A team of theoretical physicists working with Microsoft published an amazing research paper today that describes the universe like a self-learning system of evolutionary laws.
In other words: We live in a computer that learns.
The big idea: Bostrom's simulation argument has been a hot topic in scientific circles lately. The research is called "The Autodidactic Universe, ”And was published today on arXiv. It extends to 80 pages and presents a new theory of everything.
Research argues that the laws that govern the universe are an evolving learning system. That is, the universe is a computer and instead of being in a "solid state", it is perpetuated through a series of laws that change over time.
How does it work; This is the hard part. The researchers explain that the universe is a learning system by invoking other machine learning systems. Just as we can teach machines to run new functions over time, the laws of the universe are essentially algorithms that operate in the form of learning operations.
According to the researchers:
For example, when we see structures that look like deep learning architectures appearing in simple self-taught systems, we can imagine that the architecture of the functional matrix in which our universe evolves laws evolved from a self-taught system that emerged from minimal conditions.
If you think about it, it makes sense that the original law of nature would be incredibly simple, self-perpetuating and capable of learning and evolving.
Maybe the universe did not start with a Big Bang, but with a simple interaction between particles. Researchers cite this humble origin as saying that "information architectures usually reinforce the causal forces of rather small particles."
Scientists describe the ever-evolving laws of the universe as irreversible:
If the evolution of laws is real, is likely to be one-way, because otherwise it would be common for laws to revert to previous states. This is because a new state is not random, but rather must meet certain specifications, while the immediate past state has already met its specifications.
A reversible but evolving system would accidentally explore the immediate past. When we see an evolving system showing periods of stability, it is likely to evolve in one direction.
Explaining these points, researchers invoke the image of IT trying to understand how a given program arrived at an outcome. In one example, the expert could simply check the magnetic marks left on the hard drive. In this way, the results of the program are reversible: there is a history of their execution.
But if the same expert were trying to determine the results of a program by examining the CPU, that is, the entity that is responsible for executing it, it would be much more difficult to do. There is no intentional, internal recording of the functions running a CPU.
Consequences: If the universe operates through a set of laws that, although initially simple, are self-taught and therefore can evolve over time, it is impossible for humans to put these pieces together.
According to this document, rules governing concepts such as relativity may have had functionally different operational consequences 13,8 billion years ago than they would have in 100 trillion years from now. And that means that "physics" is a moving science.
Of course, these are all conjectures based on theoretical physics. Certainly researchers do not literally mean that the universe is a computer.
According to the study:
We examine whether the Universe is a learning computer.
Part of the theory seems to suggest that the universe is a learning computer, as the laws by which it currently operates were not set in the beginning.
We can not reverse the universe as a process, because there is no verifiable record of its processes - unless there is a cosmic hard drive floating somewhere in space.
The research is currently a draft, but researchers are doing a lot of work describing the types of neural network algorithmic systems that could create such a universe.
The team describes this work as the first steps towards a broader theory.
