Artificial Life最新文献

The scientific fields of complexity, Artificial Life (ALife), and artificial intelligence (AI) share commonalities: historic, conceptual, methodological, and philosophical. Although their origins trace back to the 1940s birth of cybernetics, they were able to develop properly only as modern information technology became available. In this perspective, I offer a personal (and thus biased) account of the expectations and limitations of these fields, some of which have their roots in the limits of formal systems. I use interactions, self-organization, emergence, and balance to compare different aspects of complexity, ALife, and AI. Even when the trajectory of the article is influenced by my personal experience, the general questions posed (which outweigh the answers) will, I hope, be useful in aligning efforts in these fields toward overcoming-or accepting-their limits.

This letter presents the idea that neural backpropagation is exploiting dendritic processing to enable individual neurons to perform autoencoding. Using a very simple connection weight search heuristic and artificial neural network model, the effects of interleaving autoencoding for each neuron in a hidden layer of a feedforward network are explored. This is contrasted with the equivalent standard layered approach to autoencoding. It is shown that such individualized processing is not detrimental and can improve network learning.

Epigenetic tracking (ET) is a model of development that is capable of generating diverse, arbitrary, complex three-dimensional cellular structures starting from a single cell. The generated structures have a level of complexity (in terms of the number of cells) comparable to multicellular biological organisms. In this article, we investigate the evolvability of the development of a complex structure inspired by the "French flag" problem: an "Italian Anubis" (a three-dimensional, doglike figure patterned in three colors). Genes during development are triggered in ET at specific developmental stages, and the fitness of individuals during simulated evolution is calculated after a certain stage. When this evaluation stage was allowed to evolve, genes that were triggered at later stages of development tended to be incorporated into the genome later during evolutionary runs. This suggests the emergence of the property of terminal addition in this system. When the principle of terminal addition was explicitly incorporated into ET, and was the sole mechanism for introducing morphological innovation, evolvability improved markedly, leading to the development of structures much more closely approximating the target at a much lower computational cost.

Developing reliable mechanisms for continuous local learning is a central challenge faced by biological and artificial systems. Yet, how the environmental factors and structural constraints on the learning network influence the optimal plasticity mechanisms remains obscure even for simple settings. To elucidate these dependencies, we study meta-learning via evolutionary optimization of simple reward-modulated plasticity rules in embodied agents solving a foraging task. We show that unconstrained meta-learning leads to the emergence of diverse plasticity rules. However, regularization and bottlenecks in the model help reduce this variability, resulting in interpretable rules. Our findings indicate that the meta-learning of plasticity rules is very sensitive to various parameters, with this sensitivity possibly reflected in the learning rules found in biological networks. When included in models, these dependencies can be used to discover potential objective functions and details of biological learning via comparisons with experimental observations.

Insect-inspired navigation strategies have the potential to unlock robotic navigation in power-constrained scenarios, as they can function effectively with limited computational resources. One such strategy, familiarity-based navigation, has successfully navigated a robot along routes of up to 60 m using a single-layer neural network trained with an Infomax learning rule. Given the small size of the network that effectively encodes the route, here we investigate the limits of this method, challenging it to navigate longer routes, investigating the relationship between performance, view acquisition rate and dimension, network size, and robustness to noise. Our goal is both to determine the parameters at which this method operates effectively and to explore the profile with which it fails, both to inform theories of insect navigation and to improve robotic deployments. We show that effective memorization of familiar views is possible for longer routes than previously attempted, but that this length decreases for reduced input view dimensions. We also show that the ideal view acquisition rate must be increased with route length for consistent performance. We further demonstrate that computational and memory savings may be made with equivalent performance by reducing the network size-an important consideration for applicability to small, lower-power robots-and investigate the profile of memory failure, demonstrating increased confusion across the route as it extends in length. In this extension to previous work, we also investigate the form taken by the network weights as training extends and the areas of the image on which visual familiarity-based navigation most relies. Additionally, we investigate the robustness of familiarity-based navigation to view variation caused by noise.

The predominant explanations for including chromosomal recombination during meiosis are that it serves as a mechanism for repair or as a mechanism for increased adaptability. However, neither gives a clear immediate selective advantage to the reproducing organism itself. This letter revisits the idea that sex emerged and is maintained because it enables a simple form of fitness landscape smoothing to explain why recombination evolved. Although recombination was originally included in the idea, as with the other explanations, no immediate benefit was identified. That a benefit exists if the dividing cell(s) form a simple colony of the resulting haploids for some time after reproduction is explored here and shown to further increase the benefits of the landscape smoothing process.

On this 30th anniversary of the founding of the Artificial Life journal, I share some personal reflections on my own history of engagement with the field, my own particular assessment of its current status, and my vision for its future development. At the very least, I hope to stimulate some necessary critical conversations about the field of Artificial Life and where it is going.

Can machines ever be sentient? Could they perceive and feel things, be conscious of their surroundings? What are the prospects of achieving sentience in a machine? What are the dangers associated with such an endeavor, and is it even ethical to embark on such a path to begin with? In the series of articles of this column, I discuss one possible path toward "general intelligence" in machines: to use the process of Darwinian evolution to produce artificial brains that can be grafted onto mobile robotic platforms, with the goal of achieving fully embodied sentient machines.

The evolution of living beings with continuous and consistent progress toward adaptation and ways to model evolution along principles as close as possible to Darwin's are important areas of focus in Artificial Life. Though genetic algorithms and evolutionary strategies are good methods for modeling selection, crossover, and mutation, biological systems are undeniably spatially distributed processes in which living organisms interact with locally available individuals rather than with the entire population at once. This work presents a model for the survival of organisms during a change in the environment to a less favorable one, putting them at risk of extinction, such as many organisms experience today under climate change or local habitat loss or fragmentation. Local spatial structure of resources and environmental quality also impacts the capacity of an evolving population to adapt. The problem is considered on a probabilistic cellular automaton with update rules based on the principles of genetic algorithms. To carry out simulations according to the described model, the Darwinian cellular automata are introduced, and the software has been designed with the code available open source. An experimental evaluation of the behavioral characteristics of the model was carried out, completed by a critical evaluation of the results obtained, parametrically describing conditions and thresholds under which extinction or survival of the population may occur.