Epistemic Status: Not a physicist and obviously speculative

I was pondering cosmology the other day and it struck me that the laws of physics seem surprisingly well constructed to prevent any kind of observer-accessible infinity from being observed, accessed, or utilized in any way. This is interesting because naively the laws of physics, and the mathematics we use to represent them, do in fact contain infinities. For instance, we use continuous real numbers to describe quantities and variational calculus to derive laws of motion. Any kind of continuum limit with real numbers necessarily involves infinities such as infinitesimal motions, infinite precisions etc, but quantum mechanics intervenes to prevent these infinities becoming practically accessible to us. Similarly, the mathematical apparatus of quantum field theory itself relies upon an infinite dimensional Hilbert space with continuous (hence real-valued) transformations. However, the Holevo bound and various other no-cloning results prevents us from ever extracting more than a single bit out of the theoretically infinitely precise real-valued state of a qubit. This then applies more broadly. General relativity allows the creation of singularities – points of infinite density – however they must always1 be cloaked in event horizons preventing an observer from non-irreversibly interacting with them. And at the largest scale, the cosmological constant produces an accelerating universe that strictly limits the causal reach of any observer’s worldline to a finite volume.

Moreover, what I find striking here is how these apparently disparate laws often appear to conspire to close loopholes left by other laws of physics, almost as though preventing operationally accessible infinities is a fundamental consistency condition on the laws of physics themselves.

More specifically:

Quantum mechanics enforces finite distinguishability despite (theoretically) continuous states. Despite a theoretically continuous and real-valued state, various quantum effects preclude infinite information being able to be accessed or extracted from any finite volume of space. Specifically:

  • A qubit possesses a continuously valued state, so one could conceivably try to encode and then extract an infinite binary string from its state. However, the Holevo bound states that a qubit can only carry at most a single bit of classical information. Various no-cloning theorems then enforce that we cannot make unbounded copies of the same qubit to extract more than this single bit.

  • Relatedly, the Heisenberg uncertainty principle bounds the amount of precision possible in our measurements. As we attempt more accurate measurements of some quantity we necessarily lose information about other measurable states. However, even for a single state there is a generalized uncertainty principle that defines a true objective minimum length scale. The idea here is that to resolve a state with increasing levels of precision, we need to probe with increasingly small wavelengths. Smaller wavelengths possess increasing energy and momentum. However, eventually general relativity intervenes, since we cannot concentrate increasing amounts of energy into a volume without changing the geometry of that volume and ultimately causing the collapse into a black hole. This process sets the planck length, which is the minimal length-scale such that the energy required to resolve that length-scale collapses into a black hole2.

  • Similarly, energy density also bounds the processing speed of a quantum state. Specifically, the Margolus-Levitin theorem relates the minimum time taken for a quantum state to evolve into an orthogonal perfectly distinguishable pure state to its energy. This means our theoretical quantum computer can, at maximum, perform operations at a speed that increases linearly with its energy. But again, general relativity prevents us from operating with infinite speed and hence performing an infinite amount of operations in a finite time.

Special Relativity enforces an observer-independent finite speed limit of causal interactions. This one is pretty obvious. The speed of light is not just the speed of light. It is a fully general finite rate of causal information transmission, which bounds the radius by which some observer can interact with other parts of the universe to a finite rate. If there was no such speed limit, an observer could travel ‘infinitely far’ and thus interact with arbitrarily distant parts of the universe within finite time. The mathematical structure of special relativity itself can essentially be derived from the two principles of finite speed limit and observer independence/no privileged reference frame, which is what derives its distinctive Lorentzian dynamics.

General relativity imposes finite resource bounds within a finite region. the core insight of general relativity is that energy and momentum change geometry in an observer-independent way, and in a way that is inherently self-limiting. The more energy or information we concentrate in a region, the more curved that region becomes such that ultimately, the region becomes entirely closed as a black-hole. This ultimately prevents an observer from accessing the information within that region, at least if they ever want to interact with anything outside of that region ever again.

More fascinatingly, is the fact that it turns out that we can derive general relativity from precisely such an information-theoretic argument. Specifically, Jacobson shows that if we assume that the amount of information within a finite volume is finite and scales with the area, obey the usual laws of thermodynamics such as the Clausius relation, and that the laws of physics must remain invariant across observers without a privileged reference frame, then that is sufficient to derive GR through a surprisingly simple entropic argument. This strongly suggests to me that GR really is, in some sense, an informational force concerned with relating information and geometry and ultimately to enforcing finiteness within a volume.

From here, we get the famous Hawking-Bekenstein black hole entropy bound which makes precise the point that the maximum entropy possible in a volume is both finite and scales with the area, rather than the volume itself, which is fundamental to the broader holographic principle3.

A positive cosmological constant implies a finite total accumulated information, even across infinite time. This then brings us to cosmology. General relativity enforces finiteness in space, but it is the cosmological constant that enforces finiteness in time. Imagine we lived in a non-expanding universe with a cosmological constant of zero. Assuming that the spatial extent of the universe is actually infinite, if you set off in some direction you could eventually travel through and interact with an infinite amount of energy or information. Given infinite time, your causal horizon becomes infinite as well. Conversely, if you sent out probes that beamed back energy, in this flat universe, they could eventually reach ever more distant galaxies and beam their energy back resulting in an unbounded amount of energy passing through at the origin over infinite time. You could thus construct an infinite computer which takes the energy of this beam, uses it to perform some irreversible computation, and then emits waste heat forever. Moreover, if you properly balance incoming and outgoing energy rates, this computer could compute forever without accumulating energy and hence collapsing into a black hole.

A positive cosmological constant stops this possibility because it causes space to expand between you and other galaxies, leading to a finite causal horizon where there are sufficiently distant galaxies you can never reach, even travelling at the speed of light with infinite time. This means that there is only a finite amount of mass/energy ever accessible to an observer. While general relativity prevents the instantaneous information or negentropy of a volume being infinite, the cosmological constant prevents the integral of this entropy over infinite time from becoming infinite.

Thermodynamics then enforces finite irreversible computation: Supposing that despite the cosmological limit, we nevertheless concentrate an enormous amount of energy in a volume which we plan to use for computation. Thermodynamics combined with the positive cosmological constant prevents us from turning this into an infinite amount of computation. Specifically, Landauer’s principle sets the minimum thermodynamic cost for the irreversible transformation of a bit which depends on the temperature. If the temperature tends asymptotically to zero, then over infinite time, it becomes costless to perform an irreversible computation, and hence an infinite amount of computation can be performed. This is Dyson’s proposed infinite computation scenario – where a civilization can obtain infinite subjective computation in a universe with asymptotically vanishing temperature by simply spreading out individual computations over increasingly long intervals.

However, the positive cosmological constant again prevents this infinity. With a positive cosmological constant, the universe itself becomes a de Sitter space with a finite cosmological event horizon. This event horizon itself has a fundamental non-zero positive temperature, even after all initial matter and energy has decayed and been redshifted away by the cosmological expansion. This means that there is a fundamental and inevitable floor on temperature and thus on the cost of a bit erasure, and hence the total amount of computation ever possible to perform4.

To sum up quickly. Quantum mechanics prevents us from extracting or processing infinite distinguishable information from theoretically continuous states, at least without infinite energy. General relativity enforces that only finite energy or information can exist within a finite volume. Special relativity enforces that an observer can only causally interact with a finite volume within a finite time. The cosmological constant enforces that, even across infinite time, an observer can only ever interact with a finite amount of energy and information. Landauer’s principle, combined with positive asymptotic temperature from the cosmological constant, enforces that only a finite amount of irreversible computation can be performed with finite energy, even given infinite time.

Together, these laws seem to lock together into a perfectly self-consistent structure which rules out infinities of any kind becoming operationally accessible to an observer, even over infinite proper time. Importantly, this isn’t the effect of a single global law which prohibits infinities. Rather, different laws of physics ‘conspire together’ to prevent any infinity being exploited, even if it is technically allowable within a single theory alone. The classic example here is that quantum mechanics allows infinitely precise measurements with infinite energy, but general relativity enforces finite energy density.

The information-theoretic derivations of general relativity and the foundations of quantum mechanics are especially suggestive here, in my opinion. Einstein’s field equations can be derived from the simple principle that the total accessible information in a volume scales with the area, coupled with observer invariance. Similarly, much of the mathematical structure of quantum mechanics can be derived from axioms of probability theory plus continuous reversible transformations in the state.

Very speculatively, this suggests to me that we might be looking at this backwards. The way these disparate laws interlock to prevent accessible infinities may not be an accident, but rather it might be one of the fundamental consistency conditions which our surface laws of physics arise from in the first place. The general relativity and quantum mechanics derivations show that what appear to be substantive and somewhat idiosyncratic laws of physics can be derived independently from a small set of propositions and constraints concerning the extraction, transmission, and ‘processing’ of information by an observer, combined with general geometrical principles of symmetry and invariance. The fact that we can derive the foundations of two pillars of modern physics from such simple axioms is extremely suggestive that perhaps we can go deeper with this and ponder whether even more of physics can be rederived from such deeply fundamental principles. I have no idea if we can derive e.g. the standard model from such axioms and I doubt that it is so simple. Nevertheless, it seems clear to me that ‘the universe’ in some sense cares very deeply about preventing any observer from accessing any infinity and this points towards a deeper principle, which while maybe not sufficient in itself5, serves as a fundamental constraint on the kinds of physics that are possible.

  1. Technically, this is just a hypothesis and some solutions of GR appear to allow such ‘naked singularities’. My strong prediction is that these accessible singularities are not actually realized or realizable in practice. If this does not hold then this general view is false but also this causes many downstream issues for GR and our standard notions of causality. ↩

  2. There are subtleties here in that many QFT theories assume infinitely many local modes and harmonics. To truly rule this out and understand the interaction between these quantum effects and the gravitational effects, we need a full theory of quantum gravity, which we do not possess. However, it would be very shocking if any theory of quantum gravity enables infinitely precise quantum measurements. ↩

  3. Why it scales with the area is still deeply mysterious to me, although this is likely a skill issue on my part. ↩

  4. Note, that this only applies to irreversible computations. As far as I know, it is in principle possible to perform infinitely many reversible computations. Of course we can model an irreversible computation as a reversible computation with a larger state, such that all the ‘irreversibly lost’ bits become part of a larger reversible state. However, there is a limit to this strategy again due to GR – we have to store these extra state bits somewhere and if we store too many we collapse into a black hole. This eliminates a potential loophole to infinite irreversible computation. However, if we are okay with reversible computation then we can reach an infinity (in theory). This suggests that reversible computation somehow ‘does not count’ from the perspective of the universe or thermodynamics, which makes sense intuitively to me, since if the computation is fully reversible, then it cannot really ‘make any progress’, but I don’t know how to properly formalize this intuition. This is an interesting direction though, because it exposes a slight flaw in our conception of information as entropy. For instance, a reversible computer could cycle randomly through an infinitely long series of state transitions – however there is an important sense in which the ‘useful information’ thus generated is finite. ↩

  5. This is clear because it is trivial to come up with alternative ‘laws of physics’ which enforce finiteness for any observer while also differing strongly from our own laws of physics. This implies that while accessible finiteness for all observers may be a constraint on our laws of physics, it is not sufficient to uniquely define our laws of physics. This means other similarly fundamental principles are necessary to uniquely specify our universe. ↩