to ask questions to Chris Fields, and see answered questions from others.
"Physics as Information Processing" ~ Course Syllabus
Type of Session
Type of Session
Lecture by Chris Fields
Session 1 (18 May) will review some history, starting with Boltzmann’s relation between entropy and energy, and continuing through Church’s and Turing’s models of computation, Shannon’s information theory, the “black box” methods of Ashby et al., Pearl’s definition of a Markov blanket, Bekenstein’s area law for black holes, ‘t Hooft’s and Susskind’s formulation of the holographic principle, the ideas of objects, interfaces, and virtual machines in computing, up to Friston’s 2010 and 2013 papers introducing active inference. The goal of this session is to establish the idea of a Markov blanket as a communication interface.
Wikipedia is a good general resource. If you are not familiar with them, review the Wikipedia articles on entropy, the Church-Turing thesis, information theory, and the Markov blanket.
The first statement of the holographic principle is in Gerard ‘t Hooft’s informal paper “
Session 2 (15 June) will ask: why use quantum physics to understand active inference? The answer is the discreteness of information and the necessarily limited resolution of all physical measurements. These naturally lead to a discrete-eigenvalue representation of interaction, i.e. to quantum theory. We will see how quantum theory generalizes the holographic principle, how holographic screens function as Markov blankets, and how all physical interactions between separable (non-entangled) systems can be viewed as communication. This answers the question “to what does the FEP apply?” with “everything measurable.”
Wikipedia has good articles on entanglement and on separable states (i.e. non-entangled states); both include more information than will be needed here.
The first sections (through 3.1) of Fields, Glazebrook, and Marciano, “
Session 3 (13 July) will introduce the ideas of quantum reference frames (QRFs) and their representation using hierarchies of binary classifiers. These formal structures provide a semantics for measurements, and hence provide the basis for a theory of meaning for interacting agents. The language of QRFs allows a particularly straightforward and intuitive definition of variational free energy, and so allows a fully-general, quantum formulation of the FEP. We will see that the FEP is a classical limit of the principle of Unitarity, the fundamental principle of quantum theory.
Sect. 3.2 and 3.3 of “The physical meaning of the holographic principle”,
Session 4 (10 August) will introduce the idea of a topological field theory, a field theory that does not assume a background spacetime. These provide a natural way to represent sequential measurements in terms of Feynman paths – i.e. of thinking about measurements as probing “every possible” way a system could have evolved. These methods allow a completely general description of multi-agent communication that allows the agents to employ both classical and quantum communication channels. We will focus here on how to think about composite agents, e.g. multicellular organisms and their nervous systems.
Sect. 3.4 of “The physical meaning of the holographic principle” or for all the details, “
Session 5 (14 September) will introduce the idea that spacetime is an error-correcting code that organisms (or other observers) with sufficient computational resources use to organize their experiences. This makes spacetime observer-relative, and raises the question of what computational resources an observer requires to be able to “see” spacetime. It also shows that the FEP is intimately linked to the still-open question of how to formulate an acceptable quantum theory of gravity.