The universe may be learning
The laws of physics are not immutable.
Analogies between the universe and biological life have a certain mysticism to them that can make the more rational types of us balk. Simply because a certain structure, such as the neuronal and dendritic pathways of the brain, resembles the complex structures of the universe at the scale of billions of lightyears, does not mean it is alive, conscious, or thinking.
There may be forces that do try to enforce a certain resemblance between these structures, but this is only a surface pattern.
That has not stopped scientists, philosophers, and computer scientists from imagining, however, that the universe may be only one instance of a long line, perhaps an eternal line, of evolving universes.
Lee Smolin of the Perimeter Institute in Waterloo, Ontario, Canada has been one of the most vocal proponents of an evolving universe. In this theory, the Big Bang was simply the emergence of the present universe from the womb of its “mother” universe. The womb in the mother universe would likely be a massive blackhole, swallowing all the matter needed to fill the present universe from the mother, much in the same way that a baby receives its nutrients from its mother through the umbilical cord. Thus, the event horizon of that black hole would be like an egg shell forming and capturing matter, only for this universe to burst out in a inflation of spacetime.
Andy Weir, author of the bestseller, The Martian, carried this analogy to an extreme in his viral short story The Egg, in which the main character dies and finds out, by interacting with a mysterious, God-like being, that he is in fact destined to reincarnate as every conscious being in the universe, and that this experience of billions or trillions of lives is preparation, perhaps, to create his own universe — his, hers, its, own Egg.
Analogies aside, a real possibility is that the universe may have evolved from predecessor universes. The primary criteria for evolution would be the same as on our own planet: survival to reproduction. Thus, our universe may at some point produce an “egg”, a black hole that swallows up all matter and energy and recycles it into a new Big Bang. Indeed, one might hope this is the case, despite a general lack of evidence.
While this theory may seem fantastic, the connection of the present universe with a black hole in another universe solves a number of present conundrums in astrophysics of the early universe.
The first is called the “horizon problem” which is the apparent tight correlation between regions of the Cosmic Microwave Background (CMB) in different regions of the sky. If all the matter in the universe fell into a black hole, it would have had plenty of time to interact and become correlated “before” the Big Bang. (“Before” here is a bit of a misnomer because, from the perspective of the mother universe, our time is its space dimension into the black hole.)
The second is called the “flatness problem” which is why the universe appears to be spatially flat, with Dark Energy as the current non-explanation. Topologically, a black hole singularity connected to the Big Bang of our universe would solve that because the only way that can happen is in a spatially flat universe. Thus, the flatness of space is inherited from the way the universe was birthed.
One might wonder, if the universe evolved, what did it evolve from? One possibility is that, from universe to universe, the ability to survive and reproduce has evolved. Perhaps the earliest universes were very short lived. Perhaps they inflated themselves out of existence or collapsed in on themselves. Maybe they were empty of matter. That our universe is so finely tuned to allow our existence may not be some accident or the consequence of a multiverse of 500 trillion universes, as some believe, but rather a consequence of uncountable aeons of universes evolving new ones with laws of physics and physical constants capable of supporting a universe that can create complex matter and even complex beings.
Do human beings or other sentient beings have a role to play in the continuation of this universe as The Egg suggests? Are we too a consequence of this evolution of universes? Perhaps. We do not know.
A more recent production out of the Perimeter Institute among others is the equally startling suggestion that, if there is only one universe, that universe is not evolving but learning. (Technically, evolution is a learning process but a collective one that depends on a population of beings and not only one.)
This is the premise of The Autodidactic Universe, an 80 page paper mixing high energy physics, philosophy, and computer science published on arxiv in April 2021, about two months ago as of the time of writing. Smolin is listed as one of seven authors, all with stellar academic credentials.
We can expect this paper to be going through the peer-review process right now given the academic authorship, and it does not read like the scribbling of mystics nor the hasty product of an over enthusiastic Friday lunch conversation. Rather, it proposes in detail with mathematical equations and physical arguments that the universe may learn in the same way that machine learning algorithms do today.
Much of the argument rests of the resemblance of mathematical equations for quantum physics and gravity to matrix algorithms commonly used in machine learning such as Recurrent Boltzmann Machines (RBMs). The suggestion is that, therefore, those equations describe a machine learning process on the scale of the universe where an initial blank slate matrix operation learns, by some constraints, to create the universe we see around us.
That does not imply the universe is conscious, anymore than evolution is conscious. Nor does it imply that the universe is alive, though it may be hard to avoid that conclusion if we also believe the universe reproduces. What it does mean, if true, is that the laws of physics are neither fixed nor arbitrary. Instead, they are consequences of an autodidactic, meaning self-learning or unsupervised, learning process.
The goal of the paper is to ask the philosophical question: “why are the laws of physics what they are?” In a sense, they seem almost arbitrary, and yet they create a universe that is startlingly beautiful, complex, and well-balanced. As with life, we are forced to one of two conclusions: either there is a Demi-urge who chose things to be as they are or the universe learned or evolved that way. That is not to say that if the universe learned or evolved that precludes a Creator. It simply makes the Creator more removed from the active process of creation, a Prime Cause, but not an immediate cause of all that we see when we look at the world.
The paper rests its argument on the idea that all the laws of physics we see or measure are effective laws. They are laws that are reduced from a whole host of possible laws by some unknown process of symmetry breaking and selection.
If this is so, however, why do the laws of physics seem so stable? One answer might be that the constraints that led those laws to become as they are are stable as well. Learning processes in general are irreversible, meaning that once they reach a certain level of expertise, they cease to learn much more until something changes about their environment.
This is a meta theory for why general relativity, Yang-Mills fields, and other fields both have a similarity to one another and are examples, under various constraints, limits, and symmetric breaking, of what are called cubic matrix theories, the simplest nonlinear theories. These matrix models lead to quantum gauge theories that govern forces, matter fields, gravity field, and cosmological embedding. Out of a vast landscape of possible theories, these are what emerged because they produced structures of laws that persisted.
Because it is an unsupervised learning system, it has no goal in mind. Rather, it is simply exploring the space of possibilities and learning and “looking” for regions of stability. In other words, whatever survives, survives. These laws of physics, therefore, may exist because they lead to stability.
An example of this process is a feedback loop between your thermostat at home and your A/C. Temperature rising causes your thermostat to engage an air conditioner. The air conditioner lowers the temperature and the thermostat turns it off. This is a stable, cyclic process. If, however, the thermostat or A/C breaks, the process is broken and becomes unstable. This may invoke another process such as a repair technician being called to restore the balance. Thus, from the mere premise of stability and certain rules of behavior, you get laws that create that stability as well as correction mechanisms to restore it when it fails. That is a kind of unsupervised learning. The system is the way it is, not because it is trying to achieve anything, but because the system persists. When the persistent system fails, it learns a way to restore stability.
This also means that if the universe learns it does not unlearn — that is, the laws evolve in one direction and not the reverse. New laws of the universe are never random but must meet some learning criteria that all past laws met. On the other hand, a system that is largely reversible will just explore all the possibilities. In biology, evolving systems tend to be irreversible for long periods of time, as the environment constrains a species not to revert to previous states. Since biology derives from physics, we would expect that physics would, perhaps, impart irreversible evolution onto biology.
A problem with this argument is that, as far as we know, the laws of physics are reversible in time. Irreversibility is a sort of meta-law, a consequence of statistics that we do not fully understand even now. Yet, it is precisely this statistical law that suggests that, if the universe learns, it learns in an irreversible way. As the universe learns, it must discard information about previous states that it might have used to unlearn. There is simply no room inside the universe to contain that information. Therefore, the process of learning becomes irreversible as it cannot remember its past state.
When it comes to questions like why chirality, CP violation, and baryogenesis, why hierarchies of scale, constants of nature, ratios of parameters, masses, charges, and so on, we can either say it is all random and blame the multiverse, we can suggest a more limited multiverse, or we can suggest a learning universe, or some combination. Without other universes to observe, it is impossible to really say why the laws of physics are what they are. Yet, the autodidactic universe is a compelling philosophical argument, far more satisfying than anthropic or multiverse arguments in my opinion.