Does Information Exist?

Thesis

Information is not a ghostly substance hidden behind matter. It is a family of ways of describing distinctions, correlations, and possible states, made physically consequential by the systems that store, transform, and erase them.

A measure of uncertainty

Supporting/contextual references: [info-shannon-1948]

In 1948, Claude Shannon gave “information” a deliberately narrow job. A communication system has possible messages, a source assigns probabilities to them, and a receiver learns which message was selected. Shannon’s central quantity measures how much uncertainty is reduced by that selection. For a fair choice between two alternatives, the answer is one bit. A choice among four equally likely alternatives carries two bits. The unit is not a tiny bead of knowledge; it is a scale for distinguishing possibilities.

This operational starting point matters because the word information often arrives carrying more than Shannon intended. A bit can be represented by a voltage, a magnetized region, a mark on paper, or a difference in light. What matters mathematically is not the material alone but a reliable distinction between alternatives. Shannon’s theory can then calculate capacity, noise, compression, and error-correction without asking whether a transmitted message is true, beautiful, useful, or understood.

That restraint is a strength, not a defect. It lets information theory describe a telephone line, a hard drive, and a biological signal using common mathematics. It also marks a boundary. A highly surprising string of symbols may contain much Shannon information while saying nothing. Random noise can be information-rich in the technical sense and meaningless to a listener. The theory counts distinguishable possibilities; it does not, by itself, supply significance.

Information is not meaning

Supporting/contextual references: [info-shannon-1948]

Meaning appears when a distinction participates in a practice, a history, or a system of interpretation. The sequence 01000001 can be the letter A under one encoding, the number 65 under another, or simply a pattern of voltages in a circuit. The physical pattern has not changed; the role assigned to it has. Meaning is therefore not an extra substance sprinkled over bits. It is a relation between a representation, a context, and an interpreter or use.

This does not make meaning unreal or merely private. A traffic signal means something because people, institutions, and machines share conventions and respond to it in regular ways. A gene sequence matters in a cell because molecular machinery has been shaped to interact with it. In both cases, meaning depends on embodied arrangements. But Shannon’s measure alone cannot tell us that a red light means stop or that a sequence helps produce a protein.

Several kinds of information can coexist in one event. A footprint carries information about pressure and movement; it may also serve as evidence in an investigation. The first is a physical correlation, the second an inference made by an observer. Treating all of these as identical encourages category mistakes. Treating them as wholly unrelated misses how reliable correlations become useful signs.

No information without a medium

Supporting/contextual references: [info-landauer-1961] [info-bennett-1973]

A message must be instantiated somewhere. This commonplace becomes theoretically important when information is connected to thermodynamics. In 1961, Rolf Landauer argued that the logically irreversible operation of erasing one bit has a minimum thermodynamic cost. For an ideal system in contact with a heat bath at temperature T, resetting an unknown bit requires dissipation of at least kT ln 2 of heat under the usual assumptions. The principle concerns a physical reset, not every act of forgetting and not the energy required for every computation.

Landauer’s argument links logical possibilities to physical states. If a memory can be in either of two macrostates and a reset maps both to one, distinct histories are compressed into the same final state. The environment must carry away a corresponding entropy increase in an idealized reversible limit. Later work on reversible computing clarified that many logical operations can, in principle, avoid this particular cost when they preserve enough information about their inputs. Real devices have additional engineering losses, but the conceptual point remains: information processing is not exempt from physical law.

The result does not turn information into a new kind of matter. It says that a particular operation performed by a particular physical system has thermodynamic consequences. The same abstract bit can be hosted by different media, and the cost depends on temperature, protocol, error tolerance, and what operation is actually performed. “Information is physical” is therefore best read as a warning against disembodied computation, not as proof that information is the universe’s hidden substance.

Quantum distinctions

Supporting/contextual references: [info-shannon-1948] [info-deutsch-1985]

Quantum theory changes what it means to encode and learn a distinction. A classical bit is modeled as one of two distinguishable states. A quantum bit, or qubit, can occupy a superposition of basis states, and several qubits can be entangled so that their measurement statistics cannot be represented as independent local choices. These are not simply smaller, faster classical bits. The possible states and the operations allowed on them have a different mathematical structure.

Yet a qubit should not be pictured as a little container holding every classical answer at once. Measurement yields outcomes with probabilities, and unknown quantum states cannot in general be copied perfectly. Quantum information theory tracks what transformations preserve, what correlations can be shared, and what observations are possible. It has practical consequences for cryptographic protocols, error correction, and computation, but it does not license every claim made in the name of “quantum.”

The quantum setting also sharpens the distinction between information and meaning. Entanglement is a physically tested pattern of correlations. Whether those correlations encode a message depends on how systems are prepared, measured, and interpreted. A formal state vector can predict observations without being a sentence about the world. Here, as in Shannon’s theory, structure and significance are related but not interchangeable.

The black-hole ledger

Supporting/contextual references: [info-bekenstein-1973] [info-hawking-1976] [info-deutsch-1985]

Black holes made the status of information unusually urgent. In the 1970s, Jacob Bekenstein argued that a black hole should have an entropy proportional to the area of its event horizon, rather than to its volume. Stephen Hawking’s calculation that black holes emit thermal radiation made the puzzle sharper. If a black hole forms from a detailed quantum state and eventually disappears into apparently featureless radiation, what happens to the distinctions that described the original state?

The tension is between two successful-looking frameworks. Ordinary quantum evolution preserves information in the sense that a complete initial state determines a complete final state through a reversible unitary transformation. Hawking’s semiclassical calculation suggested radiation described by a thermal spectrum, with no obvious record of the detailed state that collapsed. If both descriptions were carried all the way to the endpoint, the conflict would not be a matter of losing a library’s contents; it would challenge the rules by which quantum probabilities are computed.

Decades of work have produced strong clues and influential proposals, including holographic descriptions, black-hole complementarity, and calculations in quantum gravity that reproduce an entropy curve consistent with information recovery. These developments have changed the landscape, but they have not made every conceptual question simple or settled every dispute about what the recovered information means. The black-hole information problem remains a problem about the compatibility and domain of our theories.

It is important not to overread the slogan that the universe “stores everything.” In a quantum theory, information can be distributed across correlations and may be inaccessible to any practical observer. Conservation in a formal model is not the same as human recoverability, semantic preservation, or an eternal archive. A ledger can balance while its entries cease to be readable by any particular agent.

Substance, structure, or description?

Supporting/contextual references: [info-shannon-1948] [info-landauer-1961] [info-bekenstein-1973]

What, then, would it mean for information to exist? One answer is modest and hard to dispute: particular systems exist in states that differ, correlate, and can support reliable inferences. Information names patterns in those differences under a chosen set of possibilities. On this view, information is real in much the way a map, a boundary, or a temperature is real. It is not an independent ingredient alongside electrons, but neither is it arbitrary once the physical system and question are fixed.

A stronger answer treats information as ontological, perhaps as more fundamental than matter. This position can be suggestive. Physics often describes entities through state spaces, symmetries, and relations rather than through familiar substance. Quantum theory and gravity may eventually be unified by principles about information. But a successful informational formulation would still need to tell us which structures exist, how they interact, and why the predictions match experiments. Calling the fundamental furniture “information” does not answer those questions by itself.

A more deflationary answer says information is only a description imposed by observers. That goes too far in the other direction. The entropy of a gas depends on a coarse-graining, but the molecules, their collisions, and the reproducible limits on compression do not spring into being when we name them. Descriptions are perspective-dependent without being unconstrained. The interesting middle ground is that information is relational: it belongs to a physical arrangement viewed under distinctions that can be made stable and consequential.

This middle ground also leaves room for biology and culture without confusing levels. A cell can use molecular differences as signals; a community can make marks into language; a person can turn a pattern into a memory. At each step, new regularities and meanings become available because a system has capacities and histories. The existence of those meanings does not imply that every physical process is secretly communicating, nor that a formula about entropy contains a theory of value.

Open research directions

Supporting/contextual references: [info-deutsch-1985] [info-bekenstein-1973] [info-hawking-1976]

Information remains an active research boundary because its roles differ across physics, biology, and cognition. Quantum-gravity programs ask whether spacetime can be derived from entanglement or other informational structure, while condensed-matter and nonequilibrium researchers ask how robust macroscopic information emerges from microscopic dynamics. These proposals are mathematically serious, but none has shown that information is a substance or a mind.

Biology and cognitive science pose another set of live questions: how do molecular correlations become signals for a cell, how do neural representations acquire reference, and what makes a record available to more than one agent? Artificial systems add practical tests of whether a model merely compresses patterns or uses them with grounded, flexible significance. The evidence supports investigation, not a final informational ontology.

A distinction worth keeping

Supporting/contextual references: [info-shannon-1948] [info-landauer-1961]

Information is powerful partly because it travels between disciplines. It gives engineers a language for noise, physicists a language for state and entropy, and living systems a language for regulation and inheritance. That portability is also a risk: a precise term can become a metaphor in another context. “The universe computes” may be an image, a speculative ontology, or an empty substitution for explanation.

The most defensible conclusion is neither that information is unreal nor that it is a cosmic mind. Information exists wherever distinctions are physically realized and correlations can be maintained. Meaning enters through organized relations and interpreters.

A bit is not a soul. It does not feel, intend, remember, or care merely because it can be copied, erased, or entangled. But neither is a bit nothing. It is a compact account of a real difference that a physical system can preserve and use. The philosophical work begins when we ask what kinds of difference matter, to whom, and under which conditions.

Sources & references

Supporting/contextual references, not claim-level proof.

1. Claude E. Shannon — A Mathematical Theory of CommunicationBell System Technical Journal 27(3), 379–423; 27(4), 623–656, 1948.
2. Rolf Landauer — Irreversibility and Heat Generation in the Computing ProcessIBM Journal of Research and Development 5(3), 183–191, 1961.
3. Charles H. Bennett — Logical Reversibility of ComputationIBM Journal of Research and Development 17, 525–532, 1973.
4. Jacob D. Bekenstein — Black Holes and EntropyPhysical Review D 7(8), 2333–2346, 1973.
5. Stephen W. Hawking — Breakdown of Predictability in Gravitational CollapsePhysical Review D 14(10), 2460–2473, 1976.
6. David Deutsch — Quantum Theory, the Church–Turing Principle and the Universal Quantum ComputerProceedings of the Royal Society of London A 400(1818), 97–117, 1985.