Evolutionary algorithms are designed to find impressive solutions in complex search spaces. Meeting this aim requires that the heuristic guiding search aligns with the structure of the search space, i.e. the effectiveness of rewarding properties of individuals (like fitness or novelty) depends on how those properties are distributed. Interestingly, researchers can rarely access ground truth about such connectivity, especially in settings like evolutionary robotics (ER) where search spaces are large and an individual’s behavior could potentially inform search in many different ways. This paper raises the intriguing possibility of adapting or simplifying existing ER domains such that we know everything about the search space’s structure, to enable us to develop intuitions and quickly explore new search algorithms. The proposed approach is to pair an expressive (but limited) encoding with a benchmark ER domain, and precompute the behavior of all possible individuals. Such precomputation enables evaluation as a look-up table, and the further precomputation of normally-intractable quantities, like exact rarity of behaviors and a variety of evolvability metrics. Evolution can then be driven and gauged by such properties with extreme efficiency. The hope is that insights gleaned from this sandbox can inspire new and effective approaches that generalize to when everything is not known.

Of all the fascinating properties of life on Earth, among the most incredible is that it is open-ended. Life has continued its evolution into virtually endless diverse and often increasingly-complex forms for more than a billion years. Photosynthesis, flight, and human intelligence are but a tiny sampling of the boundless feats of evolution, often far exceeding anything yet built through human engineering. In short, evolution on Earth is as close as we have seen to a never-ending algorithm – a prolific generator that continues to invent and diverge over eons without ceasing. The field of artificial life, which dares to explore beyond the confines of conventional optimization, is the natural home for the study of open-endedness. Indeed, open-ended evolution is an active field of research within our community. Yet why has the field not caught fire despite its profound potential to transform our understanding of search algorithms and even ourselves? The argument can even be made that open-endedness may be one of the few, if not the only, viable path to brain-level complexity, and hence a vital link in the pursuit of AI and machine learning. It therefore deserves the status of a grand challenge, and all the mind-share and talent such a status entails. This talk argues for why the field is now poised to be elevated, and how the alife community sits front and center of this new frontier.

Developing a comprehensive theory of open-ended evolution (OEE) depends critically on understanding the mechanisms underlying the major evolutionary transitions; such periods of rapid innovation, such as the Cambrian explosion, have resulted in exactly the kind of diversity and complexity deemed the hallmarks of strong OEE. This paper introduces a new domain for studying major transitions in an evolutionary robotics context. Inspired by the popular Minecraft video game, the new Voxelbuild domain centers on agents that evolve the capacity to build arbitrarily complex block structures with minimal objectives. Initial experiments demonstrate both the rich expressive potential of the new domain and, intriguingly, the occurrence of major evolutionary transitions in at least some runs, thereby providing a unique opportunity to probe how and why such transitions occur or fail to occur in different runs of the same system.

Because the kind of open-ended complexity explosion seen on Earth remains beyond the observed dynamics of current artificial life worlds, it has become critical to isolate and investigate specific factors that may contribute to open-endedness. This paper focuses on one such factor that has previously received little attention in research on open-endedness: the minimal criterion (MC) for reproduction. Originally proposed as an enhancement to novelty search, the MC is in effect a different abstraction of evolution than the more conventional competition-focused fitness-based paradigm, instead focusing on the minimal task that must be completed for an organism to be allowed to produce offspring. The MC is interesting for studying open-endedness because in principle its strictness (i.e. how hard it is to satisfy) can be varied on a continuum to observe its effects. While in many artificial life worlds the MC strictness is implicit and therefore difficult to vary systematically, in the previously-introduced Chromaria world, the MC is designed to be set explicitly by the experimenter, making possible the systematic study of different levels of MC strictness in this paper. The main result, supported by visual, quantitative, and qualitative observations, is that the strictness of the MC can profoundly affect open- ended dynamics, ultimately deciding between complete stagnation (both with extreme strictness or complete relaxation) and orderly divergence. This result offers a lesson of particular importance to worlds whose MCs are not explicit by exposing an area of sensitivity within open-ended systems that is easy to overlook because of its implicit nature.