowhatsits I use the SL physics. This means the fish don't swim through other objects, but the first few fish in a shoal that encounter an obstacle are stopped by it and the following fish automatically swim around. It's looks very natural.

I also use the physics for movement, applying force impulses in calculated directions.

Ah right, I didn't realise that. I've seen other fish that don't use physics and kind of presumed that most wouldn't. I remember when you let some of your fish off on Numbakulla when it was being used as a sandbox - they were great little critters

One thing I would like to see eventually is more input events to LSL. For example the ability to sense light and sound sources and intensities.

My fish would love that, I think their biggest problem atm is boredom...

You could mimick this to an extent by having objects that also chat out on specific channels information that the 'critters' can understand.

LOL

I have some birds that people can feed by rezzing seed on the ground. The birds come down from the sky, hop around and peck up the seed. Perhaps your fish could benefit fro being fed, maybe some floating food ?