Abstract

This paper proposes to embrace alternative concepts for genetic programming that focus on the idea of symbiosis and symbiogenesis as an addition to and refinement of currently used genetic algorithms for music creation. Those typically use Darwinist principles of natural selection implemented in a fitness function as a central principle to generate new individuals. It is suggested to consider symbiotic concepts for the context of music generation with a focus on participatory performance systems. The paper finally suggest four early-stage ideas as starting points for a discussion about potential implementation of these symbiotic realities in evolutionary-inspired music coding.

Introduction

Genetic algorithms and evolutionary computation

Genetic algorithms (GA) have been used for solving optimization problems since the 1960s and are mainly inspired by Darwin’s evolution theory that is based on natural selection processes [1]. Typically, they start with a set of initial solutions, that are recombined via a crossover operator and then mutated. The best solutions are then kept for future iterations via a fitness function, very much like Darwin’s idea of the survival of the fittest. Kramer argues that today’s variants of biology-inspired algorithms evolved along four main historical threads which explains the difference of genetic algorithms [2]. In his view, only one of those variant can still be distinguished from the remaining three, which is genetic programming (GP), which evolves machine learning techniques and differs in terms of their representation [2].

Evolutionary computation in art and music

During the last three decades, genetic algorithms have faced a more frequent use for generating art and music. Starting in the 1980s, evolutionary art has been used for generating pictures [3]. For the field of music, Goldberg is considered one of the first to present a genetic algorithm for composition tasks in the early 1990s [4]. The field emerged throughout the 2000s and the diversity of advances can be observed in Miranda’s anthology Evolutionary Computer Music [5]. More recently, Loughran and O’Neil reviewed applications of evolutionary computation to music composition tasks and categorized 47 genetically inspired approaches from the years 1998 to 2018 [6]. Specifically, they observed the musical problem, the type of evolutionary algorithm, the musical representation, and the fitness function used in each of the projects [6]. In the context of music, fitness functions are often created as interactive interfaces, that let humans evaluate the musical output in real-time.

Symbiotic relationships and symbiogenesis

In her book Symbiotic planet, Lynn Margulis summarized her arguements for a symbiotic view on evolution alternative to the selection-focused, Darwinist theories, and attempted to connect her symbiotic theories to concepts of planetary homeostasis. This so-called Gaia-theory proposes that life on the planet is not only feeding from planetary resources, but a necessity for upholding the earthly environment that in turn creates the basis for life [7]. Margulis argued her lifetime for symbiosis being a core evolutionary driver tightly intertwined with environmental adaption and natural selection processes, and her serial endosymbiosis theory is today widely accepted. As Sapp puts it in the Introduction to his book Evolution by association: a history of symbiosis, our evolution "is not due solely to the accumulation of gene changes within species. In fact, we evolved from, and are comprised of, a merger of two or more different kinds of organisms living together. Symbioses is at the very root of our being" [8].

The role that symbioses played during evolution remains a matter of academic debate, as a discussion between Magulis and evolutionary biologist Dawkins at Oxford showed. Responding to Dawkins critique of symbiotic relationships in the process of selection pressure "(...) why on earth would you want to drag in symbiogenesis when it’s (...) unparsimonious, uneconomical?", Margulis responded "Because it’s there." [9]. An in-depth discussion of alternative evolutionary concepts can also be found in [10]. Ultimateletly, it seems the role that symbiosis has played and continues to play in evolution is a matter of focus on processes that both have been proven to be apparent: natural selection and symbiogenesis both generating evolutionary novelty. Fig. 1 shows a visualization of the concept of symbiogenesis [11].

Three different forms of symbioses are commonly differentiated: parasitism, mutualism, and commensialism. Commensial symbiotic associations refer to animals of different species that share the food caught by the other animal. The term is used in a broader sense where the "benefit to one of the symbionts may be nutritional or protective" [12]. In a mutual symbiosis, both partners benefit from the relationship and there is a reciprocal exchange of nutrients or protection [12]. Parasitism, however, is "a symbiosis in which one of the symbionts benefits at the expense of the other" [12].

Reflecting the theory of symbiogenesis, the question arises whether these realities are acknowledged in current genetic programming for music generation, and what could be suitable concepts to implement those symbiotic principles? Is there a beneficial way to simulate short and longer-term symbiotic relationships that are helpful for musical applications? How can we combine interdependent populations in a way that helps the creative outcome? And what are the problems that an interspecies-level concept can solve that intraspecies concepts cannot? Finally, how can it be helpful to integrate environmentally inspired concepts such as the Gaia theory that seem to be also connected to symbiotic life?

To start addressing those questions, the paper will first review some related work, before four suggestions are made for applying the biological concept of symbiosis for music generation, specifically for live performance context and with audience participation in mind. These suggestions are at an early idea stage, which is why they are presented together with further questions. Finally, these concepts will be further discussed and the paper concludes with some limitations and points to further work planned around the implementation of symbiosis in genetic coding for music.

Related work

Genetic algorithms can be seen in a multitude of variations, and they are closely related to the concept of multi-agent systems for music generation [13]. Tatar and Pasquier reviewed 78 systems, many of which use evolutionary computing as their architecture [14]. However, multi-agent systems typically use individual agents in the sense of simulating (human) musical behaviour and interspecies symbiosis has not been observed in those simulations. The same accounts for the use of swarm intelligence that are built on individuals that co-evolve and affect each other rather than focussing on their interconnection via 2nd order species in symbiosis [15] [16] .

Dahlstedt and Nordahl created a world of co-evolving creatures that recreate based on a genetic algorithm reproducing single children. The creatures have listening capacities and can be seen as a multi-agent system, that creates music as a result of their individuals that work together, trigger each other, and create a small part each [17]. Strictly speaking this concept does not involve symbiotic relationships, even if the environment balances all consumed and produced energy levels, that the creatures take in and digest [17], and that is due to the fact that the single creatures do neither share food or benefit in other symbiotic ways.

With their system Locus Diffuse, Hoy and Van Nort created a multi-user instrument that incorporates the food-searching behavior of slime mold [18]. The work is inspired by both computational simulation of slime mold behavior [19], as well as real-life application of using slime mold as a sound-generating source [20]. The system is an interesting approach, since it is situated "at the crossroads of sonic ecosystem design, agent-based musical systems, multi-user instruments, and networked performance" [18].

Implementing the idea of symbiotic relationships, Hirasawa et al. found that their Genetic Symbiosis Algorithm (GSA), that took into consideration various mutual symbiotic relations between individuals was able to find more flexible solutions compared to conventional methods [21]. They introduced a symbiotic parameter to their genetic algorithm, that was used after the crossover and mutation stages before the fitness function. Their approach however is limited to intraspecies symbiotic mutualism and did not consider other forms of interspecies symbioses such as parasitism or commensialism or interspecies mutualism.

On a rather analytical level, Moriaty uses the concept of symbiosis for theorizing the interaction between musical performers and dancers. She maps the different kinds of symbiotic relationships to the ones that are in place between dancers and musicians when performing together on a stage [22]. The three different symbiotic forms (parasitism, mutualism, and commensialism) are used as taxa describing interactive relationships between dancers and musicians. This biological metaphor offers a valuable new perspective upon the performative relationships, however are they not sonified or used in other generative ways [22].

Lastly, Lorway’s participatory performance system Autopia shall be mentioned, which uses an interactive fitness function that is controlled by audience members [23]. Her melody-focussed live coding approach controls an intitial population of sine waves that are consecutively shaped by a gamified audience voting rationale.

Suggestions for introducing symbiosis in genetic algorithms for music

Moving forward, some early-stage ideas for strategies to implement ideas from symbiogenesis to evolutionary coding are attempted. This is specifically done with a performance-based scenario in mind, in which the audience is recognized as a participating entity.

Suggestion 1: Representing quantities of audience participation as mitochondria

Mitochondria have evolved as the energy-generating symbiotic bacteria in eukaryiotes during the evolutionary process [8]. They are equipped with their own DNA and could be conceptualized as symbionts representing the participatory energy of the users of a multi-user instrument. Instead of representing musical elements such as tones, melodies, rhythm, or harmony, the quantities of participation generated from the server data in networked instruments could be seen as a measure for the overall musical activity represented by a separate, but connected system of mitochondria present in all individuals of a population in the genetic algorithm. This idea could be extended to other holobionts (hosts and their microbiota) [24], in the sense that the microbiota affect the host collective on an inter- and supra-individual level, which affects the the evolution of a population over time.

Suggestion 2: Symbiotic collaboration as part of the mutation operator

Inspired by mathematical models for cellular interaction as described in [25], symbiotic collaboration could also be implemented as part of the mutation function. Returning to the example of mitrochondria, which are passed on maternally with the egg, evolving microbiota in eukariotes could be implemented as a seperate mutation operator that reflects symbiotic relationships and evolves alongside the separate genome of the nuclei. Those symbioses could also vary in their stability over time and represent both short and long-term symbioses, that represent more stable collaborations between species.

Suggestion 3: The eukaryotic cell as a metaphor for musical performance

Rather than simulating symbiogenetic processes, the eukaryotic cell could itself be taken as a metaphor for simulating interactive concert situations. The nucleus of such cells could represent the staged musical performers, that also hold the DNA of such an event. Audience members is such a scenario could be seen as (different) microbiota, living symbiotically within these cells. The cell constituting the closed environment such as a concert hall or other type of venue shielded from the outside world, could represent the cell membrane.

Suggestion 4: Appreciating environmental conditioning (Gaia) within the fitness function

Following Margulis’ attempt to relate symbiogenesis to the planetary environment [7], the evolution of populations could also influence the surrounding resources that in turn enable and influence its genome. In that sense, the symbiotic relationship affects the homeostatic of the environment creating a circular loop between evolutionary processes of the population and their habitats. A further exploration of the Gaia hypothesis as suggested by Lovelock and Margulis [26] is considered beneficial.

Conclusion

Symbioses are interspecies relationships and collaborations that can spread across individuals and form mergers of more or less stable durability. This paper suggested to review those associations as a biological metaphor and approach to extending currently used genetic algorithms and evolutionary programming strategies that are often inspired by Darwinist evolution biology. It was found that symbiogenesis in evolutionary computation is neither fully acknowledged by mutation operators, nor is it a form of collective intelligence typically represented by swarms or as part of multi-agent systems. This is why it is considered beneficial to discuss how we might implement this type of biological reality within future music systems.