Hey,

Could I use the PDG to set up some simple artificial life simulations?
So I create some ‘creatures’ with randomised parameters which have to achieve a goal, say get across a line.
Then if they achieve this goal they can be used as a template for another generation of creatures. And then keep iterating.

I suspect this is way beyond my technical level but could be an interesting design approach perhaps.

Hi Simon,
Is this a practically oriented question (can it be done in principle or in practice with TOPs) or a more academically oriented one?
You're right in thinking this is technically challenging, although if you google “virtually evolved creatures” you'll find lots of demonstrations, like this one [www.youtube.com], some newer and some as old as computers.

I feel confident in suggesting a different approach if you're looking to come up with solutions of credible creatures that would be both artistically and biologically believable. Something that is not as embarrassing as the “horses” in the Avatar movie, which has two pairs of front limbs. Usually, redundancy in biological creatures is not present unless said redundancy is worth more for the survival of the individual than the cost of keeping it.

Nature's showing us that the individual is not worth a lot, so a lack of anatomical redundancies is a lot more cost effective for the species' continuation than equipping the individual with redundancies, which would assure a longer life span, but which would also result in a less diverse gene pool due to lost opportunity of gene mutation. So this would be another reason for why individuals are neither immortal nor designed to live for a longer time than “necessary”.

Aside from studying biology and evolution, general knowledge of maths and physics will go a long way. Since volume varies by cubic power with length, it's obvious that shortening and thickening limbs would provide a much smaller area for heat dissipation while keeping a constant volume, which is great if you live in a cold climate. Having longer shins and forearms, if you're a creature that makes ends meet by running, will get you greater linear speed at the ends of your limbs for the same angular speed, which makes you run faster, obviously. Displacing a volume with a medium less heavy than the medium of the displaced volume, will get you lift (that's why “atmospheric whales” concept is stupid). I could think of other things and these are just basic physics considerations, but if you're gonna consider biological realism, most Hollywood movies will seem ridiculous if don't also turn in your brain at wardrobe along with your coat. So to me it seems that we have to educate ourselves, the artists, or movies will keep doing what a director says “yep, that's alien enough” (to a blob) or “realistic enough” to anything for that matter.

Evolution and its mechanisms is a hobby horse of mine, so excuse my rambling, but it's fun to do so sometimes
If you're looking to generate virtual creatures for academic purposes, roll up your sleeves and prepare to give up a good portion of your career to this. Otherwise, if you're just an artist, consider what I've said above.
Cheers!

Thanks for your answer.

I just want to experiment with it rather than create plausible VFX creatures.

But a simplified version of my question would be;

1) Can you make processes conditional (say a particle reaches point A then it writes a BGEO). I assume it can do that…
2) Could you create feedback loops in Tops, so if the particle reaches point A could those initial parameters be used as the basis for the next generation?

I think I just need to start poking around in Tops really.
Also, I absolutely love that Karl Sims work. I think I saw it on this mini-documentary: https://www.youtube.com/watch?v=wQQ2NHECcvQ [www.youtube.com]

I'd like to attempt to bump this thread (failing that, I'll repeat Simon's question).

I want to use PDG to version out trees, animals, insects, jellyfish for a film about using AI for evolution.

For each base model, the TOPS parameters that get randomised will be base form (a, b or c), number of legs/branches/eyes, scale, point colour, texture, scales/skin. This could be further complicated by choosing from a variety of base models for each auxiliary feature.

I'd like, as a principle of my project, to design the flora and fauna, but augmented by AI bias or randomised biases. The alternative is to use a gen AI to just concept them as images, which I don't love. In that case I'd prefer to get an output text list of attributes for each animal and just design it; if it came to that, I could do it with just a detail wrangle in a for-loop.

NO, I want a bunch of semi-random designs output as 3D models with randomised attributes, preferably with the point/prim attribs and groups intact so I can edit them and apply materials.

Am I crazy, or is this the exact use case for which PDG could be useful?