[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: 'not an artist' article
On Tue, Jul 04, 2000 at 11:34:51AM -0500, Erik wrote:
> On 04-Jul-2000 Chris Purnell wrote:
> > On Mon, Jul 03, 2000 at 07:26:37PM -0500, Steve Baker wrote:
> >> Yep - that's the $64,000 question.
> >> I spent some time today thinking about it - and I'm pretty sure it could
> >> be done.
> >> The snag is to come up with a set of genes with the right 'powers' over
> >> the mesh to come up with credible critters - without limiting the resulting
> >> shapes arbitarily.
> were the problem specified correctly to the algorithm, no arbitrary limits
> would be required. But making a lifeform is not a simple problem to define :)
Taking Steve's example where the problem is "make me a cute bunny rabbit".
No sane person here is going to attempt to specify that to a computer.
The algorithm we have chosen is genetic evolution using a human for the
selection. The remaining problem is one of implementation. Arbitrary
limits should be used if they can simplify that. They may also reduce
the number of generations required to get a useful result.
> > I've also spent some time thinking about it. I think any thing limiting
> > the results to human and anthropomorphic characters would help the time
> > it takes to get a useful result and may also help programming it.
> > I liked your idea about evolving the bones. I think fixing the number of
> > bones and their conections is the way to go. And having the genes control
> > the length of the bones. I'm not sure about constraints about the relative
> > lengths of the upper and lower parts of the limbs.
> If you define the bones and their connection as a static thing, you limit the
> power of the machine to generate soemthing unique, as well as requiring the
> programmer (or modeller) to model a functional animated skeleton. That kinda
> defeats the purpose imho :)
I think we should keep form and function seperate. If you to use an
evelotionary system to get it to learn how to walk than I think that
should be a seperate program.
> > This is where I get stuck. Putting the flesh on the bones.
> There are plenty of articles around for putting flesh onto a bone structure
Can you give me any pointers to look for them? URLs, search stings, etc.
I'd want something that could give my human characters a realistic looking
> > I was thinking of enforcing left-right symmetry so we only have to
> > generate half a character.
> one of my fish is not symmetrical... Maybe instead of enforcing an arbitrary
> limit, simply introduce a symmetry component to a gene? I d'no if a symmetry
> gene would be useful, since those few creatures in nature that are no entirely
> symmetrical are still primarily symmetrical.
No creature in nature is entirely symmetrical but the differences are
usually down to imperfections in the grouth process rather than the
And the only left right asymetry in my characters that I'd want would
be in their hair style.
> Has the genome project released any technical documentation? I remember hearing
> that they had completed something, mebbe it was just the human DNA? Fruit flies
> were figured out quite a while ago (14 genes iirc). Since they are examining
> what we are trying to emulate, maybe we should look at their results and see
> what they found the basic building blocks to be and how they find them
> attached. I'm sure the most natural results would come if we could use nature
> as a blueprint :)
If you start using real genes as a bases for your virtual genes you may
find that you need to grow each individual in each generation in the same
way that real individual. That is will cells. This may be the way to go
only using a lot less cells that in real animals.
Christopher John Purnell | A friend in need's a friend in deed
http://www.lost.org.uk/ | A friend with weed is better
--------------------------| A friend with breasts and all the rest
What gods do you pray to? | A friend who's dressed in leather
To unsubscribe, e-mail: firstname.lastname@example.org
For additional commands, e-mail: email@example.com