Artificial Intelligence

1-hour Evolution of an AI ecosystem

This is a side-project that I’ve been working on for nearly 4 years now (on and off). The simulation runs on Unity.

The creatures (called bibites) initially spawn with an empty brain and a basic genetic code, but have a random chance of getting mutations, that will change their traits and/or behaviors, that they can then pass to their offspring.

Those mutations can range from a slight change in size to new connections being added to their brain (thus changing their behavior).

The more they eat, the more eggs they can lay, propagating their genes.

Their brains are inspired by the rtNEAT method, so, except from their senses (input neurons) and their possible actions (output neurons), every component of their brain are evolved (like developing a connection between seeing food and going forward) and selected through natural selection.

I have bigger plans for this project now that that I’m done with my studies and I have a little more free time. I will upload a video to explain it further if there is enough positive feedback and/or interest.

Music is :
Inscencer Long (1 hour) by P C III
Downloaded from

Please follow and like us:

Comments (16)

  1. Next change a random character in a random line of code, and rerun the software. Repeat.
    How many steps do you get before the software crashes? Do you ever get a beneficial change or new feature?

  2. i watched the whole thing in half an hour

  3. Would have like to seen a graph for traits and outputs etc

  4. S'ils étaient capable de plus de diversité génétique et de réactivité vis à vis de leurs congénères.
    Par exemple face à plusieurs types de nourriture comme :
    Des morceaux immobiles
    Des morceaux mobiles
    Des cadavres de leur congénères morts
    Des morceaux issus de cadavres.
    Voir des morceaux issus d'individus afin d'attirer des proies !

    Également diversité génétique comme :
    Une attirance pour la couleur (après ajout d'un gêne de la vue ?)
    Une manière d'attirer la nouriture (ils arrivent bien à attraper par moments)
    D'être cannibale, charognards ou des chasseurs. Voir de se nourrir uniquement par petits morceaux sur les autres sans les tuer.
    Également d'en voir se nourrir d'oeufs …

    Il y a beaucoup de facteurs à prendre en compte je pense et qui pourraient donner sur le long terme des espèces distinctes occupant des niches précises.
    Et cette vidéo bien que très intéressante, montre aussi qu'on peut aller plus loin dans la réflexion.

    A noter que j'ai l'impression que très vite on voit le tout se stabiliser sur deux "espèces" et dès l'apparition d'une troisième l'une des deux précédente à tendance à s'éteindre.

  5. What's the name of this song?

  6. This is really really cool

  7. You should make this a website

  8. 1 hour is a good time slice but speed it up before posting so we can see it in x4 or whatever looks good

  9. Next step: total world domination ! Don't give them access to wifi..

  10. looks pretty shit, but you got the general idea

  11. This is awesome! I would definitely like to find out more about how things work, particularly how the brains evolve and how input/output is encoded/decoded to/from the neural code

  12. Reminds me of googles polyworld experiments. You should definitely check out their video and paper on it!

  13. ok but how do they see? Is size the only physical mutation possible and what does being small or large do? Are there carnivores in their population? If so, do the carnivores only eat ones that were born already, or can they eat eggs too?

Comment here