Vostinar, Anya E., Emily L. Dolson, Michael J. Wiser, and Charles Ofria. “Identifying Necessary Components for Open-Ended Evolution.”
URL1 URL2
A central goal of the field of artificial life is to build evolving systems that capture interesting dynamics of natural systems, producing evolutionary outcomes such as sophisticated navigation behaviors, novel cooperative strategies, complex ecosystems, or major evolutionary transitions, to name but a few. Such “open-ended” systems are sought after
for a number of reasons: 1) For artificial life researchers, the presence of dynamics that are seen in biology but not artificial life raises the possibility that there is some fundamental and as-of-yet unidentified quality that artificial life systems are missing (Korb and Dorin, 2011). 2) For biologists, access to systems exhibiting complex and nuanced evolutionary processes allows rapid experimentation and facilitates understanding them on a mechanistic level (Tenaillon et al., 2016). 3) For evolutionary computation researchers, insights from open-ended evolutionary systems have the potential to expand the classes of applied engineering problems that we are able to solve with evolutionary algorithms (Hara and Nagao, 1999; Potter and Jong, 2000). Biological evolution is the only process known to have produced general intelligence; replicating this process would provide incredible insights into our own origins, as well as allowing us to harness these dynamics to spur breakthroughs in artificial intelligence. While various artificial life systems have recreated individual dynamics – such as the evolution of complexity, cooperation, and competition (Lenski et al., 2003; Goldsby et al., 2012; Zaman et al., 2011) – these accomplishments have been in highly controlled circumstances. The overarching goal of open-ended evolution research is to create a
system where all of these dynamics can emerge more organically, as in the biosphere.