After one more year of intensive work and numerous test runs, a new major update for is finally polished and ready. It offers possibilities I had only dreamed of before. 🪐
YouTube:
#ArtificialLife
#generativeart
#Cuda
Here is another simulation run (~ 2 days). The blue creatures here are equipped with sensors not only for the fruits, but also for their peers. They now seem to act in a more coordinated way.
Day 5 of my evolution simulation:
I've doubled the size of the simulated world and the number of particles, so it's now running a bit slower.
There seem to have been no more mass extinctions since yesterday. Some of the creatures have developed even more extremities and are more
Day 4 of my latest non-stop evolution simulation:
Crab-like creatures have evolved and are roaming all over the simulation. Otherwise, the funny critters with spiral horns can still be found in a related form. Here is a spatial and temporal close-up. I will probably scale up the
I'm in the process of setting up a new evolution simulation using the insights from the last long run. The initial creatures will be placed in a challenging environment where survival requires behaviors with non-trivial sensory inputs. has already been
Some visual impressions to the next major upgrade of ALIEN (under development): Genetic editor, NN-controlled agents and a simplified messaging and energy transportation system between the cells will be coming ...and many other exciting things too!
#ArtificialLife
#generativeart
.. running the simulation with the latest setting a bit longer and new mutants have emerged. Maybe it's deceiving me, but somehow the entire scenery looks more and more like alien life forms ...
Happy to present some advancements to the new particle engine for the
#CUDA
-based simulation tool "alien"!
It'll be unlocked in the upcoming major release and hopefully provide a good fundament for
#ArtificialLife
and
#Evolution
simulations therein.
There's still a lot to do...
In the following simulation there are regrowing plants that form fruits that can be eaten by other organisms (equipped with sensors and NN). After more than one day of simulation, they have clearly optimized their feeding behavior in this complex physical environment.
A long-running evolution simulation of a larger world in ALIEN. Everything here originates from a single genome.
In the more sparsely populated regions, organisms develop differently than in the denser ones.
Day 8:
The creatures still keep evolving and thus the character of the world. There are notable changes: The spiral thingies are being more and more displaced or at least their once imposing horns are reducing in size. Instead, one can now see many longer zigzag extremities. What
Day 6 and perhaps the last day of my current evolution simulation:
My crawlers have changed a bit again but no completely new mutants have emerged... oh well, although some have developed impressive bubble-like thingies, for example.
I'll stop the simulation for now and improve
Testing a new setting with different rules for the energy supply.
Lots of little creatures have formed a boring mud. But at second glance, one can see that a small ecosystem has evolved around and use them as food.
Here is an interesting close-up:
While testing various simulation parameters, I spotted this beautiful scene (in a panoramic view) of larger agents, flying through clouds of energetic particles.
Day 7:
I decided to let the simulation run a little longer to observe whether or not it remains in a constant state from an evolutionary perspective after a longer period of time.
Fortunately, this is not the case. From a morphological point of view, some individuals have
Welcome to my creepy-creature-non-stop-evolution-simulation diary 😀
Day 3: The diversity of the simulated world has increased and the gray goo has dissolved again. Statistics indicate that many creatures had evolved using a sensory system. But after a while they lost it. I
"Dark Forest" simulation:
A gloomy world inhabited by soft-body replicators, mysterious self-defending creatures and much more.
Made with ALIEN v3.0 to demonstrate new engine capabilities in real-time.
Full HD version:
#CUDA
#ArtifialLife
#GenerativeArt
I keep on experimenting with evolution simulations on my graphics processor... The next update of ALIEN will contain interesting new options and features to guiding evolution. Here is some weird stuff formed after 8 hours on an rtx4090.
A time-lapse through the co-evolution of two species simulated with Artificial Life Environment v3.3.
Full length and quality:
#CUDA
#ALife
#GenerativeArt
I'm working on the possibility of specifying particular predefined geometric shapes (and their sizes) in the genome of the organisms. They can be combined and easily mutated. Here is a first test run (massively slowed down) with combined triangles.
The next update for is ready! 🎉
ALIEN version 4.4 offers many exciting extensions to the used model: genomes can encode repetitive structures, 2 new cell types are introduced (reconnector and detonator cells), new visualizations and a lot more.
The revised genome editor in ALIEN allows to select predefined (as well as individual) geometries for parts of the phenotype. The simulation running in the background reveals mutants that have emerged from organisms which all had carried the same genome.
I've added an SPH fluid solver. I hope it could be useful for evolution simulations where the agents adapt their movement patterns according to the surrounding medium.
now provides a command line interface. This could be useful for automatically running and evaluating simulations for various parameter settings via scripts as well as for performance measurements.
This is an excerpt from a much longer video I put on YouTube.
It shows a simulation of a self-sustaining and expanding network of nodes and filaments emerging from a single spore. It's well worth watching in high resolution.
#CUDA
#ALife
#GenerativeArt
@Perpetual_Horny
The simulation in this video was made with an alpha of the next ALIEN version. See my Discord for active discussions about development as well as Windows builds. The source code is freely available on GitHub .
Evolution in a maze:
The maze connects the inside with the outside world, which are governed by different environmental conditions.
Hi-Res:
#ALife
#evolution
#CUDA
This demo video was created to give you an overview of the state of development of the upcoming major version (still in progress!).
The already implemented features can be tried out here:
The simulation tool has been renewed and should now be easy to use.
@marek_rosa
Here is an example of creatures using their visual input. But the underlying genomes were handcrafted in the editor and not created by evolution.
@ZoldenGames
Yeah they don't seem to move, but they can contain a lot of internal logic (cells for neural nets, energy transfer, sensors, armor...). Need to inspect it in more detail.
@KhayalRaJinn
The simulator and the used genetic system support the behaviors you mentioned (see for an example). The reason why it is not observed here is that the creatures rarely use visual inputs. I have ideas to improve that but I will test it in an other
@marek_rosa
Here is an example of creatures using their visual input. But the underlying genomes were handcrafted in the editor and not created by evolution.
Thrilled to see alien-project featured in
@ThomasCabaret84
's latest video: 🚀Exciting content on the emergence of replicators in cellular automata and particle systems, along with a fascinating autopoietic system by the creator. English subtitles available
Ongoing development is mainly inspired by the feedback received, e.g. settings that include different spatial regions with different physical conditions (thanks
@ThomasCabaret84
).
Thanks also to
@sableRaph
for his lovely and fair critique in some of his past Creative Code News!
@DeryaTR_
Thank you! The pressure is somehow configurable in the simulator. Normally, agents constantly lose some kind of energy over time (by releasing particles into the environment) and must attack/consume others or absorb energy particles to gain new energy for longer survival and to
@LabYosi
A simulation of a tiny ecosystem consisting of different species (herbivores and plant-like material). All structures here are composed of particles whose interaction results in the overall behavior. More details are available on the GitHub page () and docu
@marek_rosa
Thanks! Upscaling an alife-sim has always been a dream for me. Maybe one can start with portals connecting arbitrary many worlds in a certain topology.
I'm still not quite there where I wanted to be before tackling that challenge. ...still in the process of improving the model.
@sina_lana
I'm actually collecting observations on what influence the size of the world has on the evolution of the system (at least with this parameter constellation). In a smaller setting (15% of the size) with otherwise same parameters,the system collapsed in the same amount of time. 2/3
@ZoldenGames
They vary strongly in size. Normally there are between 50,000 and 120,000 creatures.
I also keep track of the generation of each individual. Currently several thousands (also varying greatly since small creatures can reproduce quite fast).
Behind the scenes:
The thermo-mechanics is driven by a particle engine while the
#ALife
stuff is computed on top of it by an information transportation and action layer.
The behavior of the tiny machines results from the interaction of their building blocks.
@sina_lana
In this video about 40h (still running on my PC, currently almost 3 days). The simulation is configured with certain food chain rules: e.g. same mutants can not consume each other, larger can consumer smaller better, ... The rendering shows same mutants with the same color. 1/3
@ZoldenGames
The mechanical energy is not conserved in this sim. The forces are calculated using SPH as well as velocity Verlet for friction and springs between bonds.
Additionally, a cell here (= one particle) can be equipped with a higher-level function (neural net, muscle, attack/defense
@pfrendl
The blueprints of the creatures (including the plants) are encoded by genomes in their cells. In this simulation, mutations were restricted to the parts of the genome that describe the neural networks. The genomes were first created using the built-in genome editor.
@examachine1
@nburn42
In an old paper I used GA for evolution (unfortunately in German; I can send you the full text).
But here I've cultivated self-replicating machines via long-running simulations, as described here
and .
@DMTryptaminesX
I will make it available soon. The simulation shown uses features that are still work in progress and will be provided with the next update.
@sina_lana
So far I've implemented some widgets for diagrams, which gives certain general information about the temporal development for the number of creatures, energies, internal processes and so on (broken down by colors). But it's not really an evolution tree. In this case it look like
@iandanforth
For each particle the engine provides the connected particles and some other properties (pos, vel, energy, state, metadata, ...). Forces are calculated and applied at each time step, but not stored anywhere.
Thank you
@curious_archive
for using and referencing ALIEN in your latest video.
It's an exciting journey through different simulators, concepts and dimensions. 😍
I enjoyed watching it!
@ZoldenGames
The used genome model has a tree-like structure. Each "sub-genome" encodes a network of connected cells in a reference configuration (here the triangles). These can only be constructed by certain cell types. The cells are moving to the reference configuration by acting forces.
@greig_oldford
Thank you! The data structure behind the collision detection is essentially a 2D array of linked lists of pointers, whereby the particle position is mapped to a slot in the array. There is no long range interaction and this works quite fast on a GPU.