Portable Genetics Format (PGF)

Hi!

I'm a Systems Specialist, with a strong background in virtual environment research. I am quite familiar with AI, AL, robotics and all forms of advanced mad science.

I would like to work with the alife commmunity to define a standard, efficient, portable genetics format that will be applicable to all forms of artificial life.

Let's begin by providing lists of characteristics which you feel are necessary when defining artificial lifeforms.

As we are desiging this format, we will examine the most efficient methods of encoding the data, and develop the format to allow for a core set of characteristics, which will work for all forms of autonomous creations. Built into this format, we will then allow for an unlimited amound of expansion, based on species.

If we look at creatures from a taxonomical level, we should be able to define a method for DNA use in second life, and elsewhere, through the development of a cross-platform compatible genetics format.

Please begin by listing your needs.


Thanks,

Klaus Bertone

What might be useful would be an examination of the existing methods by which such information might be stored, manipulated and propogated within SL. And what limitations these techniques might have. SL has a great potential to be a platform for experiments in AL, but it is early days.

I'm not sure what the makeup of members of this list is, with regard to previous experience or just interest and a desire to experiment/explore. As this is your first post, and you only joined in the last few days, could I suggest that you might introduce yourself, and your motivations before attempting to become the standards guru. There are several groups in SL that you might want to join as well.

Search for "artificial" in the Groups find tab will give: Artificial Life, Artificial Intelligence, Artificial Second Life Institute for example.

Maybe it would be useful to raise a debate as to whether people want such a format, rather than developing their own. However, this does lead to a general question of whether people want to collaborate; this is a commercially sensitive environment after all, with mechanisms for protecting constructions, as these could become viable products. It would be good to share experiences/results, but this isn't something we should just assume. Even in the academic community there are some details that are reserved in publications.

The biggest problem we have currently within the LSL language is that there is no way to hard store data. You can't write to a note card and there is no disk storage available. The only way of permanently storing anything is through an external server linked with XML-RPC or email.

To me, the point of autonamous AL is to make them independant of external influences such as this. I don't want all my creatures dieing because a link to an external server went south for winter...

So we're left with persistant memory which limits us to 4 byte numbers, text strings or lists. In other words, our DNA model needs to be as simple as possible.

Currently I'm using the 24 bits* available in the integer type as this can be passed directly from parent to child along with the script. The individual bits are also given multiple meaning, ie 3 bits used to pass an impirical value can also be read as boolean values to turn on or off other unrelated features in the child.

I've also experimented with overlapping bits, which produced some interesting results, though not usefull at the time. This is where the child reads bits 0, 1 and 2 as one value, then reads bits 1, 2 and 3 for then next value, etc,. Using this method you can load a large number of variables with data in the child, however the more variables you "create" from the data, the less control the parent has when modifying the DNA being passed.
The specific advantage though is that certain behaviour aspects will become automaticaly linked without having to write any code at all. Imagine 3 behaviours where the third is dependant on the first two. You pass 4 bits each for the first two behaviours, defining a value 0 - 15 for each one, then the child creates the third value from 2 bits each from the defined DNA. so if either | both of the defined values change, the derived value changes in responce. The derived value can also be weighted by chosing which bits and how many to use.

Having a common DNA standard for AL has been discussed in this forum before but never really got anywhere. Personnally, the more I think about it, I'm not sure if its a good idea or not. By definition you would be restricting future AL projects by pre-defing DNA. While the standard would have to be expandable to make allowance for future progress, any future progress would be forced to recognise the past entries in the DNA standard, thus imposing limitations.


*Edit: I of course meant 32 bits... it was a long day...

Hi!

// =====================================================
// BEGIN ROLE DEFINITION AS STANDARDS GURU
// =====================================================
Firstly, Mr. Mike, I have read all of the posts to this forum, and have just as much right as any researcher to push the envelope of the field. I do not intend to ever dumb down my posts, just so you can read them. It is my sincerest wish that you take your own advice, and consult a lexicon if something is not clear to you. I'm terse. If you want excess verbiage read Steven King.

There is no room in my thread for your politically correct intelligent design nonsense. I get paid handsomely at work to make large ideas fit in small heads. I do not intend to take the trouble when I'm doing this for MY own enjoyment. Your corporate doublespeak is quite effective for treading water, if that is your preferred method, please move aside so others may swim freely. Artificial Life is about EVOLUTION, plain & simple. It is not a theory when it is hard-coded.

You sound like the bureaucrat in DA HHGG, who posed the question, "Maybe people would like fire fitted nasally?"...

I've been a bare-metal-hacker for 26 years. I have a DIY attitude. If something is broke, I fix it. Hand me a roll of duct tape and get me a coffee while you're up.

Thaaaank you.
// =====================================================

Hi Everybody else, what a beautiful day in the neighborhood, nice weather we're having, eh?!

Quite frankly, Surina really knows her stuff, and is one of the obvious core members here. I would like to work directly with her, Johnny Noir, Argent Sonecutter, and any others who are up to snuff, to design a protocol & working methodology for A-Life genetics. It would be nice to have something this important to future technology be the offspring of SL.

As far as meat & potatoes go, 32 bits can go a long way with bit-wise manipulation. It's actually better that we are limited, as there is no option for bloat. As far as scripting space, 16K-64K was enough room for all of the classic Atari video games, so unless you feel you can wear their boots comfortably, it's silly to complain about it.

I've invited Jed Margolin, the creator of Atari Battlezone to join SL, and will be directing him and other talented programmers to this group, provided that the RL personalities here are for real, and don't apply the (G)EEK word to themselves too literally. A hacker is one who makes the impossible happen, a (G)EEK is a luser who's purchasing patterns inspire Japanese clothing designers to make pockets deeper in their domestic clothing lines. Why worry about if someone will confuse you with a criminal... reclaim the word, and use it proudly.

Over the course of this week I will be putting together a draft document which will outline taxonomical attributes. I would like to see further fleshing out of the characteristics which you, as local experts, feel to be most necessary. During this time, please brush up on your Boolean logic, and share any creative ideas which spontaneously come to you while doing so. Let's also try to solve the persistent storage problem by creatively using features of LSL in ways they were not indented to be used.

A benefits/drawbacks list regarding email would help us to really evaluate what it COULD be used for. Remember, a gestation period is normal in all living things, instant transference of genetic material may not really be necessary, instead, just set a flag that something is pregnant or is maturing, then send email with a better defined genetic makeup... establish a timeout period, in case of server instability. If the email doesn't come for a while, the lifeform will remains pregnant, or has a prolonged maturation period until it receives further instruction. This is just a 'for instance', lets examine all of the good & bad in the email method.

As described in my first post, there is no built-in limitation. We will define a core set of attributes first. All creatures will have this. Next, add-ons will be applied in a taxonomic fashion (Kingdom, Phylum, Class, Order, Family, Sub-family, Genus, Species, Sub-Species...). If you have no idea about what I'm talking about, check out a book on Biology, and re-read...

This will allow for a complete database of animal types. It is quite likely that if we invest the time doing it right, and prepare specifications documents, we may have our ideas incorporated into SL itself. This would allow us further freedoms to define how artificial lifeforms fit into the overall schema of SL.

The important thing here is SCOPE. Think scope!

Local attributes, global attributes... think on these things.

If we do this right, we can let RL automata (robots with wireless email) gain functionality from SL evolution experiences. At this point, we write our own ACM entry.


In closing, those of you who would like to try to get a working standard together, join in.

... & Mike, please loosen your tie, to let the blood flow & we'll get along splendidly.

= )


L8R,

Klaus Bertone

uh...on an unrelated note...you should be just as proud of being a geek as you are a hacker. Mind set isn't all that different.

Hi!

Let me politely clear things up.

Geek==Dork
Geek==Nerd
Geek==Loser
Geek==Dockerboy
Geek==Consumer electronics hording goatee bearing Captain Kirk
Geek==Goaty
Geek==Billy G.
Geek==Gofer-do-this-that-for-half-wage
Geek==Visual Basic Programmer, who says it is a programmer
Geek==Tape those glassess together, oh yeah!
Geek==McFly
Geek==It's ok to call the geek-squad, or any other technical support, for that matter.
Geek==I use an IE & word, ergo I'm a computer user
Geek==ipods are K-RaD-a-koO0L!


jhsgfiuergknrb, c
zcsvbdb
^^ *!
xbfdbnnngmhfmmmn ..sdxg===-sg~ SYSTEM OVERLOAD

ARRRG.

The main difference here is that as a hacker, I'll be the first to point out that "mind set" is all one word, or should at least be hyphenated.

Lets try to keep this thread focused on hacking, not 'geeking'.


END OF TRANSMISSION.


Thank you.

Klaus

Hi,

I kinda watch all this alife stuff from the sidelines since I'm partly too busy and partly too lazy to get involved at the moment, but it interests me hugely.

This post is partly to say 'hi', since I might try and join in to whatever extent I can later. The second part of it is to say this:

Different people have different defenitions of words: some people I know refer to the meal in the middle of the day as dinner, some refer to their evening meal as dinner, some use the term to refer to whatever the biggest meal they're eating that day is. I've found lots of people use the term geek to describe someone who has a very stong interest in something, not necessarily technical even. And lets not forget, the term originally refered to a type of circus performer.

Sorry, just had to get that off my chest - people thinking language is set in stone is one of my pet peeves.

Thirdly, I'd like to suggest the idea that a standardised genetics format should treat prims simmilarly to cells or organs (not literally, that's just the best way I can think of summing up the idea). Don't know how that fits with standard alife practices or what anyone is intending to do (or even if it's been discussed already), but something about it screams out to me.

Klaus. I thought I'd wait to respond to your second post. Wait until I could make some sense of it. I've read it again. I've read my post entitled "A suggestion". I just can't see where your comments link to my own. Sorry. I've tried. It just comes across as factually inaccurate, hurtful and insulting. Was that your intent?

If you are refering implicitly to the 'off topic' forum post I made, entitled "Evolutionary Design" at:

/112/d7/71778/1.html

then you may have missed the point, as it was satire in the context of several heated debates there on Intelligent Design. I think an apology is due. This forum hasn't had a flame before. It seems a shame that this is the first.

Hi!

An apology! A flame! A shame?!!

Egads man, get a hold of yourself!

Re-read your post and count the # of times you declared that I should do something. Of course I take offense.

I'm trying to do something productive here, and instead of soothsaying, you are naysaying.

It should be rather apparent that my use of the terminology, "Intelligent Design", was metaphorical. Based on context, it should be obvious, that it was directed at the idea of "waiting for the gods to hand it down from above, and then maybe it will evolve...". Not the gods on Mount Olympus , mind you, but those who dwell in the corporate campus & the ivory towers.

(See, now you are making me dumb my posts down... and you wonder why you stepped over my irk-threshold ...).

As should also be obvious, I don't like to work that way. If everyone was like this, (as our RL society is surely heading toward...), nothing would ever get done, because you have to waste time waiting for someone else to tell you how to do something. I like the direct approach... use the scientific method, and be done with it.

If Thomas Edison had waited around for someone to tell him that he was doing the right thing, you would be sitting at home reading a book by candlelight right now.

Even if someone else does the hard work for you, who says they will do it right, &/or let you use it for free?

It should also be obvious that my posts contain tongue-in-cheek humor. Who else comments text blocks (gawd.) I believe that I did say that we could get along splendidly, provided that certain conditions were met, did I not?

= )

Didn't you even notice that you yourself suggested the title, "Standards Guru"?...

I'll let you slide since you are, after all, a mac user.

Still, being an islander, I would have presumed that you would be familiar with Douglas Adams...

But, I'm still waiting for my coffee, & while you're at it you can fax me a pizza.

Now, enough of this nonsense. The next post from you better have some detailed technical information relevant to this thread, instead of the banter of non-productive critique, or it's straight to the headmaster with you.

= )

hee.

Look at it from my perspective, at any given moment, it is quite likely that I am at least one gallon of coffee ahead of you for the day... Of course everything is really funny and/or extremely irritating all at once.

Perhaps you should try out my methodology, it is quite effective at solving 20,000 things simultaneously, and doesn't have all of the side effects of modern drugs ending in 'a'.

In any case, what the hell???, I don't see anybody else posting a resume here to verify credentials.

It should be obvious from the first post that I have the background knowledge to put together a data format & specification. Why would I bother if I didn't? In fact, why would I bother to join a specific group & post at all? Time is precious, man... we're all tombstones waiting to happen. I only give a timeslice to activities that I value.

That being said, its time for you to stop being the sensitive-artist-type. It wastes your time & my time, & in the immortal words of Jeff Spicolli, it is wasting OUR time.

Get back to work.

= )


Klaus

Hi!

I presume that you are referring to an attempt to fax me a pizza.

Windex & Goo-Gone should help significantly, then try again with less cheese.

Thanks for trying, you now have a taste of the futility that millions face daily.


Have a wonderful day.

Klaus

In an attempt to get this thread back on topic I have a question.

Why?

One of the defining properties of spiecies in RL is that while their DNA may follow a shared format, it isn't actually transferable from one to another. You can mate a pig with a sea-gull as often as you like (or are able) but you'll never produce a flying pig...

DNA only has a common format because all creatures share common ancestry. Also, the DNA itself is more than a simple set of variables that determins what a program should do, it IS the program.

So the LSL scripts we put into our cute AL creatures are in themselves the DNA and we already have a standard for LSL.

Hi!

This is a very good point, especially regarding the current state of SL.

After a lot of thought on the matter, I had arrived at similar conclusions, based on the way things are as of now.

However, to be able to be forward thinking, it might behoove us to look at how we would like automata to behave in the future, within SL & elsewhere.

SL is nice, because a full lab is available to the experimenter. If we take advantage of this, and prepare a taxonomical de-facto standard, it could become part of the future of SL. This could be done by arriving at a base standard (which allows for add-ons in a common format), and publishing these creatures to the script examples web site. This would effectively build a database of automata which follow shared standards & can be used in many sims. Essentially, it would be a common hereditary point.

As I had mentioned earlier, the ability of these creatures to email to the outside world, could, in fact, be useful for passing learned behavior to other online or robotic automata.

There are certain pieces of software (El Fish & Sim Life) which have successfully allowed for interbreeding. Maybe we can get those software engineers involved. There was also an online world about four years ago that allowed creatures to be created, & released into an environment. These creatures would send periodic status reports to your email addr when significant events occurred in their 'lives'.

I would say that the eventual goal of creating such a standard would be future integration into the SL software, where certain creatures could be indigenous to certain regions, and could be purchased when purchasing land.

More on this later.

-Klaus

Some interesting ideas here,

In practice wouldnt the DNA be a combination of scripts and attributes?

Suppose it was agreed that each creature has two functional scripts. The BRAIN script decides where to move to, and a MOVE script which physically moves the creature. You have a standard command format so that any BRAIN script can tell any MOVE script where to go.

In addition each script has a gene sequence or 'string', a set of attributes set to certain values. For the brain script this might take the form of how far the brain should decide to move, and for the move script, how fast it moves.

A minor change would be to exchange gene strings, so the move script travells a little faster. A major change would be to actually change a script. That is, the creature drops the roll along the floor script for one that hops.

One advantage of that approach would be that creatures could end up with a mixture of scripts written by different people, which could be interesting.

The disadvantage is that I've increased the number of standards to be agreed. You would have to agree a standard set of script functions (Brain, sensors, movement?), a standard to allow those scripts to talk to each other and a standard gene string which may be different for each script type. Hmm, I may have just talked myself out of this idea.