(Semi)Formal Artificial Life Through P-systems & Learning Classifier Systems: An Investigation into InfoBiotics

自从图灵在50年代早期提出生命系统模式形成的模型以来，人工生命（ALife）已经取得了巨大的进步。图灵首先展示了一个简单的耦合反应-扩散方程系统如何通过化学不稳定过程产生化学浓度的空间模式。后来，j·冯·纽曼（J. von Newman）证明了可以制造自我复制的抽象机器，而a·林登迈尔（A. Lindenmayer）则引入了l系统来模拟人工植物。在过去的20年里，大部分的人工生命研究都是用一种更特别的自下而上的工程方法来完成的，通过设计或进化控制系统中实体的局部相互作用的规则来产生某些紧急行为。在这种情况下，涌现被解释为系统内的一个过程，这个过程不能仅仅通过检查规则来预测，而只能通过运行模拟来观察。生命中最早的一些里程碑是T. Rays的Tierra， J. Holland的Echo和L. Yaeger的Polyword。这些早期的系统都是基于基于个体的建模框架，这是高度抽象的，并且在代理执行交互的环境的模拟细节（即物理和化学定律）方面非常有限。西姆斯的虚拟生物和研究，如框架或游泳者，将更准确（尽管仍然武断）的物理现实融入了ALife系统。反过来，环境相互作用细节的增加使得更丰富的突发过程得以观察到。最近的工作通过增加发育过程、差异基因表达和基因调控网络，将更详细的生物学纳入其中，赋予生命模拟更大的真实性。因此，随着计算资源越来越容易获得，我们的生物知识越来越深入，越来越多层次的生物、化学和物理细节以自下而上的方式被纳入到ALife模拟中。分析生物技术、计算生物学、生物信息学和微生物学的最新进展正在改变我们对生物系统复杂性的看法，特别是它们执行的计算（即信息如何处理、传输和存储），以便在动态的、有时是敌对的环境中生存、适应和进化。我们建议捕捉这些最新的生物学见解，特别是那些与细胞生物学有关的见解，以开发复杂的细胞样系统的生命模拟。此外，虽然我们建议坚持自底向上构建人工智能系统的传统工程方法，但我们希望将当前的研究实践扩展到一种更加计算形式化和严格的方法来设计和实施人工智能研究。在这个提议中，我们寻求对自下而上的人工生命研究方式进行根本性的反思。到目前为止，这方面的大部分研究都有很强的特别成分，几乎没有形式化。我们提出了一种新的（半）正式的细胞人工生命方法，我们称之为信息生物。InfoBiotics提出，正式信息学方法、进化和学习以及生物学和生化见解之间的协同作用是更有原则的生命研究实践的先决条件。该提案背后的驱动研究问题是：1。将信息学作为人工细胞生命研究的一种有原则的方法，需要怎样的形式信息学、进化和学习范式以及生化见解的结合？2。为了从信息生物的角度提出并能够回答与科学相关且有意义的生命问题，需要在前两者之间取得怎样的平衡？

Artificial Life (ALife) has advanced enormously since A. Turing proposed in the early 50s models of pattern formation in living systems. It was Turing who first demonstrated how a simple system of coupled reaction-diffusion equations could give rise to spatial patterns in chemical concentrations through a process of chemical instability. J. von Newman, later, demonstrated that it was possible to build self-replicating abstract machines while A. Lindenmayer introduced L-systems for modelling artificial plants. The bulk of ALife research in the last 20 years has been done with a more ad-hoc bottom-up engineering approach by designing or evolving the rules that govern the local interactions of the entities in the system as to produce certain emergent behaviour. Emergence in this context is interpreted as a process within the system that could not have been predicted from merely inspecting the rules but that it is observed only by running the simulation. Some of the earliest landmarks in ALife were T. Rays' Tierra, J. Holland's Echo and L. Yaeger's Polyword. These early systems were all based on an individual based modelling framework, which were highly abstract and quite limited in the simulated details (i.e. physical and chemical laws) of the environment where the agents performed their interactions. K. Sims's virtual creatures and research like framsticks or swimmers incorporated a more accurate (albeit still arbitrary) physical reality into the ALife system. In turn, this increase in the detail of the environmental interactions allowed richer emergent processes to be observed. More recent work incorporated a more detailed biology through the addition of developmental processes, differential gene expression and genetic regulatory networks endowing ALife simulations with greater realism. Thus, as computing resources became more accessible and our biological knowledge deepened, more and more levels of biological, chemical and physical details were included in a bottom-up fashion into ALife simulations. Recent advances in analytical biotechnology, computational biology, bioinformatics and micro-biology are transforming our views of the complexity of biological systems, particularly the computations they perform (i.e. how information is processed, transmitted and stored) in order to survive, adapt and evolve in dynamic and sometimes hostile environments. We propose to capture some of these more recent biological insights, in particular those related to cell biology, as to develop sophisticated ALife simulations of cellular-like systems. Furthermore, while we propose to stick to the traditional engineering approach of building ALife systems from the bottom-up we would like to extend current research practice towards a more computationally formal and rigorous approach to the design and implementation of ALife research. In this proposal we seek a fundamental rethink on the way bottom-up Artificial Life research is conducted. Until now, much of this research has had a strong ad-hoc component with very little formalisations. We propose a new (semi) formal cellular Artificial Life methodology, which we call InfoBiotics. InfoBiotics proposes that a synergy between formal informatics methods, evolution and learning and biological and biochemical insights are a pre-requisite for a more principled practice of ALife research. The driving research issues behind this proposal are:i. What combinations of formal informatics, evolutionary and learning paradigms and biochemical insights are needed for a successful development of InfoBiotics as a principled approach to Artificial Cellular Life research? ii. What is the balance of each of the former that is needed in order to ask and, be able to, answer scientifically relevant and meaningful ALife questions from an InfoBiotics perspective?