From the article, link to the original paper on arXiv:<p><a href="https://arxiv.org/abs/2104.03902" rel="nofollow">https://arxiv.org/abs/2104.03902</a><p>I was wondering whether there was a short statement that summarized how autodidactic systems differ from the Wolfram Physics project approach to the "universe as a computer" and I found one in the paper:<p>>In this section we discuss several protocols for autodidactic systems, as well as some preliminary computational experiments. A protocol defines a dynamic architecture with rules for change that lend themselves to interpretations as learning strategies or reward functions. Generally speaking, the autodidactic paradigm suggests none of these may be fixed a priori, since one history guided by a set of rules might transform itself to be guided by altered rules, though in practice it is often helpful to limit the number of ways in which the system may evolve. Autodidactic systems are distinct, however, from emergent cosmological models that apply a priori growth rules consistently, as in [41].<p>>[41] Stephen Wolfram. A Project to Find the Fundamental Theory of Physics. Wolfram
Media, 2020.<p>Completely uninformed, but phenomena like General Relativity emerging from a priori growth rules (IIRC claimed by the Wolfram Physics project) seems more parsimonious than rules that change over time.