Optimal Spatial Deconvolution and Message Reconstruction from a Large Generative Model of Models

We introduce a general-purpose univariate signal deconvolution method based on the principles of an approach to Artificial General Intelligence. This approach is based on a generative model that combines information theory and algorithmic probability that required a large calculation of an estimation of a `universal distribution' to build a general-purpose model of models independent of probability distributions. This was used to investigate how non-random data may encode information about the physical properties such as dimension and length scales in which a signal or message may have been originally encoded, embedded, or generated. This multidimensional space reconstruction method is based on information theory and algorithmic probability, and it is agnostic, but not independent, with respect to the chosen computable or semi-computable approximation method or encoding-decoding scheme. The results presented in this paper are useful for applications in coding theory, particularly in zero-knowledge one-way communication channels, such as in deciphering messages sent by generating sources of unknown nature for which no prior knowledge is available. We argue that this can have strong potential for cryptography, signal processing, causal deconvolution, life, and techno signature detection.

Interacting Behavior and Emerging Complexity

Can we quantify the change of complexity throughout evolutionary processes? We attempt to address this question through an empirical approach. In very general terms, we simulate two simple organisms on a computer that compete over limited available resources. We implement Global Rules that determine the interaction between two Elementary Cellular Automata on the same grid. Global Rules change the complexity of the state evolution output which suggests that some complexity is intrinsic to the interaction rules themselves. The largest increases in complexity occurred when the interacting elementary rules had very little complexity, suggesting that they are able to accept complexity through interaction only. We also found that some Class 3 or 4 CA rules are more fragile than others to Global Rules, while others are more robust, hence suggesting some intrinsic properties of the rules independent of the Global Rule choice. We provide statistical mappings of Elementary Cellular Automata exposed to Global Rules and different initial conditions onto different complexity classes.