What this covers
This February 2022 episode collects questions submitted by Patreon supporters, culled and occasionally grouped by Sean Carroll. The format is conversational and ranges widely, touching on quantum mechanics, spacetime, thermodynamics, philosophy of science, and democracy. Carroll works through conceptual puzzles and misconceptions across these areas, often circling back to how emergence, coarse-graining, and the limits of human observation shape what we can know and how we reason about it.
The substance splits into three distinct territories. First, quantum and relativistic physics: Carroll clarifies what wave functions really are (properties of fields, not spreading particles), explains why entanglement is generic rather than exotic, corrects the myth that virtual particles literally pop in and out of existence, and walks through how spin emerges from symmetry groups independent of relativity. Second, thermodynamics and the nature of time: he argues the arrow of time is macroscopic and entropy-driven, rooted in the universe's initial conditions, and defends the eternalist view that past and future moments are equally real. Third, epistemology and decision-making: Carroll argues that falsifiability is overblown as science's defining feature, that Bayesian Reasoning handles uncertainty without infinite regress, and that naturalism emerges from evidence rather than assumption. He is willing to entertain that miracles could become credible if evidence shifted appropriately. Throughout, he emphasizes that coarse-grained descriptions capture objective facts about nature—not observer conveniences—because they preserve just the information needed for prediction. The episode also includes practical reflections on PhD thesis-writing in physics, the Swiss watch industry's survival strategy, and the preconditions for functioning democracy.
Carroll argues that emergence and higher levels of description are objectively real features of nature (not observer-dependent conveniences), that fundamental physics rarely needs revision to explain higher-level phenomena like consciousness or life, and that Bayesian reasoning—not absolutist principles—is the right tool for thinking about everything from miracles to catastrophic risk to democracy.
- Coarse-grained descriptions like fluid mechanics carry just enough information to make accurate predictions—an objective, non-generic fact about nature
- Quantum field theory and the wave function are taken as physically real, with classical/particle pictures emerging from them
- Good Bayesians assign priors, update on evidence, and refuse to treat metaphysical or political possibilities as strictly impossible or certain
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Intelligently designed emergent systems differ systematically from un-designed (evolved) ones: a designed machine must be held in an engineer's mind and built for a future purpose, making it less complex but far better at its specific goal; an evolved organism is shaped only by what works in the present moment with no future planning, making it more complex, more robust, more self-repairing, and more multi-purpose, but rarely optimal at any single task.
“the cow can be infinitely more complex than the car can, really when you get down to the detail level.”
Carroll's sharpest 'aha' moment came on a plane ride home from a seminar: learning from Don Page that the Lorentz group in 2+1 dimensions has the geometry of anti-de Sitter space let him prove via simple geometric arguments that you cannot build a time machine (closed timelike curves) without starting with one—illustrating how a single provocative fact can unlock a problem he and collaborators had struggled with.
“what I realised on the plane ride home by sketching things is that this answers all of our questions.”
Emergence is an objective property of nature rather than an observer-dependent phenomenon, because (1) the physical information accessible to any observer—e.g. the few photons carrying information from cream and coffee—is far too limited to access microscopic details, pointing everyone toward the same coarse-graining, and (2) the coarse-grained information you do retain is precisely what is needed to predict what happens next, which would not be true of a random subset of microscopic information.
“that is why Dan Dennett calls emergence real patterns. It's the word 'real' there that really matters. It's not an observer-dependent phenomenon for a couple of reasons.”
The arrow of time is a macroscopic phenomenon arising from increasing entropy in the universe, not from time-reversal symmetry; just as space has no arrow yet exists, time can exist without an arrow, and at the universe's eventual equilibrium the arrow of time would disappear even though time itself persists.
“The arrow of time is a macroscopic phenomenon that comes about from the fact that entropy is increasing in the universe. It has nothing to do with time reversal symmetry.”
Bayesian updating does not require 100% confidence in evidence nor an infinite regress of uncertainties about uncertainties; you simply recognize that your perception may not map directly onto reality (illusions, hallucinations, mistakes) and attach error bars to your evidence and likelihood functions, after which you can stop—no need to add uncertainties to the uncertainties of the uncertainties.
“your interpretation of the evidence you get as something is happening in the world should always come with some error bars, with some uncertainties attached to it. But once you attach those uncertainties for some good reason, I don't think you need to attach uncertainties, the uncertainties and uncertainties to the uncertainties”
Entanglement is the generic case in quantum mechanics, not a special event: because there is one wave function for an entire collection of particles rather than separate wave functions per particle, unentangled states (where the joint wave function factorizes) are the special exception; entanglement therefore occurs all the time, as soon as two particles interact or one particle decays into two.
“So entanglement is by all means the generic case in the space of all possible quantum goings on.”
Spin does not arise from unifying special relativity and quantum mechanics—it appears in non-relativistic quantum mechanics too (a single spin-half particle / qubit); spin emerges systematically from the representations of symmetry groups, and even the non-relativistic Galilean symmetry group including rotations would yield spin-half and spin-one representations, so there is no necessary connection to relativity.
“it doesn't arise from unifying special relativity in quantum mechanics. It's perfectly okay to have spin in non-relativistic quantum mechanics.”
Wave-like field oscillations naturally want to spread out, so a free electron's wave function would disperse all over space; particles appear localized largely because most are bound in atoms, where the proton's attraction confines the electron's wave function into localized orbitals, and because heavier objects (nucleus, Earth) spread far more slowly, making macroscopic localization effectively complete.
“it is the attraction of the proton that confines the wave function of the electron.”
The particle/anti-particle dichotomy is a sloppy rule of thumb: some particles (photon, Higgs) have no anti-particle, while particles carrying a conserved charge like electric charge or baryon number have an anti-particle with opposite charge; what actually matters is not particles vs anti-particles but which interactions can happen, so neutral dark matter particles may have no anti-particle yet still annihilate into photons.
“they claim there's this thing called particles and this thing called anti-particles, which is a bit of a cheat. The reality is, some kinds of particles have anti-particles and some don't.”
Entropy is a macroscopic concept that gains relevance only from coarse-graining and incomplete information; there is no reason to think about entropy at the microlevel, and a sand castle is not more 'ordered' than a sand pile at the level of individual grains—if you knew all the microdata, you would never need to talk about entropy at all.
“There is no reason to think about the concept of entropy on the microlevel, entropy is a macroscopic concept. It's a concept that gains relevance from course graining, from thinking about systems from the point of view of incomplete information”
Virtual particles are not really popping in and out of existence; what is really going on is quantum fields, and virtual particles are interior lines of Feynman diagrams—a calculational and intuitive device for the effects of vibrating fields. The popping-in-and-out language is especially misleading for the vacuum, whose quantum state is stationary and unchanging in time, so nothing is actually appearing and disappearing there.
“Not really, is the short answer because really what's going on are quantum fields, that's always the answer.”
Saying 'you should maximize utility' is almost content-free because any decision-making algorithm can be mathematically cast as maximizing some number; the substantive and questionable step in practical utilitarianism is assuming there is a single quantity comparable and addable linearly across different people, which is what Carroll is most skeptical of.
“Any decision-making algorithm can always mathematically be cast as maximizing some number, and then you can call that utility”
Science does not assume naturalism a priori (methodological naturalism), but rather concludes naturalism because it is the best fit to the data; science is methodologically empiricist, so a good Bayesian should not rule out supernatural possibilities in advance—if certain phenomena were better explained by miracles than by physical law, that should shift credence toward miracles.
“I never bought into this very common claim of methodological naturalism as part of science. That science has to assume naturalism from the start.”
Falsifiability is overblown as a demarcation of science: Popper was really gesturing at two features—a good theory should be definite (allowing some outcomes and forbidding others) and empirically judgeable—but treating literal falsifiability as the criterion fails, as shown by theories with free parameters (cosmic strings, dark matter, String Theory) that could hide from observation yet still be confirmed if detected, which would create an absurd philosophical knot if you had pre-declared them unscientific.
“the whole idea of falsifiability is a little bit overblown.”
Making genuinely new research progress is dominated by lonely sitting and thinking, with talking to colleagues and reading papers being important but secondary; projects range from programmatic ones where the path to a paper is clear once you have the idea, to open-ended ones (like the structure of Hilbert space in quantum gravity) where you can spend a year and publish nothing.
“if you're making progress on something that is new and interesting at a research level, you're doing a lot of lonely sitting and thinking”
You can get a rough idea of physics from verbal descriptions, but not an accurate one—not because the words omit numerical coefficients, but because without understanding the mathematical concepts you don't really know what phrases like 'the curvature of spacetime' mean (Einstein's equation uses a specific representation, the Einstein tensor, not all curvature), so verbal expressions are incomplete unless you grasp the concepts behind the equations.
“if all you have are those words and you have no idea what the mathematical concepts are, then you don't really know what is meant by the phrase 'the curvature of spacetime'.”
The Swiss mechanical watch industry cratered in the quartz crisis of the late 1960s-70s because quartz watches were cheaper and more accurate, but the industry survived by repositioning mechanical watches as elegant technological marvels and works of art—jewelry that men can wear—creating a modern hobby with price points from $100 to over a million dollars.
“the Swiss mechanical watch industry cratered. The quartz crisis, it's called, because why would you spend more money on a watch that is not as accurate?”
In theoretical high-energy physics, the PhD thesis as a text goes almost unread—you can staple together your papers with an introduction—so spending time rewriting chapters into a coherent narrative is a poor use of time compared to getting unpublished research submitted to journals; however, this varies wildly by subfield, with some programs requiring a single-author paper as the thesis.
“as a text, as a collection of words and thoughts, yeah, almost no one's gonna read it.”
In the eternalist (block universe) view, all moments in time are equally real, but they are not real 'at once' or at a single time—just as all points in space exist without being at the same place; you are located at one moment and count your position relative to others, so there is no universal present even though past and future moments are real.
“all moments are equally real. The current moment you're in is real, but future moments are also real, past moments are also real. They're not all real at the same time or at once, that would make no sense.”
Carroll and Aidan Chatwin-Davies proved, under certain assumptions, that the Second Law of Thermodynamics is the cosmic no-hair theorem: if you assume a gravitational system whose entropy is proportional to event-horizon area with a finite upper bound, you will evolve to de Sitter space—even without using Einstein's equation—mirroring how entropy increases until equilibrium.
“we proved a theorem that says literally that the second thermodynamics is the cosmic no-hair theorem.”
Functioning democracy depends on the willingness to work with and try to persuade people you fundamentally disagree with; if you treat half the country as an enemy rather than a fellow citizen, democracy cannot work, and Carroll worries this commitment is disappearing on both sides—though Republicans are far more responsible for the current threats.
“If you consider half the country to be your enemy, then there's no way for democracy to work under those circumstances.”
Any reversible theory—one where information is conserved from moment to moment and can be run forward or backward—always admits some set of operations on the state that constitutes a time-reversal symmetry; particle physicists' claim that time-reversal symmetry (T) is violated in the standard model only reflects their particular badly-chosen definition, since CPT is always conserved in any Lorentz-invariant quantum field theory and can be taken as the true time-reversal operation.
“if your theory is reversible... then there is always a set of things you can do to the state that you can call time reversal, that will be a symmetry of the system.”
Cosmic expansion does not need explanation by entropy increase—Einstein's equation and the Friedmann equation already explain it via the relationship between energy density and spacetime curvature—and the simultaneous expansion and entropy increase are both driven by the universe's special low-entropy, high-density initial conditions near the Big Bang, not by one causing the other.
“that single fact, the initial conditions of the universe in which the universe was both dense and low entropy set up simultaneously the fact that the universe is expanding and the fact that entropy is increasing.”
Quantum fields are properties of the universe located at every point in spacetime and coexist with spacetime itself—they are not created or destroyed at the Big Bang nor do they expand with the universe; only the particles, which are excitations of the fields, can be created or destroyed, while the fields have a value (possibly zero) everywhere.
“Fields are properties of the universe that are located at every point in space time. So they are co-existent with spacetime itself, they're not created or destroyed, the fields themselves.”
Einstein invented General Relativity by thinking only about reconciling Newtonian gravity with special relativity, and the resulting equation relating spacetime curvature to matter and energy is more or less unique given the desire for a relatively simple equation; the amazing fact is that this equation, derived from solar-system considerations, also works for pulsars, black holes, cosmic expansion, and the microwave background—understandable via how field theories work on long length scales.
“the equation he comes up with is more or less unique, given you want a relatively simple equation that relates the curvature of space time to matter and energy.”
It is plausible that a 50/50 quantum measurement outcome in a single atom could occasionally tip a neuron and thereby influence macroscopic human behavior, but Carroll suspects such cases are rare because the brain contains so many atoms that individual quantum outcomes usually wash out; importantly, quantum amplification is not a necessary part of explaining consciousness, which could in principle be explained classically.
“I suspect individual quantum choices or measurement outcomes usually wash out. But if you're right there on the precipice, halfway between one possibility and another, maybe a single quantum measurement event really does matter.”
To assess apparent miracles as a Bayesian, you ask whether the frequency at which you observe such events is higher than you'd expect in a miracle-free world; if the events are just as frequent as a purely physical world predicts, they are not evidence for violation of physical law, but if their observed rate exceeds the physicalist prediction, that legitimately counts as evidence against the physicalist theory.
“what they should be asking themselves is, if they're weren't any miracles, how often would we be seeing these things that we're seeing.”
Determinism is probably not correct as an operational matter for our universe because quantum mechanics introduces intrinsic randomness; even deterministic interpretations (Many-Worlds or hidden variables) cannot let you predict what you will observe next, so a Laplacean predictor would require deterministic laws unknowable to any human and a computer existing outside the universe to avoid self-reference paradoxes.
“we have very strong evidence that it's not possible to have a complete quantum state description of the universe and deterministically predict what will happen next, what will be observed next.”
Ethical arguments work essentially the same way as mathematical proofs—both derive conclusions from agreed starting points using logic—and mathematical proofs are just a subset of logical arguments; science adds empirical error bars but still uses deductive logic the same way, so once you agree on principles, ethical reasoning proceeds by the same logic, with the footnote that different logical systems (modal, higher-order) can yield different conclusions.
“I think that mathematical proofs are just a subset of logical arguments... once you agree on your principles, if that's how you're reasoning... then you use the rules of logic to derive conclusions from there.”
The scenario Carroll worries about most is not a dramatic collapse into 1984 but a quiet failure where democracy falls, power and wealth concentrate gradually, and people don't really care or get outraged—life goes on, trains run, classes are taught—until concentration becomes intolerable enough to provoke revolt; this complacency is a darker and more plausible danger than overt tyranny.
“we can't imagine that if democracy falls, it will instantly be like 1984.”
While human brains differ in capability just as bodies do, we too quickly jump from that mild claim to imagining magical ability; for the overwhelming majority of cases, including super-genius mathematicians, what looks like effortless insight is actually the visible final product of a great deal of hard work that does not make good cinema but is the real engine of progress.
“for the overwhelming majority of cases, including super genius mathematicians, what you're actually seeing is the result of a lot of hard work”
For very unlikely but very bad scenarios (simulation dystopias, quantum immortality, LHC destroying Earth), the rational response is not a calculated conviction that badness times probability is small, but a humble recognition that the probability is so low you genuinely cannot estimate it—and you must symmetrically admit equally unlikely good outcomes—so such scenarios are not worth worrying about.
“the real thing is not a conviction that the probability is so low that when you multiply it by the badness, it's still small, but a humble recognition that the probability is so low that you don't know what the probability is”
Clearly-identified opinions are not misinformation and generally should not be suppressed, but these questions should not be treated absolutely; the right question is not to invent an inviolable principle but to ask what is best in a complicated situation—Carroll's impulse is to let people hold and spread opinions while remaining open to rare circumstances where that would be genuinely harmful.
“no, opinions are not misinformation and they generally should not be suppressed.”
The finite-but-huge dimension of the universe's Hilbert space (about 10 to the 10 to the 120) comes from gravity, not the number of particles: applying the Hawking black-hole entropy formula (area of horizon over four in Planck units) to the de Sitter horizon of our accelerating universe gives an entropy of about 10 to the 120, and the Hilbert space dimension is the exponential of that entropy.
“where does this finite but big number come from if it has is nothing to do with the number of particles in the universe. The answer is gravity.”
Even if the anthropic principle works for some fine-tunings (perhaps the up/down quark masses), it cannot explain all of them, because several fine-tunings have no connection to the existence of life: the Higgs mass hierarchy problem, the extreme low entropy of the early universe (far lower than life requires), and the strong CP problem (theta-QCD being smaller than 10^-9 with zero connection to life) all demand dynamical or other mechanisms rather than anthropic selection.
“the entropy in the early universe was way, way, way, way, way lower than it needs to be to account for the existence of life, so you need more than simply an anthropic explanation.”
The cosmic microwave background anisotropies are not due to gravitational lensing—lensing contributes only a small amount on small angular scales—but rather to density fluctuations in the primordial plasma just before recombination 380,000 years after the Big Bang; the further claim that these density fluctuations originate in early-universe quantum fluctuations is a speculation, not established knowledge.
“The vast majority of what you see when you look at CMB anisotropies is not from gravitational lensing. What it's from is density fluctuations of the primordial plasma just before recombination”
In the 15 years since Carroll's dark matter/dark energy lectures, the rough picture (5% ordinary matter, 25% dark matter, 70% dark energy; dark matter likely weakly-interacting slow particles; dark energy likely vacuum energy) is unchanged, and the only progress has been negative—better limits from not finding dark energy density changes over time and not detecting dark matter particles—leaving plenty of unexplored parameter space.
“The only progress that's been made has been of a negative variety.”
Mass is not conserved—this has been known since Einstein's E=mc², which is why atomic weapons can convert mass into energy—rather it is energy that is (roughly) conserved; when the Higgs field changes value at the electroweak phase transition it gives mass to ambient particles by transferring energy from the Higgs field to those particles.
“mass is not conserved. We've known that mass is not conserved for a long time, ever since Einstein explained to us E=MC squared”
Carroll's leading candidates for global catastrophic risks over the next 50 years are nuclear war (operational weapons remain and proliferation only increases) and bioterrorism/pandemics (COVID already caused over 20 million excess deaths, within an order of magnitude of 100 million, and a more lethal pathogen that is infectious before symptomatic would be devastating), with solar flares as an under-considered wildcard given we have no data on millennium-scale events.
“I would say the two most obvious choices are nuclear war, one way or the other. We still have a lot of operational nuclear weapons on this planet.”
String Theory has been more fruitful than Loop Quantum Gravity: Loop Quantum Gravity conservatively tries to quantize Einstein's classical theory and hasn't gotten very far, while String Theory unexpectedly predicted gravity when string theorists were studying the strong interactions and has yielded profound results like holography and complementarity—results the Loop Quantum Gravity community did not discover—so if forced to choose, Carroll picks String Theory while remaining humble.
“If the theory hands you gravity when you weren't even looking for it, that's a promising kind of sign”
There is a default of enjoying a simple, relaxed life, and pushing yourself out of your comfort zone yields rewards that are less tangible and immediate—so it is a competition between your present self and your future self, where you may have to push yourself to do something and only afterward be very glad you did, just like going to the gym.
“it's a competition between your present self and your future self.”
The concept of agency makes sense only as a high-level emergent description of an immensely complex situation where you cannot model all the moving pieces, making it useful to ascribe choices to an agent; a single atom obeying the Schrodinger equation is not complex enough to warrant that language, so agency should not be assigned to individual quantum events.
“Agency, in my mind, specifically makes sense as a description of a situation where there is some immensely complex situation where you can't actually model all of the moving pieces that are doing their thing, and therefore, it becomes useful as an emergent higher level description to ascribe an agent making choices”
One's philosophical stance on mathematical realism vs nominalism—whether numbers are as real as electrons or merely a way of talking about the physical world—is not idle, because it changes how a theorist develops further theories; Carroll's nominalist leanings have led him to contemplate physics theories using only mathematics weak enough not to invoke their own consistency, illustrating how metaphysical stances feed back into actual research.
“it's literally the case that a stance towards mathematical realism is affecting the research work that I do in that case.”
In poetic naturalism, different true descriptions of the world have domains of applicability, and in familiar cases of emergence these form a nested hierarchy where higher-level theories (sociology) have smaller domains that fit inside those of lower-level theories (particle physics)—but this nesting is not necessary; one can imagine non-intersecting or partially overlapping domains, which may be relevant to strong emergence and downward causation.
“It happens to be the case that in many, many known and understood and familiar examples of emergence, all of the theories are entirely subsets of other theories.”
Consciousness will most likely be explained as an emergent higher-level phenomenon, analogous to how natural selection was understood as a logical consequence of imperfectly replicating information decades before the underlying molecular physics—so it will be illuminated by information theory, cybernetics, and brain science rather than by altering or appealing to fundamental physics.
“I would be absolutely shocked if anything about fundamental physics was useful in explaining consciousness.”
The criticism that String Theory predicts every possible low-energy physics (and therefore nothing) is sloppy: the many ways to compactify extra dimensions yield different low-energy physics, but String Theory comes into its own at high energies near the Planck scale where it makes definite predictions for string scattering, so the underlying Popperian point—that it says some things would happen and others would not—is satisfied even if we can never run those experiments.
“people make the criticism that, therefore String Theory predicts every possible thing, therefore it predicts nothing, therefore it's not a theory at all. And I don't believe that either.”
Carroll is untroubled by the prospect of multiple physical models describing nature: if they fully and perfectly describe the same thing they are simply the same model in different language (like Hamiltonian vs Lagrangian mechanics), and if they have overlapping but non-equivalent domains of applicability that together cover all of nature, that is fine too—the goal is accurate description over the widest domain, not a single universal vocabulary.
“you have multiple conceptually different models that are exactly like you said, describing fully and perfectly what nature does, then they're the same model. They're just the same thing going on in different language”
Carroll's view of racial justice in America is that a great deal of racism persists, often in hidden but impactful ways (not overt discrimination but quieter policies and actions with devastating differential effects by skin color), and that the real problem is racism itself, not the occasional overreach in accusations of racism—'you can't make an omelette without breaking some eggs.'
“racism can be kind of hidden, but still really impactful. It doesn't need to be, "Blacks cannot drink out of this water fountain."”
Spending a few hundred dollars per person on a top tasting menu is a rational way to budget pleasure: unlike the world's fanciest hotel or car which Carroll will never access, the best restaurant meals are arguably as good as any meal Jeff Bezos could eat, yet cost about the same as a vacation day at Disneyland—so for a foodie it is well worth it.
“No matter how rich you are, it's just as good as a meal that Jeff Bezos could possibly eat. And it's 500 bucks, which is again, a vacation.”
Sports leagues are far too conservative about changing obviously bad rules; even without knowing the details, Carroll strongly suspects the NFL should change its sudden-death overtime rule (where the coin-toss winner has historically almost always won), since sudden-death may suit soccer but makes no sense in American football.
“I have this feeling that sports leagues are way too conservative about changing obviously bad rules”