Open-ended evolution
Biological evolution has been running for around four billion years and has not run out of new kinds of thing to make. Nearly every simulation of evolution settles down within days and stops producing anything genuinely new. Working out why is one of the defining open problems of artificial life.
What the term means
Open-ended evolution describes a process that keeps generating novelty indefinitely, rather than converging on a good solution and staying there. The difficulty starts with defining it well enough to test. "Keeps getting better" will not do, because there is no fixed standard to get better at. "Keeps changing" will not do either, since a population drifting randomly changes constantly and produces nothing.
Attempts to sharpen it usually reach for a few properties: new kinds of entity appearing over time, continuing increases in the complexity of interactions, the arrival of genuinely new ways of making a living, and transitions where things that were individuals become parts of a larger individual. None of these has a settled measure everyone agrees on.
Maynard Smith and Szathmary catalogued the moments where biology changed what counts as an individual: independent replicators into chromosomes, separate cells into eukaryotic cells, single cells into multicellular bodies, individuals into colonies. Each is a case of former competitors becoming a unit that reproduces as one. No simulation has convincingly produced such a transition without it being built in.
Why simulations stall
Several causes are usually suspected, and they are not exclusive:
- A fixed objective. If there is a fitness function, the process ends when it is maximised. There is nowhere further to go by definition.
- A static environment. If the world does not change, adaptation is a one-off problem and gets solved. In biology, the largest part of any organism's environment is other organisms, which are also adapting, so the problem never holds still.
- A closed possibility space. If the genome has a fixed length and a fixed meaning for every position, there is a finite set of possible organisms, and the search eventually covers it.
- No room to grow. Novelty often costs energy or space before it pays. A world with no slack kills everything unusual before it can become useful.
Where Emergent Life stands
This project addresses the first two of those causes and not the others, and it is worth being clear about which is which. There is no fitness function, so nothing can be maximised and stopped. The environment is not static: resources deplete and regrow, terrain material is conserved but can be moved by organisms, and disturbed ground recovers over time through ecological succession, which is set out in reshaping the world.
The possibility space, however, is not open. The genome has a fixed number of genes and the network has a fixed shape, so the set of possible organisms is large but bounded. Structure does not evolve. That is a real limitation on any claim of open-endedness, and it is a deliberate one, taken because a fixed shape is what makes two organisms comparable and most measurements here depend on that comparability.
Emergent Life has not demonstrated open-ended evolution, and given a bounded genome it is not positioned to. What has been measured is more modest: lineages diverging under a stated divergence rule, populations adapting to conditions the rules never name, and environmental changes persisting beyond the organisms that made them. The measurements, including the runs that produced nothing, are on research notes.
References and further reading
Primary sources for the claims on this page. Where this project's own results are referred to, they link to the page carrying the evidence instead.
- Maynard Smith, J., & Szathmary, E. (1995). The Major Transitions in Evolution. Oxford University Press.
- Taylor, T., Bedau, M., Channon, A., et al. (2016). Open-Ended Evolution: Perspectives from the OEE Workshop in York. Artificial Life, 22(3), 408-423. doi:10.1162/ARTL_a_00210
- Packard, N., Bedau, M. A., Channon, A., et al. (2019). An Overview of Open-Ended Evolution. Artificial Life, 25(2), 93-103. doi:10.1162/artl_a_00291