Is the universe a self-learning computer?

By

The simulation hypothesis is a philosophical and scientific theory, which posits that life and the universe as we know it is generated from artificial intelligence. Commonly referenced in jest and humour, the idea of us living in a simulation is one that is often bandied around as a conspiracy theory; a fantastical attempt to justify the laws that govern how the universe develops, an idea that seems so grounded in science fiction that it appears to be just that – fiction.

However, a recent research paper, developed by physicists supported by Microsoft, may be challenging the scepticism of the fundamental aspects of this hypothesis. Alongside other pioneering scientists, cosmologist and theoretical physicist Dr Stephon Alexander recently proposed an idea called the Autodidactic Universe. This suggests that the universe is self-learning and adapting through a self-taught system, similar to machine-learning algorithms.

Alexander recently proposed an idea … that the universe is self-learning and adapting through a self-taught system

The question ‘Why is our universe so special?’ often appears in discourse surrounding the formation and development of our universe. As Alexander discusses in his interview with NewScientist, if the laws and forces that govern our universe were slightly different, everything would be different. Stars wouldn’t be able to burn hydrogen, carbon wouldn’t be produced, and life would simply be unable to exist in the way it does now.

Why is it that these laws of nature have emerged in the exact way we experience them? If a series of potential laws were equally as likely to be implemented, why are these laws the ones that were ‘chosen’?

The Autodidactic Universe theory aims to explain why the universe is the way it is, applying the physical laws of the universe to a matrix mathematical framework.

Our ability to learn is due to our capability to accumulate information that informs problem-solving decisions, building knowledge over a long period of time. If the universe is self-learning, it has the capability to develop itself and perpetuate through a changing series of laws. Analogous to machine-learning systems, the Autodidactic Universe theory describes a feedback system, where the beginning stages of universe development may have influenced the further stages, with the aim of reaching a more stable state.

The difference between a learning and evolving universe is a key distinction. Evolution works through a process of ‘survival of the fittest’. An evolving universe would suggest that there are large numbers of universes, but only the ‘fittest’ survived.

An evolving universe would suggest that there are large numbers of universes, but only the ‘fittest’ survived

A universe that is self-learning, however, would not go through this process. Instead, the Autodidactic Universe has the ability to store, in a matrix-type form, versions of itself, so when mistakes occur the consequences are stored as memory without acting out in real time. Moving forward through each iteration, the
self-learning universe improves and stabilises, unidirectionally replacing older physical constants with improved and more suited laws of physics. This is a concept in itself that would have huge implications for the unification of physics if proven.

If true, this would suggest that the universe may have originated from humble beginnings, with simple interactions between particles that then gradually developed into the complex system we know our universe to currently be.

So, does this research add weighting to the argument that we all are living in a simulated reality?

Well, not quite. Although analogous to recent computing advances in neural networks, the theory only draws comparison between artificial intelligence systems and the development of the laws of physics. It would be a bold step to suggest that we are actually living in a computer, at least without further research being undertaken. Instead, Alexander explains that this theory doesn’t exclusively relate the universe to artificial intelligence. He describes
the potential for some biological control on universe development. Who’s to say a ‘superbrain’ type universe couldn’t also develop such complicated neural networks?

The conclusion to this research is that we still, in fact, do not know how the universe developed. However, the wide scope of theories surrounding this topic, whether we are in fact the creation of a super computer or ‘superbrain’, opens the door to further development of a broader theory. By drawing comparison
between physical theories and machine learning architectures, Alexander and his co-researchers have kickstarted an exciting new concept within theoretical physics. It has by no means provided us with the answers, but instead started a fascinating discussion.

Illustration:

Leave a Reply

Your email address will not be published. Required fields are marked *

 

This site uses Akismet to reduce spam. Learn how your comment data is processed.