Abeba Birhane, David J. T. Sumpter
When mathematical modelling is applied to capture a complex system, multiple models are often created that characterize different aspects of that system. Often, a model at one level will produce a prediction which is contradictory at another level but both models are accepted because they are both useful. Rather than aiming to build a single unified model of a complex system, the modeller acknowledges the infinity of ways of capturing the system of interest, while offering their own specific insight. We refer to this pragmatic applied approach to complex systems — one which acknowledges that they are incompressible, dynamic, nonlinear, historical, contextual, and value-laden — as Open Machine Learning (Open ML). In this paper we define Open ML and contrast it with some of the grand narratives of ML of two forms: 1) Closed ML, ML which emphasizes learning with minimal human input (e.g. Google’s AlphaZero) and 2) Partially Open ML, ML which is used to parameterize existing models. To achieve this, we use theories of critical complexity to both evaluate these grand narratives and contrast them with the Open ML approach. Specifically, we deconstruct grand ML `theories’ by identifying thirteen ‘games’ played in the ML community. These games lend false legitimacy to models, contribute to over-promise and hype about the capabilities of artificial intelligence, reduce wider participation in the subject, lead to models that exacerbate inequality and cause discrimination and ultimately stifle creativity in research. We argue that best practice in ML should be more consistent with critical complexity perspectives than with rationalist, grand narratives.

Read the full article at: arxiv.org

Somayeh Kalantari, Eslam Nazemi & Behrooz Masoumi
Computing (2022)

Nowadays, the study of self-organizing systems has attracted much attention. However, since these systems are run in dynamic, changing, and evolving environments, it is possible that undesirable behaviors that are contrary to the system goals occur. Therefore, it is necessary to provide mechanisms to guide the self-organizing system. However, several approaches were proposed to guide self-organizing systems, more effective approaches are required due to the variation of the contexts in which they are deployed and their complexity. This paper aims to use the self-advising property to provide guidelines about the context of self-organizing systems. The agents of these systems are guided implicitly by using the guidelines provided. In the proposed approach, contextual data is made by an advisor agent that produces them based on the agents’ behavioral entropy. The proposed approach is evaluated using a case study based on the NASA ANTS mission. According to experiments, the proposed approach causes adaptation activities’ costs to decrease at all radio ranges. Besides, in some radio ranges, i.e., 110 and 120 GHz, the guiding state’s adaptive time is less than the no-guiding state’s adaptive time. The evaluations also show that the ruler agents’ mean entropy in the guiding state is less than the no-guiding state in 75 % of radio ranges. This approach’s success in reducing the agents’ entropy indicates its ability to guide self-organizing systems.

Read the full article at: link.springer.com

W. Ross Ashby

The brilliant British psychiatrist, neuroscientist, and mathematician Ross Ashby was one of the pioneers in early and mid-phase cybernetics and thereby one of the leading progenitors of modern complexity theory. Not one to take either commonly used terms or popular notions for granted, Ashby probed deeply into the meaning of supposedly self-organizing systems. At the time of the following article, he had been working on a mathematical formalism of his homeostat, a hypothetical machine established on an axiomatic, set theoretical foundation that was supposed to offer a sufficient description of a living organism’s learning and adaptive intelligence. Ashby’s homeostat had a small number of essential variables serving to maintain its operation over a wide range of environmental conditions so that if the latter changed and thereby shifted the variables beyond the range where the homeostat could safely function, a new ‘higher’ level of the machine was activated in order to randomly reset the lower level’s internal connections or organization (see Dupuy, 2000). Like the role of random mutations during evolution, if the new range set at random proved functional, the homeostat survived, otherwise it expired.

One of Ashby’s goals was to repudiate that interpretation of the notion of self-organization, one commonly held to this day, which would have it that either a machine or a living organism could by itself change its own organization (or, in his phraseology, the functional mappings). For Ashby, self-organization in this sense was a bit of superfluous metaphysics since he believed not only could his formalism by itself completely delineate the homeostat’s lower level organization, the adaptive novelty of his homeostat was purely the result of its upper level randomization that could reorganize the lower level and not some innate propensity for autonomous change. We offer Ashby’s careful reasoning here as an enlightening guide for coming to terms with key ideas in complexity theory whose genuine significance lies less with facile bandying about and more with an intensive and extensive examination of the underlying assumptions.

Read the full article at: journal.emergentpublications.com

Also at http://femto4.chem.elte.hu/entropia/InfoTheoryPapers/Ashby1962.pdf 

Researchers say we’re on the cusp of “GoPro physics,” where a camera can point at an event and an algorithm can identify the underlying physics equation.

Read the full article at: www.quantamagazine.org