Can the mind fake the universe?
- Transfer
The objective reality and the laws of physics themselves arise from our observations in accordance with the new concept, which turns upside down what we consider fundamental.
Sophie Hebden
FQXi Awardees: Marcus Muller 1
January 18, 2019
Creation of the cosmos.
Credit: Yuri Akurs, iStock of
Marcus Müller could be mistakenly accused of being the greatest individualist in the world. The radically new view of quantum physics on reality, which he offers, seems to lead to the fact that the world we perceive arises from our observations. He admits that this is a very difficult idea to understand. “No one has yet come up with a science fiction story that has at least somehow illustrated it,” Müller says, “it follows from our theory that only observations are fundamentally real.”
Marcus Muller
Muller, who works at the Institute of Quantum Optics and Quantum Information in Vienna, Austria, and his philosopher colleague Michael Kaffaro of the University of Western Ontario, Canada, cast doubt on one of our basic assumptions: reality is objective and independent of us. Muller and Kuffaro do not claim that the world is illusory or unrealistic, but claim that something more fundamental is behind it and that their approach can explain the origin of reality and how fundamental physical laws arise.
Quantum theory has already made physicists take the role of observer seriously. For example, before we measure a quantum system, it can have conflicting properties, for example, be in two different energy states. When we make observation, we force the system to take a certain state, linking the act of observation with reality itself. Muller and Caffaro claim that they simply follow this path, bringing it to its logical end. Their goal is to develop a basis for describing reality without assuming the existence of ordinary objects with properties determined by physical laws 2 . But how to solve the ambitious task of building the Universe without any building blocks or instructions for its assembly?
“What does it mean from the point of view of mathematics that there are no laws of nature? This is like asking, but what if you only have math? Could this give you something that allows you to predict what you’ll see, that is, some probabilities of observation without assuming anything else? ”Says Müller.“ That was my starting point. ”
A mathematical method that can help do this is already available. The concept of “algorithmic probability” is already used by those who work on artificial intelligence and machine learning. It describes the likelihood of observation occurring over the entire spectrum of possibilities, and programmers use it to create robots that can “learn” from observing the environment and make decisions based on past results 3 .
Dump in the brain
Using this approach, Muller obtained a coherent description of reality from descriptions of sets of observations made one after another. It starts with an observation that is encoded in some way. This method itself does not matter, so let's say that we use binary strings with zeros and ones. Observation can be a dump of data at some point in time describing your state with a long string of bits. Mueller takes a mathematical tool from the field of artificial intelligence and asks: how likely is it that a random computer program will produce these bits by pure chance? This is called the algorithmic probability of this observation.
Mueller postulates that what is happening depends on the algorithmic probability due to what has been observed in the past, and argues that this process can shape reality itself. New observations that do not fit into observations in the past, perhaps worlds with five or six dimensions are discarded, because their probability is much less than the observed world with three spatial dimensions. Observations consistent with what we saw in the past, and which correspond to the description of the world in the usual sense, with ordinary laws, are most likely. Therefore, from our observations we get an “emergent world” 4 .
“It turns out algorithmic probability is a very useful concept,” says Muller. Physical laws and laws arise in the same way as the correct answers that robots give as a result of training. “Laws tend to stabilize, and you observe these patterns around you and the world with such laws,” Müller explains. “This has the same mathematical foundation.”
Renato Renner , a theoretical physicist at the Federal Institute of Technology, Zurich, Switzerland, notes that in physics there is a long history of attempts to formulate first-person theories 5. But unlike most attempts, Müller was able to develop “a well-defined theory within which non-trivial results can be obtained,” says Renner. “It's just great!” “Speaking simply, Muller provided a quantitative version of Occam's razor : it assigns high probabilities to future possible events that have the simplest explanation.”
Boltzmann brain
Along with an explanation of why laws of nature exist at all, Mueller's approach can also solve some notorious puzzles in cosmology and quantum theory. One of the strangest is known as the " Boltzmann brain problem ." This concept dates back to the end of the 19th century, when the physicist Ludwig Boltzmann published a theory stating that the entropy or disorder of a closed system always increases. Think about the hot coffee in the drawer: the clutter in this system will increase over time, as the thermal energy of the coffee will dissipate in the surrounding air. But there is always the possibility that part of the system will fluctuate from disorder to order.
Boltzmann believed that the extraordinary complexity of life on Earth, including our brain, is the result of these random fluctuations, and that they can occur anywhere in the universe. He predicted the emergence of self-aware entities, later called Boltzmann brains, spontaneously appearing and disappearing. What is even stranger, cosmologists in recent years have calculated that many models of the universe suggest that the number of Boltzmann brains should significantly exceed the number of normal human ones. How can you be sure that you are not a Boltzmann brain, and that your next appearance will not happen in the depths of space?
Mueller's theory gives an encouraging answer: to continue to exist as usual on Earth is much simpler than to suddenly appear in space. And if something in the information-theoretical sense is simpler, then the algorithmic probability says that it is more likely.
The ideas have not yet been fully worked out - Muller has not yet published them in a peer-reviewed journal. But the concept aroused great interest, and invitations to the conference surged. Rob Speckens , a theoretical physicist at the Perimeter Institute in Canada, believes that algorithmic information theory will find many applications in physics in the future, and applauds Mueller for being one of the few people to accept the challenge. “This is very original and thought-provoking material,” he said.
But in theory there are roughness saysRudiger Schack , a mathematician from Royal Holloway, University of London, UK, who is working on an alternative interpretation of quantum theory, which also gives the observer a central role called QBism (on the wiki ). Although Schack admires the mathematical rigor of the project, he is concerned about the lack of a clear definition of agents — the people who make observations in Mueller’s theoretical constructs. “The agents of Müller’s theory are oddly diminished entities defined by a random process,” says Schack. “The defining feature of agents is that they make choices and act.” In Muller’s theory, on the contrary, the concept of action, as well as the world in which agents act, are secondary consequences of the theory. “It doesn't suit me,” summarizes Shack.
Müller agrees that the concept of “agent action” is not part of the fundamental description in his theory, but he does not see this as a flaw. “Choice of action is a secondary concept that should not be a fundamental part of any physical theory,” says Muller, “just like emotions or free will. This view is consistent with traditional thinking, for example, in cosmology. "
Renner, who saw the preliminary version of Müller’s work, claims to be one of the most interesting articles he has read in the last couple of years ( arXiv: 1712.01826 6 , short version of arXiv: 1712.01816), making you think. “In view of the conceptual problems that we face in our current physical theories, especially in quantum theory,” Renner says, “such radically different approaches are, in my opinion, extremely necessary.”
Notes by the author of the translation
1. FQXi - Virtual Institute for Basic Research. Supports research in fundamental physics and cosmology through the provision of grants. More than $ 100,000 has been allocated to support this project.
2. The author of the article, following the developers of the theory, uses the term objective reality, but it is more about reality perceived by the sensory organs, described by the laws of physics. Initially, the theory introduces definitions of two types of existence of objects: P-existence and M-existence. On the basis of P-existence, a P-world is defined, which can be interpreted as the existence of objective reality in the metaphysical sense of this concept. M-existence can be generally understood as the existence of a universal computing environment, which, when applied to a human observer, can be interpreted as a computational neural network environment of the brain. Although the authors of the theory do not draw such parallels in an explicit form, and in general try not to abuse broad generalizations in the work, which is typical for works on similar topics.
3. Perhaps the approach described in the article will interest specialists working on AI issues, for this reason, the hubs include Artificial Intelligence and Algorithms hubs. There are many publications on the subject of algorithmic complexity and algorithmic probability on the Habré , of which the publications of the aideus user are dedicated to the use of these concepts in the field of AI, see 1 , 2 , 3 . Perhaps all this is of little relevance in the light of recent advances with the use of ANNs in the development of AI, nonetheless.
4. Emergent in the sense of the emergence of a new integral quality, previously absent from parts.
5. The closest in spirit to this work are the publications of A. Kaminsky on subjective physics . In general, this topic includes the work and discussions related to the role of the observer and, in particular, his consciousness in physics, starting with discussions on the problem of measurements in QM during its creation, and ending with the prophecies of Uncle Roger :) nowadays. According to the author of the translation, the problem of the observer is essentially interdisciplinary in nature, and cannot be fully solved only within the framework of physics.
6. The article turned out to be somewhat watery, as it sometimes happens with popular science articles :), so for those interested I added a translation of the abstract of the work itself:
Spoiler heading
In accordance with the modern concept of physics, it is assumed that any reliable physical theory should describe the objective evolution of a unique external world. However, this assumption is disputed by quantum theory, which suggests that physical systems should not always be understood as possessing objective properties that are simply revealed during measurement. In addition, as stated below, some other conceptual problems in the foundations of physics and related fields indicate possible limitations of the current perspective and motivate the study of alternatives. This article proposes an alternative approach that begins with the (strictly formalized) concept of the “observer state” as its primary concept, and it does not initially assume the existence of a “world” or physical laws. This can be reduced to one postulate, namely, Solomonov’s induction, which correctly predicts future observations. Using tools from the algorithmic theory of information, it is shown that the resulting theory predicts that it seems to observers that there is a world that evolves in accordance with algorithmically simple, computable, probabilistic laws. Unlike the standard representation, objective reality is not assumed with this approach, but rather arises as an asymptotic statistical phenomenon. The resulting theory solves puzzles like the problem of the Boltzmann brain in cosmology, makes specific predictions for thought experiments involving duplication and computer simulation of observers, and predicts new phenomena, such as “probabilistic zombies,” controlled by observer-dependent probabilities. The approach identifies some phenomena of quantum theory (violation of Bell's inequality and lack of signaling) as typical consequences of the information-theoretic features of the agent’s memory and suggests that we switch our attention in the foundations of quantum mechanics to the question “what is really happening?” to questions about algorithms, causation and computational models. This work is not intended to be a “theory of everything”, but rather is a conceptual proof of principle for a new approach to fundamental issues, which puts subjective observation in the spotlight. The approach identifies some phenomena of quantum theory (violation of Bell's inequality and lack of signaling) as typical consequences of the information-theoretic features of the agent’s memory and suggests that we switch our attention in the foundations of quantum mechanics to the question “what is really happening?” to questions about algorithms, causation and computational models. This work is not intended to be a “theory of everything”, but rather is a conceptual proof of principle for a new approach to fundamental issues, which puts subjective observation in the spotlight. The approach identifies some phenomena of quantum theory (violation of Bell's inequality and lack of signaling) as typical consequences of the information-theoretic features of the agent’s memory and suggests that we switch our attention in the foundations of quantum mechanics to the question “what is really happening?” to questions about algorithms, causation and computational models. This work is not intended to be a “theory of everything”, but rather is a conceptual proof of principle for a new approach to fundamental issues, which puts subjective observation in the spotlight.