AMA | November 2025
What this covers
Sean Carroll answers questions from Patreon supporters across physics, philosophy, and public reasoning. The range spans quantum mechanics, cosmology, black holes, emergence, epistemology, and applied ethics—touching also on music, animal welfare, and media. The thread binding them together is a defense of a coherent worldview: Many-Worlds Interpretation for quantum reality, physicalism for what exists, Bayesian Inference for how to think, and adherence to scientific consensus tempered by humility as the standard for public reasoning. Carroll argues that the Many-Worlds Interpretation is the simplest interpretation consistent with quantum formalism, that entropy is objective despite requiring coarse-graining, and that deviation from consensus must be earned through evidence rather than asserted—a principle that applies equally to technical physics and to evaluating claims in domains like medicine and psychology where p-hacking creates systematic bias.
The conversations turn contentious and surprising in several directions. Carroll defends Many-Worlds Interpretation against the charge that it violates Occam's razor—arguing that simplicity of equation matters more than multiplicity of worlds—and sketches how moral reasoning can survive it despite all outcomes being real. He explains why Gödel's incompleteness theorem does not constrain physics, why entanglement cannot enable faster-than-light signaling despite intuitions to the contrary, and why the absence of megastructures in the cosmos is evidence not of human loneliness but of the failure of certain science-fiction assumptions. He confronts the Boltzmann Brain problem as cognitively self-defeating rather than merely uncomfortable, examines how music may function as learned cultural knowledge rather than mathematical truth, and weighs the ethics of animal suffering against future deprivation. Technical points—why larger black holes can have water-like densities, how field falloff encodes spatial dimensionality, why energy is not conserved in an expanding universe—are interwoven with foundational questions about emergence, the God hypothesis, and whether strong calculational skill in physics guarantees philosophical insight.
In this AMA, Carroll defends a coherent physicalist, Everettian, naturalist worldview—arguing that quantum mechanics is best understood via Many-Worlds, that entropy and emergence are objective despite involving coarse-graining, and that good epistemics (Bayesian reasoning, deference to consensus with humility) matter as much in science as in public life.
- Many-Worlds is the simplest interpretation consistent with the formalism and is not falsified by ontic-wavefunction results
- Entropy depends on coarse-graining but is still objective, not merely observer-relative
- Scientific credibility requires deviation from consensus to be earned through evidence, not asserted
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.
Massless particles like photons do not pose a challenge to physicalism: physicalism does not require mass, rest energy, or sharp localization, and a photon is a physical excitation of a quantum field carrying energy and momentum, so being massless and delocalized is fully compatible with being physical.
“If something like a photon has no mass, no rest energy, no proper location, and can't even be said to "exist" at a single point in spacetime, is it physical?”
Because laypeople lack the expertise to evaluate technical claims directly, they should rely on a combination of the speaker's reputation and how far the claim deviates from scientific consensus, recognizing that deviation from consensus must be earned through evidence rather than asserted.
“For technical subjects like Physics, how does a layman spot bullshit if they don't have the expertise needed to properly consider the information? Should we rely solely on the speaker's reputation and how much they deviate from consensus?”
Gödel's incompleteness theorem concerns unprovable true statements within sufficiently powerful formal arithmetic systems and does not straightforwardly transfer to physics; invoking it to explain wave-function collapse or to bar a deterministic theory of everything is a category mismatch unless the physical theory is genuinely cast as such a formal system.
“I wonder if something similar might apply in physics... Perhaps it could even be one of the possible explanations for the wave function collapse, that we cannot fundamentally prove, what happens.”
Modern scientific culture, particularly in fields like medicine and psychology that rely on statistical significance thresholds, incentivizes false positives through p-hacking, whereas fundamental physics is less affected; this raises concern about whether the reward structure of science systematically generates spurious results.
“While this seems to be much less of a problem within fundamental physics, fields such as medicine and psychology appear far more effected.”
The Boltzmann Brain problem is a genuine problem not because being a Boltzmann Brain is discomforting, but because a cosmology that predicts you are most likely a randomly fluctuated brain is cognitively unstable: it undermines the very reasoning and evidence used to arrive at that cosmology, making the theory self-defeating.
“Alternatively, maybe the Boltzmann Brain concept is a problem not because it's impossible or illogical, but because it's simply a dead end: unprovable and undisprovable? Am I missing something?”
In Mad Dog Everettianism the only fundamental entity is the wave function of the universe, so 'size scale' itself is emergent; consequently the 'up/down the size scale' metaphor for fundamentality is misleading, because the fundamental level is not located at the smallest scale but is the global quantum state from which spatial scale emerges.
“according to Mad Dog Everettianism, the fundamental is the universe, as described by its wave function - no? In which case, the fundamental is all the way *up* the size scale - or whatever one says when "size scale" is *itself* an emergent feature.”
Two thinkers can apply Occam's razor and reach opposite conclusions about quantum reality—one favoring the parsimony of a single world, the other the parsimony of Many-Worlds—because the razor disciplines theoretical postulates (equations and entities) rather than the number of objects, so simplicity of the formalism can favor Many-Worlds even though it multiplies worlds.
“it's the fact that the very same razor of occam has led you to opposite conclusions. You, to the cleanliness of many worlds, and him to that of only one.”
Even though all outcomes of a quantum experiment are real in Many-Worlds, this does not require revising moral theory, because the weighted reality of branches plays the same role as probabilities did; a bet keyed to a deterministic fact (a math conjecture) differs from one keyed to a quantum coin, but moral reasoning can treat branch weights as the relevant probabilities.
“on many different occasions, you say that the Many-Worlds interpretation doesn't bring any new insights into the purported moral theory.”
The universe has equal numbers of protons and electrons because it is electrically neutral overall; net charge is conserved and was near zero from the start, so charge-creating processes produce balanced amounts, ensuring matter is neutral rather than this being a separate fine-tuned coincidence.
“I'm having trouble understanding how we ended up with a universe that has an equal number of protons and electrons. Is there some process that ensures they are created in equal amounts?”
Because the Schwarzschild radius scales linearly with mass while volume scales as the cube of the radius, larger black holes have lower average density; supermassive black holes can have densities comparable to or below that of water, which has counterintuitive implications for how we think about black-hole 'compactness'.
“the internet told me that black holes become less dense the bigger they get. I dont believe I've heard you on this subject, would you mind teasing out this subtlety?”
Gravitational and electromagnetic field strength falls off as 1/r² rather than 1/r³ because the flux from a source spreads over the surface area of a sphere, which grows as r² in three spatial dimensions; the power of the falloff thus encodes the dimensionality of space.
“Somebody told the inverse square law is proof we live in 3 dimensions. Can you explain why this is true? Why isn't it 1/r^3 for 3 dimensions?”
When matter collapses into a black hole the mass does not 'go' anywhere problematic; the energy that was in matter is carried by the curvature of spacetime itself, so a black hole's mass is a property of the gravitational field, and asking where the matter went conflates substance with the geometric description.
“it was mentioned that black holes are not things of matter but curvatures in spacetime. So my question is, where does this mass go?”
Energy is not globally conserved in an expanding universe because conservation of energy follows from time-translation symmetry, which the expanding universe lacks; thus dark energy can drive accelerated expansion and photons can lose energy to redshift without that energy going anywhere, since there is no global energy-conservation law to violate.
“cosmic expansion violates the conservation of energy by basically creating new energy from nothing, which is okay because conservation of energy only holds in a static universe.”
Quantum entanglement cannot be used to transmit coordinated signals faster than light, so the science-fiction scheme of using entangled particles to trigger synchronized galaxy-wide revolts fails: measurement outcomes are random and locally indistinguishable, and no chooseable information passes between distant locations.
“the resistance can send entangled particles from a central location to all the other locations... because the particles are entangled, it ends up being the same plan across the galaxy.”
In real physics practice, one does not simply guess a Lagrangian and check until it works; Lagrangians are constrained by required symmetries and known behavior, so the 'just assume this Lagrangian' pedagogy compresses a disciplined process of writing down the most general form consistent with the symmetries and then fixing terms by experiment.
“does it work like this in real life? you invent a Langrangian and check until you get one right? and has use of AI on trying to - given a system desc - come up with a Langrangian been increasing?”
Hypotheses that fail to meet scientific consensus, such as intelligent-design critiques of evolution, should be clearly distinguished from established science (e.g., not shelved alongside biology textbooks), because misrepresenting their epistemic status to the public misleads non-experts even if free speech permits their publication.
“a very annoying habit of bookstores staff of locating books by Michael Behe and friends that question evolution in favour of creationism, on the same shelf real biology books are located.”
Given a prior that advanced life is common in a vast universe, the absence of observed Von Neumann probes, Dyson spheres, and other megastructures is better interpreted as Bayesian evidence that such self-replicating machines cannot achieve the capabilities imagined in science fiction, rather than as evidence that humanity is alone.
“So I conclude that the fact we haven't found any is actually evidence that such machines cannot achieve the capabilities envisioned in science fiction.”
Entropy depends on a choice of coarse-graining (macrostates), but given that choice it is an objective feature of the system rather than something purely relative to an observer's senses; you cannot make entropy go away by changing what you happen to perceive.
“Suppose though both coffee and milk were same temperature and we only saw in infrared - we'd see nothing changing, entropy would be constant. So is entropy relative to the observer - Carlo Rovelli seems to imply it is”
Although many physicists claim the best physicists make the best natural philosophers, technical calculational skill (e.g., doing hundreds of integration-by-parts problems) is partly separable from philosophical insight, so a strong physicist is not automatically the most interesting from a foundational or philosophical standpoint.
“the best physicists are going to make the best natural philosophers. But I am skeptical of that answer, because it seems to me that the best physicists are not always the most interesting from a philosophical point of view.”
The strongest moral case against eating farmed animals is not the deprivation of their future prospects but the lifelong suffering imposed on them—overcrowding, force-feeding, artificial insemination, and forced separation from offspring—so that a wild animal hunted by a predator may actually have a better life than a factory-farmed one, even though both are killed.
“Factory-farmed animals on the other hand are treated like machines for optimizing human profit, forced to live in overcrowded and unnatural conditions, force fed, artificially inseminated and forcefully separated from their cubs, etc.”
A person can be conscious and exercise agency without forming durable memories, as illustrated by interacting coherently (counting backward by sevens) during a period of which one retains no recollection; this suggests consciousness and short-term agency are dissociable from long-term memory formation.
“Obviously I was interacting with the doctor as I was saying "86 ... 79" when I came to, so I must have been showing signs of sentience in the ER. I have difficulty seeing how I could have agency without memory?”
When a Type-2 emergent macro system (like politics) arises from a micro-level theory (like physics), borrowed concepts such as 'phase change' may be more than metaphor, raising the question of whether macro systems meaningfully inherit real properties from the micro theory or whether cross-level similarities are always mere analogies.
“When we use aspects of the micro theory — like the idea of "phase change" — to describe behavior at the macro level, is this just metaphor, or are some of these elements potentially or necessarily "real" as you go up levels?”
In the Friedmann equations the spatial curvature and total energy density are linked by Omega_k = 1 - Omega_0, but this is a constraint relation rather than a one-way causal determination: neither curvature nor energy content strictly 'causes' the other; they are mutually constrained features of a single solution.
“Is there a causal relationship between the curvature and energy contents? Does one "determine" the other, or can we only say that they are related?”
Because we lack a settled theory of consciousness, we cannot be confident that even narrow, non-general AIs are entirely non-conscious, which raises the possibility—however small—that using them could be unethical; the absence of an explanatory criterion for consciousness leaves the moral status of AI systems genuinely uncertain.
“Lacking any major explanation of consciousness how can we be confident that even 'dumb' (non general) AIs are not somewhat conscious, and thus that we are not using them unethically?”
Emotional responses to music may function like a culturally shared emotional language rooted in physiology and acquired through repeated media exposure—a form of common knowledge—rather than being a universal mathematical truth, implying that 'what sounds good' is partly a repeatedly learned cultural framework.
“I suspect emotional responses to music are a form of culturally shared emotional language, rooted in our physiology, akin to common knowledge.”
To analyze the God hypothesis scientifically, it must be defined precisely enough to generate predictions about what a universe created by God would look like; without a sufficiently complete account of expected observations under the hypothesis, it cannot be assigned likelihoods and compared to alternatives in a Bayesian framework.
“how well defined does "God" have to be? How complete does your theory of how a universe would look under the God hypothesis have to be?”
Although the wave function branches constantly, everyday macroscopic decisions like choosing ice cream flavors are almost always the same across branches, because the neural processes determining the choice are effectively classical; genuine quantum indeterminacy rarely amplifies to flip macroscopic outcomes outside engineered measurement settings.
“could there be events in my brain that branch the wave function where I'm taking vanilla in half the branches and chocolate in the other? Or is more like ... I'm still having chocolate in all the branches”
Even if larger-scale PBR-type experiments conclusively establish that the wave function must be physically real (ontic rather than epistemic), this would not falsify the Many-Worlds Interpretation, since Many-Worlds already treats the wave function as the real, physically existing entity.
“if future, larger-scale experiments conclusively show the wave-function must be physically real, would that outcome falsify the Many‑Worlds Interpretation of quantum mechanics, or could Many-Worlds still survive in that scenario?”
Solving a fine-tuning problem by positing a new dynamical field is not necessarily as arbitrary as tuning a constant to a precise value, because a field can provide a mechanism that dynamically drives the quantity toward the observed value, converting an unexplained coincidence into an explained consequence—though whether it succeeds depends on whether the field genuinely reduces the required tuning.
“Do you think adding new fields to make things work is in some sense just as arbitrary as tuning the numbers just right? Or is that different?”
The Koide formula—a simple function of the charged lepton masses equal to 2/3 that became more accurate with improved measurements—illustrates the difficulty of distinguishing meaningful physics from numerology; the strength of a numerical coincidence can be partly quantified by whether it gains Bayesian credibility (e.g., improving precision after parameters are removed) rather than relying on a prior chosen to fit.
“It gets good Bayesian points for getting much more accurate (threw away an adjustment term) after better mass measurements, and now seems uncomfortably accurate for it to be random chance (correct to 6th sig fig). More broadly, where is the line between numerology and science?”
Physicists' surprise at constant values rests on a 'naturalness' prior in which dimensionless ratios are expected to fall in an order-one range relative to fundamental scales like the Planck mass, so a value like M/10^120 is judged surprising and fine-tuned, while M is judged natural—an assumption a non-physicist would not hold without that prior.
“some physicists have a prior expectation that some constants should fall into a "natural" range, where that range is somehow related to the Planck mass.”
Under the Everettian (Many-Worlds) interpretation there is no fundamental randomness; apparent quantum randomness is self-locating uncertainty about which branch you are in, more like a deterministic process whose 'seed' is the universal wave function being partially revealed by measurement, which is closer in spirit to Bayesian than frequentist probability.
“Under Everettian QM it seems that there is no such thing as true randomness. Just psuedo-randomness where we don't know the seed of the pseudo-random number generator ( in this case the seed is a vector in Hilbert space).”
Brewing tea strongly in little water and then diluting it yields the same final solution as brewing weakly in the full amount of water, because the chemical extraction depends on equilibrium concentrations and total solute, not on the order of operations—the dilution water lacks no 'essence' relative to water that brewed directly with the leaves.
“is that fundamentally different than steeping the full amount of water less strong? That extra hot water that I would add to dilute the strong tea, does that lack some essence”
The habitual framing of quantum mechanics as 'weird' or 'counterintuitive' is unhelpful and inconsistent, since Newtonian phenomena (like not feeling the Earth move) are equally counterintuitive yet not prefaced with warnings; treating QM as inherently bizarre primes audiences to expect incomprehensibility rather than understanding.
“no one start a lecture on Newtonian physics with "counterintuition warning" (e.g. we don't feel e.g. Earth moving, that's weird). Shouldn't we retire these stereotypical phrases about quantum mechanics”
Choosing how much lifespan to sacrifice for pleasure is a genuine optimization problem, since neither pure instant gratification nor pure longevity maximization is prudent and the two are not mutually exclusive; the rational stance treats marginal life-shortening pleasures (tasty food, alcohol, risky hobbies) as trades to be weighed rather than categorically accepted or rejected.
“there are many practices that clearly decrease the average lifespan while enriching one's experience (e.g. eating tasty food, drinking alcohol, BASE jumping, driving). Would you give up one year of your life to increase participation in such things?”
If body parts are progressively replaced, the question of whether one remains the same person tracks the Ship of Theseus paradox; personal identity plausibly centers on the brain, so replacing everything except the brain preserves identity, while partial brain replacement with chips raises a genuine and unresolved boundary problem about how much of the brain can change before the person changes.
“at what point, if ever, would a human person become someone/something else if enough body parts were replaced? The "Ship of Theseus" paradox comes to mind.”
A single time-varying cosmological constant could in principle account for both inflation (an extreme early value driving rapid expansion) and present dark energy (a small residual value after decay), unifying the two accelerated-expansion epochs—though making this work quantitatively is the hard part.
“could a time-varying cosmological constant explain both inflation and dark energy? I.e. the cosmological constant had an extreme value right after the Big Bang, giving rise to a rapid expansion of the universe (i.e. inflation), and then quickly decayed to near zero”
MOND (Modified Newtonian Dynamics) does not require rejecting the Equivalence Principle; it modifies the dynamical law relating force and acceleration (or gravity) at low accelerations rather than the equality of inertial and gravitational mass, so framing MOND as an equivalence-principle violation misstates the proposal.
“Does accepting MOND theories mean that we are rejecting the Equivalence Principle? If yes, does that make the various MOND solutions even less credible alternative to Dark Matter?”
Penrose's objective-collapse-based quantum consciousness model is fundamentally incompatible with Many-Worlds, because objective collapse posits a real, dynamical reduction of the wave function that eliminates branches, whereas Many-Worlds denies any such collapse occurs.
“His later Quantum Consciousness ideas depend on objective collapse, which seems directly at odds with MWI. Do you see any possible way that Penrose's objective-collapse-based Quantum Consciousness could fit within a Many-Worlds framework, or are those two views fundamentally irreconcilable?”
Claims that the universe being a simulation has been mathematically 'ruled out' via Gödel, Tarski, and Chaitin results are likely overstated, since such logical-limitation theorems do not straightforwardly bear on computational physics programs like cellular-automata models of physics; the burden is to show the argument actually constrains those proposals.
“I just read an article that claims that the Universe being a simulation has been ruled out (Mir Faizal at Univ of BC, et al)... it did reference Godel, Tarski, and Chaitin”
An open democratic society with a strong scientific culture may sit at a critical point analogous to a phase transition between authoritarianism/theocracy and anarchy, where information can propagate efficiently to create long-range correlations while preserving enough diversity and informational content—drawing on the physics of phase transitions to model healthy democracy.
“an open democratic society with a strong scientific culture might represent a critical point, like a phase transition, between two extremes: authoritarian dictatorship or theocracy on one side, and anarchy on the other?”
Faced with bad political developments, sadness, frustration, and anger at discrete decisions are more appropriate and useful than despair, because despair is paralyzing and unproductive whereas targeted negative emotions can motivate responses to specific decisions.
“I feel sadness, frustration, and anger at discrete decisions by this administration. But not despair. Should I feel despair? Do you? And is that a useful emotion?”
When only a small fraction of registered voters participate (e.g., 15% in a byelection), an organized minority can prevail and impose policies a non-voting majority would oppose, raising the tension between charitably excusing non-voters' busy lives and legitimately resenting minority rule produced by abstention.
“is it okay to be annoyed that a minority is in position to impose their will on a majority that does not agree but did not even bother to show up?”
Public calls to prohibit the development of superintelligence, signed by figures like Hinton and Bengio, merit serious engagement, but the practical force of such statements depends on whether superintelligence is well-defined and whether a prohibition is enforceable, distinguishing the symbolic value of expert warnings from a workable policy.
“There has recently been a public call for a prohibition on the development of superintelligence, signed among others by Geoffrey Hinton and Yoshua Bengio. What is your take on such statements and the subject matter?”