Game

Neurotrees Evolution

In this game you change the rules (on the right) to evolve populations of neurotrees able to live in specific conditions, such as light. It shows how genetic algorithms work.

Scientists use genetic algorithms to simulate evolution, and engineers use them to design antennas, enzymes, robots and neural networks. See more in How does this work →

Mission 1 of 3

Best score 0 / 70

1Grow
2Score
3Select
4Copy + mutate

precursor 1 precursor 2 mixed colour = mixed genes

0generation(rounds so far)

0average score(how well they fit their world)

average best

Game The science

How evolution solves problems

The game runs a genetic algorithm: it copies natural selection to find good designs that nobody draws. It needs only three things Darwin described: variation, inheritance and selection [1].

One generation, four steps

1 GrowEight genes, numbers from 0 to 1, grow each neurotree's shape.
2 ScoreA fitness function scores how well the shape fits the world.
3 SelectThe best share survive. The rest disappear.
4 Copy + mutateChildren mix two parents' genes (crossover) and change a few at random (mutation).

↻ Repeat. Nothing in the loop knows what a good shape looks like: the direction comes only from what survives [3, 4, 5].

What mutation does

No mutation: progress stops when the founders' genes are used up. Too much: good shapes get scrambled. Medium works best. The first three generations are flat because, while there is room, everyone survives.

What selection does

Keeping only the best 20% climbs fastest. Keeping 75% is gentle and slow: weak shapes keep having children.

Both charts: average score of the whole population in the Light world, average of 60 runs of this game's algorithm per setting (selection chart at 40% mutation; mutation chart keeping the best 25%).

Where it is used

Space antennasNASA evolved the antenna flown on the Space Technology 5 satellites in 2006 [10].

Enzymes and medicinesDirected evolution runs the same loop with real proteins; Nobel Prize 2018 [11].

Artificial intelligenceEvolving neural networks and their designs [12, 13].

RobotsVirtual creatures that learned to move, and robots built from evolved designs [14, 15].

Laws of physicsRecovering equations, like a pendulum's, from raw data [16].

Evolution itselfTesting ideas about selection and chance that take millennia in nature [2, 17].

Biology ↔ algorithm

Biology	Algorithm	In the game
Genome	List of numbers	8 genes
Environment	Fitness function	Light, Signals, Drought, Treevirus
Survival	Selection	Selection slider
Sexual reproduction	Crossover	Two parents mix genes
Mutation	Random change	Mutation slider

Limits

A teaching toy: real genomes have thousands of genes, real fitness is the number of surviving offspring, and real evolution also involves chance (genetic drift) [17]. Choosing by eye instead of a rule is artificial selection, as in Dawkins's biomorphs [9].

References (18)

Darwin C. (1859). On the Origin of Species by Means of Natural Selection. London: John Murray.
Lenski R.E., Rose M.R., Simpson S.C., Tadler S.C. (1991). Long-term experimental evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations. The American Naturalist 138(6): 1315–1341. doi:10.1086/285289
Holland J.H. (1975). Adaptation in Natural and Artificial Systems. Ann Arbor: University of Michigan Press.
Goldberg D.E. (1989). Genetic Algorithms in Search, Optimization, and Machine Learning. Reading, MA: Addison-Wesley.
Eiben A.E., Smith J.E. (2015). Introduction to Evolutionary Computing, 2nd ed. Berlin: Springer. doi:10.1007/978-3-662-44874-8
Lindenmayer A. (1968). Mathematical models for cellular interactions in development. I. Filaments with one-sided inputs. Journal of Theoretical Biology 18(3): 280–299. doi:10.1016/0022-5193(68)90079-9
Prusinkiewicz P., Lindenmayer A. (1990). The Algorithmic Beauty of Plants. New York: Springer.
Cuntz H., Forstner F., Borst A., Häusser M. (2010). One rule to grow them all: a general theory of neuronal branching and its practical application. PLoS Computational Biology 6(8): e1000877. doi:10.1371/journal.pcbi.1000877
Dawkins R. (1986). The Blind Watchmaker. Harlow: Longman.
Hornby G.S., Globus A., Linden D.S., Lohn J.D. (2006). Automated antenna design with evolutionary algorithms. AIAA Space 2006 Conference, paper AIAA 2006-7242. doi:10.2514/6.2006-7242
Arnold F.H. (2018). Directed evolution: bringing new chemistry to life. Angewandte Chemie International Edition 57(16): 4143–4148. doi:10.1002/anie.201708408
Stanley K.O., Miikkulainen R. (2002). Evolving neural networks through augmenting topologies. Evolutionary Computation 10(2): 99–127. doi:10.1162/106365602320169811
Real E., Aggarwal A., Huang Y., Le Q.V. (2019). Regularized evolution for image classifier architecture search. Proceedings of the AAAI Conference on Artificial Intelligence 33: 4780–4789. doi:10.1609/aaai.v33i01.33014780
Sims K. (1994). Evolving virtual creatures. Proceedings of SIGGRAPH '94: 15–22. doi:10.1145/192161.192167
Lipson H., Pollack J.B. (2000). Automatic design and manufacture of robotic lifeforms. Nature 406: 974–978. doi:10.1038/35023115
Schmidt M., Lipson H. (2009). Distilling free-form natural laws from experimental data. Science 324(5923): 81–85. doi:10.1126/science.1165893
Kimura M. (1968). Evolutionary rate at the molecular level. Nature 217: 624–626. doi:10.1038/217624a0
Van Valen L. (1973). A new evolutionary law. Evolutionary Theory 1: 1–30.

About this page: the game and its texts were created with AI (Claude, by Anthropic). It simplifies a complex topic and may contain errors; check the references for anything important.

Back to the game