Q) How big is the DB morphospace?

[EricL]

I think there are 66,571 unique "base pairs".

2^16 possible numbers
1001 possible sysvars  (including 0)
10 "basic" commands e.g. add, mult, etc.
7 advanced commands e.g. andgle, dist, ceil, etc.
9 bitwise commands e.g. >>, ++, etc.
10 conditionals e.g. > , <, <=, etc.
4 logic commands e.g. and, or, not
3 commands that store to or manipulate sysvar locations e.g. store, inc and dec
4 flow control commands cond, start, else, stop.

And I'll throw in "end" just for completeness.

2^16 + 1001 + 10 + 7 + 9 + 10 + 4 + 3 + 4 + 1 = 66,571.

(Really, we should probably call them "codons" not base pairs, since some are degenerate.  That is, there are mutliple ways to code for the same result.  Sysvar type base pairs outside 0-1000 for example can exist in the DNA, but they get MODed when used.   But I'll continue to call them base pairs here.)

So, while there are actually more than 66,569 different base pairs, some of those map to functional equivalents.  If we are trying to calculate how many unique morphs are possible with the current DB DNA language (which I am) we should probably treat the polymorphics as all being the same.   Different genomes that code for the same morph might mutate differently since their coding sequences would be different, but the morphs they code for are identical.

The max genome length today is 32000.  Thus, there are 66,571 ^ 32000=  3.2528 X 10^15434 possible morphs.  Thats a very very big number....

[shvarz]

That's because your calculation is not correct.  To calculate the number of possible genomes you need to bring 66,571 to the 32000-th power.  The number is enormous, although admittedly many of the genomes are not going to be even remotely survivable.

[EricL]

That's because your calculation is not correct.  To calculate the number of possible genomes you need to bring 66,571 to the 32000-th power.  The number is enormous, although admittedly many of the genomes are not going to be even remotely survivable.

Damn, you caught me.  I caught it just after I posted...  

[Numsgil]

I think using the entirety of the numbers as 64000 possible values is a little naive, since alot of sysvars work on either the entire domain or a significant subset of it.  But I don't know what other value I would use to replace it.  But they definately are comparable to codons instead of base pairs, so any comparison between real DNA needs to include that fact.

But yes, genome space is very huge.  Hence the phrase "open ended".   The majority of that search space is made up of sysvars and numbers, which leads to the interesting idea of: is it possible to create a numberless ALife DNA?

[Jez]

is it possible to create a numberless ALife DNA?

Should be, as long as you are not including the removal of things like *.eye5 I could do it. .repro for instance, could become;

*.robage *.eye5 div .repro

or something. If I've understood the question correctly - it would just mean using values or value sums rather than numbers.

[Numsgil]

But the sysvars are still basically numbers.  You could collapse *.robage *.eye5 div into a number.  I'm just wondering how you would build a DNA language that would let you do everything that you can do in Darwinbots without using any numbers, or at the very least hiding all the numbers from the DNA.

[Jez]

I don't know, we use maths to model our universe and the bots. I guess you'd have to think of another way to model and interpret the bots universe.

Lets say colour for ID, repro only as set figure or equal division of self, no distance or movement value (only can see summat am moving). one eye or all eyes working as one.

Yes then, base it on yes/no values rather than scalability. Probably more limited unless you used a sort of infinite tree of yes/no values.  
(am I moving) [yes] (am I moving as fast as target) [no] (am I moving faster than target) [yes] (can I move faster than this) [yes] etc

[Numsgil]

That's interesting.  I could see alot of DNA being built on boolean values.  I'm not sure it would be as versatile as the present DNA though.  Ultimately I think if you wanted to depart totally from numbers you'd need to use something like a neural net.  Various inputs matched up and modified to varirous outputs.