Self-Modifying Code in Open-Ended Evolutionary Systems

Abstract

Having a model and being able to implement open-ended evolutionary systems is important for advancing our understanding of open-endedness. Complex systems science and newest generation high-level programming languages provide intriguing possibilities to do so, respectively. Here, some recent advances in modelling and implementing open-ended evolutionary systems are reviewed first. Then, the so-called allagmatic method to describe, model, implement, and interpret complex systems is introduced. After highlighting some current modelling and implementation challenges, model building blocks of open-ended evolutionary systems are identified, a system metamodel of open-ended evolution is formalised in the allagmatic method, and an implementation prototype with a high-level programming language is outlined. The proposed approach shows statistical characteristics of open-ended evolutionary systems and provides a promising starting point to interpret novelty generated at runtime.

Curb Your Self-Modifying Code

“…Self-modifying code is hard to understand and control, and its behaviour hard to predict. It has nevertheless intriguing applications in fields such as software security (Mi et al, 2015;Shan and Emmanuel, 2011), artificial general intelligence (Steunebrink and Schmidhuber, 2012;Schmidhuber, 2003), and open-ended evolution (Christen, 2021;Banzhaf et al, 2016). The present study addresses the issue of controlling self-modifying code by introducing control mechanisms for partial code modifications and specific transitions between code and data.…”

“…Self-modifying code, however, is a useful approach in cases where such programming concepts are difficult to apply or missing. Christen (2021) recently provided a self-modifying code prototype and a formalism how it could be applied within the so-called allagmatic method (Christen, 2020;Christen and Del Fabbro, 2019) -an approach to formalise, model, implement, and interpret complex systems. Thereby, the allagmatic method serves as guidance for self-modification based on concepts defined in a metaphysical framework.…”

“…Within U, whatever is implemented, it takes as an input the entity states of the i-th milieu entities, M(t) i , of the current time iteration, t, and it outputs the entity states of the i-th entity, ê(t+1) i , of the next time iteration, t + 1. This is used to guide the generation of code with the self-modifying code prototype developed earlier (Christen, 2021). The entity state type reveals possible operations.…”

“…It means that code or some parts of the code is not only program code for execution, it is in addition also accessible as data. In the present study, a self-modifying code prototype implemented earlier is used Christen (2021). It is integrated into the allagmatic method and the partial code modification prototype described in the previous section.…”

“…This newly generated string updateRules is then sent to the method UpdateFunction, where it is treated as code. It is compiled and then executed using the previously developed self-modifying code prototype Christen (2021). In short, it converts the string updateRules into a syntax tree object (The .NET API Documentation, 2022b) and this tree is then compiled as an object without writing out a source file.…”