World as a Neural Network
* Article: The World as a Neural Network. By Vitaly Vanchurin. Entropy, 2020.
Description
Bobby Azarian:
"In 2020, theoretical physicist Vitaly Vanchurin published a landmark paper titled “The World as a Neural Network” in the journal Entropy. Where Hossenfelder described the structural organization of the Universe to be brain-like, Vanchurin argues that the world is literally a neural network, with an interconnected network of “nodes” existing at the microscopic scale that is equivalent to the network of neurons inside our skulls. This network allows the Universe not just to evolve, but to learn, and it is a hypothesis that may actually be testable someday.
“A working hypothesis is that on the most fundamental level the dynamics of the entire Universe is described by a microscopic neural network which undergoes learning evolution.”
(https://bigthink.com/hard-science/the-universe-may-be-a-giant-neural-network-heres-why/)
Discussion
Bobby Azarian:
"But does this perspective provide any additional explanatory or predictive power over previous cosmic models? Well, using this paradigm, Vanchurin was able to show how to reconcile the theory of general relativity and quantum mechanics — a major problem in physics and the goal of a “theory of everything.” One of the greatest mysteries of modern physics is that we have no idea how gravity, which Einstein’s theory of general relativity addresses, interacts with the weird world of quantum mechanics, which includes exotic phenomena such as superposition and entanglement. Vanchurin has shown that using the mathematics of neural networks, you could get the quantum behavior at one limit and classical behavior at another.
Vanchurin’s hypothesis would represent a new kind of “theory of everything,” one that includes emergent phenomena — like conscious observers — that arise due to the computational nature of reality and its tendency to evolve, learn, and grow more complex. This is different from reductionist theories of everything, which only focus on how particles and forces interact, rather than the nature of the Universe as a holistic computational system.
This conception of nature is radically different because it explains the emergence of complexity as a consequence of a Darwinian evolutionary process that is occurring at the fundamental level of reality. Biological evolution would then just be an extension of this process, which continues at larger scales and higher levels of organization, in a hierarchical fashion.
- “… if the entire Universe is a neural network, then something like natural selection might be happening on all scales from cosmological and biological all the way to subatomic scales…some local structures of neural networks are more stable against external perturbations than other local structures. As a result the more stable structures are more likely to survive and the less stable structures are more likely to be exterminated. There is no reason to expect that this process might stop at a fixed time or might be confined to a fixed scale and so the evolution must continue indefinitely and on all scales… atoms and particles might actually be the outcomes of a long evolution starting from some very low complexity structures, and what we now call macroscopic observers and biological cells might be the outcome of an even longer evolution.”
This idea of the Universe evolving through the natural selection of more stable networks over less stable ones makes complete sense if it works like a neural network because it would exploit the mechanisms described by the influential neuroscience theory known as neural Darwinism, proposed by the Nobel Prize-winning biologist Gerald Edelman. While the idea that the world as a whole can evolve and learn sounds radical, it is not just Vanchurin who has proposed this possibility."
(https://bigthink.com/hard-science/the-universe-may-be-a-giant-neural-network-heres-why/)
More information
* Article: The Autodidactic Universe. By Lee Smolin, Jaron Lanier, et al.
URL = https://arxiv.org/abs/2104.03902
Bobby Azarian:
"In 2021, a similar paper was published by some high-profile scientists, physicist Lee Smolin and computer scientist Jaron Lanier among them, called The Autodidactic Universe, which inspired sensational headlines like “Physicists working with Microsoft think the Universe is a self-learning computer.” The paper proposes that the cosmos may possess an innate ability to learn, adapt, and evolve in a manner akin to a living organism. This view diverges from conventional theories of everything, which typically treat the Universe as an inert system governed by unchanging laws. By contrast, Smolin and Lanier posit that the laws of the Universe might emerge sometime after its creation, and those laws might change or evolve as the cosmos develops and learns more about its own structure, dynamics, and possibilities.
Like the Vanchurin paper, the authors argue that the Universe shares many critical features with neural networks, and they also emphasize that the Universe evolves through Darwinian mechanisms, which are evolutionary processes but also learning processes that create information, complexity, and a hierarchical organization. If their big ideas are correct, then this emerging theory of everything has the power to unite physics and biology.
Jaron Lanier, a pioneer of virtual reality, told podcast host Lex Fridman, “There’s been a zillion papers about how you can think of the Universe as a big neural net, or how you can think of different ideas in physics as being quite similar to, or even equivalent to, some of the ideas in machine learning, and that actually works out crazy well. That actually is kind of eerie.”
(https://bigthink.com/hard-science/the-universe-may-be-a-giant-neural-network-heres-why/)