AMA | March 2021
What this covers
This Ask Me Anything episode gathers questions from Patreon supporters on topics spanning physics, philosophy, and technology. Sean Carroll selects questions where he judges he has something substantive to offer, grouping related ones together. The episode works through conceptual reframes that Carroll argues dissolve apparent paradoxes: how to understand time in relativity, where particle discreteness comes from in quantum field theory, and how to think about free will and expert credibility in intellectual life.
The physics content moves across several domains. Carroll clarifies proper time as path length as the right way to think about time in special relativity, explaining why there is no objective fact about distant simultaneity and why talk of time "flowing faster" near black holes is nonsense. He addresses how discreteness from boundary conditions — the same mechanism that produces discrete harmonics on a tied string — explains why particles appear discrete in quantum field theory. He also covers Hawking radiation, dark matter accretion, cosmological curvature, and why gravity differs fundamentally from ordinary quantum fields. On quantum mechanics, he discusses the difference between quantum entanglement and classical correlation, and contests David Albert's interpretation of the wave function by appealing to basis independence as non-physicality.
Beyond physics, Carroll addresses how to judge expert credibility through steelmanning vs strawmanning as a trust signal, and notes how selection effect in expert opinion can create misleading distributions — most philosophers are atheists but most philosophers of religion are not. He also characterizes AI risk as rooted in unpredictability and value misalignment rather than raw capability, highlighting the dangers of anthropomorphizing AI as a fallacy.
Carroll argues that many apparent puzzles in physics (time dilation, particles vs fields, quantum interpretation, free will) dissolve once you adopt the correct conceptual framing — time as proper time along trajectories, particles as boundary-conditioned modes of fields, and free will as an emergent compatibilist description — and that intellectual life requires balancing trust in experts against principled, evidence-driven heterodoxy.
- Time should be understood as proper time measured along individual trajectories, not as a universal flow, which dissolves special-relativity paradoxes
- Particle discreteness arises from boundary conditions on the wave function of quantum fields, analogous to a vibrating string
- Free will debates are about whether you obey laws, not whether laws are deterministic; compatibilism captures the useful emergent description
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.
Apparent expert distributions can be driven by selection effects: most philosophers are atheists but most philosophers of religion are not, just as most physicists may treat quantum interpretation as semantics but specialists in foundations do not, because people who hold a given view are unlikely to specialize in a subfield premised on its opposite.
“It's like most philosophers are atheists, but most philosophers of religion are not... if you're an atheist, probably not going to specialize in philosophy of religion.”
When experts disagree and you cannot evaluate their reasoning directly, you should judge trustworthiness by how they treat opposing views: whether they understand and fairly represent alternatives rather than strawmanning them, and whether they honestly acknowledge weaknesses in their own position rather than hiding them.
“listen to how they frame the controversy... how much do they understand the alternatives to their point of view, are they simply dismissing those alternatives? Are they being fair, are they strawmanning the alternatives?”
In special relativity there is no objective fact about whether a distant event (like a Mars landing) has 'already happened,' because observers moving at different speeds disagree; the only physically real notion is whether the event is in your past light cone, not whether it is simultaneous with 'now.'
“being in our past light cone is in fact a real physical thing, happening at the same time is not a real physical thing, that's exactly what special relativity is about. You cannot define simultaneity for events that are very, very far away in space.”
There is a real difference between erasing a memory and a memory not yet having been created: erasing (like pouring water over a footprint) is an irreversible entropy-increasing process where you can't recover the past, whereas running time backward removes the footprint by lowering entropy along the trajectory it came from, so Landauer's principle does not symmetrically apply toward the past.
“I think there is a difference between being erased and not yet having been created.”
Black holes mostly absorb ordinary matter rather than dark matter because ordinary matter undergoes dissipation — atoms bump into each other, emit photons, lose energy and spiral in forming a shining accretion disk — whereas dark matter interacts too weakly to dissipate, so it zooms past unless it hits dead-on, and any dark matter absorbed forms no detectable accretion disk.
“the reason why ordinary matter gets absorbed by black holes is because ordinary matter undergoes dissipation... it loses energy and spirals into the black hole. So, mostly black holes are gonna be eating up ordinary matter, not dark matter.”
The idea that every particle has an antiparticle is misleading: only particles carrying a conserved quantity like electric charge have distinct antiparticles, while particles like photons are their own antiparticles; the particle/antiparticle classification is a convenient but non-fundamental scheme, with the fundamental facts being charges, symmetries and interactions.
“Some particles have their own anti-particles and some don't. The idea of antimatter is a convenient kind of classification in some cases, but it's not a fundamental idea.”
Quantum entanglement is fundamentally different from classical correlation: with Bertlmann's socks or a shared random seed there is a pre-existing fact about each item's state, but entangled spins have no definite individual state, and crucially you can choose to measure along a different axis (X vs Z) and still get correlated results, so the correlation isn't determined until you choose what to measure — making it intrinsically quantum and not classically analogizable.
“Entanglement is not correlation. I mean, it is correlation, but it is more than that. It's different than classical correlation.”
Time as measured by observers carrying clocks (proper time) is exactly analogous to distance along a path in space: different trajectories between the same two events accumulate different elapsed proper times, so time is not a universal objective flow but a quantity measured along individual paths, with the longest time between two events being the straight-line path.
“special relativity says that the time between two events as measured by observers carrying their watches is like that, it's like distance. It will be different for different observers”
It is nonsense to say time 'flows more slowly' near a black hole or at high speed; time moves at one second per second for every observer, and you should only compare total elapsed proper time when trajectories reconvene, never compare rates of time flow across distant frames.
“Time moves at one second per second. When you're far away in relativity, you just shouldn't compare the rate at which time flows, you should only talk about the rate at which time is measured by individuals, and that rate is always one second per second.”
If the Higgs field underwent electroweak false vacuum decay and jumped to a much larger value, the amount of symmetry breaking would change but the pattern would not: SU(2)xU(1) would still break down to U(1), preserving electromagnetism and the W and Z bosons, because changing which symmetry breaks would require a different field charged under electromagnetism to get a vacuum expectation value.
“the amount of symmetry breaking would change, but which symmetry was broken would not change. There would still be electromagnetism, there'd still be the W and Z bosons”
Unlike ordinary quantum field theory where any fixed mode can hold arbitrarily large energy, in gravity concentrating too much energy in a small region collapses it into a black hole; this simple fact, prior to holography or string theory, shows gravity is fundamentally different from quantum field theory and cannot be understood just by building a cleverer field theory.
“In gravity, if you try to put a lot of energy into a small region of space, the whole thing just collapses into a black hole.”
According to the philosophy of effective quantum field theories, theories valid at one energy scale typically remain valid below that scale but may fail above it, so new particle physics is much more likely to be found by going to higher energies than lower ones; low-temperature/low-energy new physics is possible but not expected.
“quantum field theories that work at one energy scale typically work below that energy scale... but they might not work at energies above that energy scale. There really is a difference between going down in energy and going up in energy.”
Particles arise as discrete phenomena in quantum field theory the same way a tied-down vibrating string has discrete harmonics: boundary conditions force discreteness; specifically, the wave function over field amplitudes must go to zero for arbitrarily large amplitudes (since infinite energy is impossible), and this boundary condition yields a discrete set of vibrational states interpreted as zero, one, two particles, etc.
“That phenomenon is why particles arise in quantum mechanics and in quantum field theory.”
Hawking radiation shrinks a black hole because energy is observer-dependent in general relativity: although a particle of an in-falling pair is absorbed, a distant observer attributes negative energy to it, so the black hole's total energy decreases by exactly the magnitude of the positive energy carried away by the emitted particle.
“from the point of view of someone who's very, very far away, they attribute a negative energy, a negative mass effectively, to the particle that falls in. So yes, the black hole... is absorbing a particle, but it's absorbing a particle with a negative energy. So the total energy of the black hole goes down.”
The Second Law that entropy increases follows from two things: the statistical definition of entropy (high-entropy macrostates correspond to vastly more microstates) plus a boundary condition of low entropy at one end of time, which we define to be the past; there is no pre-existing definition of past and future beyond the direction of lower versus higher entropy.
“There's no pre-existing definition of past and future, there is the direction of lower entropy and the direction of higher entropy, and we label the direction of lower entropy to be the past.”
Dark energy does not push apart gravitationally bound systems like the Earth-Sun or Milky Way-Andromeda; the universe's expansion is only an approximation valid on large scales, and a spherically symmetric expanding universe has provably zero effect inside a bound region, analogous to zero electric field inside a charged spherical shell.
“as long as the universe around us is expanding spherically symmetrically around us, there is literally no effect on what goes on here in the solar system.”
Because we move much more slowly than light — traveling far more through time than through space — it is the warping of the time component of the spacetime metric, not the stretching of space depicted in rubber-sheet diagrams, that is mostly responsible for Newtonian gravity, the inverse-square law, and why apples fall and we don't float away.
“it's the warping of the time component in the metric that is actually more responsible for gravity in the Newtonian limit... The fact that apples fall from trees, the fact that you do not float away from the Earth, this is mostly because of the warping of that time coordinate”
An infinite amount of time does not guarantee that a system breaks out of equilibrium or returns to any particular state, because there are competing infinities: if the space of possibilities is bounded and dynamics are reversible the system explores everything, but if both time and the space of possibilities are infinite, a region can wander into equilibrium and never leave, just as walking forever along increasing real numbers never returns you to one.
“If you have both infinite time but also an infinite space of possibilities, then any given region of the universe can wander into an equilibrium and never leave.”
Anti-de Sitter space is a popular toy model for quantum gravity because Maldacena's mid-1990s AdS/CFT correspondence shows that quantum gravity with anti-de Sitter boundary conditions is related to (perhaps identical to) a non-gravitational conformal field theory living on the boundary, letting physicists use understood non-gravitational physics to learn robust features of quantum gravity that may transfer to our real (de Sitter-like) universe.
“quantum gravity with anti-de Sitter boundary conditions is related to, and maybe, is exactly the same as a quantum field theory without gravity, living on the boundary. That's the AdS/CFT correspondence”
Although cosmological curvature is often described as three options (positive, negative, flat), it is better understood as a continuous real number — the radius of curvature ranging from negative to positive infinity — so a universe can look flat simply because it has a very large radius of curvature; current measurements are compatible with zero but also with a large positive or negative radius, and better data could reveal slight curvature.
“what we're measuring when we measure the curvature of the universe is a real number, not a discrete choice of three different possibilities.”
Strictly speaking, the space between the Earth and Sun is not 'cold' but simply has no well-defined temperature, because temperature only properly applies to systems at or near thermal equilibrium; that region is a paradigmatic non-equilibrium situation (hot in the Sun's direction, cold and dark elsewhere), and additionally space is not a thing you can heat up since it is the medium through which things travel.
“it's not that space is cold in between the Earth and the Sun, it's that space doesn't have a temperature, strictly speaking, it's not a system that is in thermal equilibrium.”
Against David Albert's wave-function realism (which takes a function on 3N-dimensional configuration space as fundamental), Carroll argues the truly real thing is just the abstract vector in Hilbert space, because the wave function can be equally represented in position space, momentum space, or infinitely many other bases — none privileged, like coordinate systems in relativity — and there is just enough structure in Hilbert space to reconstruct the world.
“the thing that is real in my mind is the vector in Hilbert space.”
The real AI risk is not the simple-minded worry that machines become so much smarter that we become lesser beings, but that any artificially created intelligent systems will be so wildly different from human intelligence, values, and morals that they will be totally unpredictable; the common discussion is naive in anthropomorphizing AI and assuming we can 'bake in' our values, since truly self-aware systems would decide their own values.
“the real problem is that these systems are gonna be so different than what we think of as an intelligent agent, that they'll be totally unpredictable. And that's a worry.”
When you are a true non-expert and there is a consensus among experts, you should place a very high (though not 100%) credence on that consensus simply because they are experts, unless you have good informed reasons to doubt them; this is the rational default for laypeople.
“if you literally know nothing about a field, then your default should be to give most of your credence to what the consensus of experts is.”
In the many worlds interpretation, branches that descend from a past version of you contain different people who are no longer you, like identical twins; so while someone descended from you may have won a lottery, that winner is not you, and the overall collection of your descendants likely lost money.
“All those other branches have different people that descended from the same past you, but now are no longer you, they're different, just like if you had an identical twin.”
We have very good reasons to believe that future discoveries like dark matter, Higgs bosons or grand-unification particles will not produce technological change, because we cannot interact with them: the Higgs decays in ~10^-21 seconds and dark matter passes right through matter (proven by failed direct-detection efforts); the last fundamental-physics discovery with technological implications was around the 1950s, and all technological change since has been recombining already-known ingredients.
“there are very good reasons to believe the dark matter or Higgs bosons or anything like that, grand unification particles are not going to lead to technological changes, mainly because we can't interact with them very much.”
When humans design systems they optimize for comprehensibility and a limited set of foreseen failure modes, while nature, designing for no specific purpose, produces more multi-purpose, flexible systems better able to adapt to unforeseen circumstances; the key issue is robustness rather than complexity, localness, or entropy.
“since nature does not design for specific purposes, we end up with organisms that are just more multi-purpose, and in a sense, much more flexible, much more able to adapt to unforeseen circumstances.”
The cosmic microwave background has a rest frame in which radiation is isotropic, and our motion relative to it produces an observed temperature dipole; but this is not a 'preferred' frame in violation of relativity, because relativity only forbids a rest frame intrinsic to the vacuum, not rest frames defined by stuff like the CMB.
“there's absolutely nothing wrong with having a preferred rest frame in special relativity... What the real rules are, is that there's no rest frame in the vacuum.”
General relativity and quantum mechanics seem incompatible for two distinct reasons: technically, quantizing gravity is non-renormalizable so quantum corrections become large and uncontrollable with infinities everywhere; conceptually, naively applying quantum rules to GR yields the Wheeler-DeWitt equation in which the wave function of the universe does not evolve in time at all — the 'problem of time.'
“this is a fancy way of saying that quantum gravity is not re-normalizable”
Classical macroscopic forces like electromagnetism and gravity are accumulations of identical boson particles piling on top of each other; fermions cannot accumulate this way (Pauli exclusion), which is why there are no noticeable macroscopic fermion forces, and a classical gravitational wave is a huge number of gravitons piled together.
“bosons can accumulate and therefore give rise to big classical forces, whereas fermions cannot accumulate, that's why we do not have fermion forces noticeable in the real world.”
You cannot decisively violate the Born rule with a single anomalous run; getting an improbable outcome (like a million identical results when 50-50 is predicted) never crosses a threshold to declare the rule wrong, but as a good Bayesian you continuously lower your credence in the rule as observed data deviates further from its predictions, exactly as with any other physical theory.
“as a good Bayesian what you should be doing is saying... the more I deviate from the prediction of the Born rule, the lower my credence is that the Born rule become... Is... Should be accepted”
The 'cosmological constant' is a label for the vacuum energy — the energy inherent in empty space itself, present in every cubic centimeter even with no particles or radiation — so it is a real property of spacetime, not merely a number in a formula, and that is how it can be dark energy.
“it's the energy inherent in empty space itself, it's a thing, it's not just a number in a formula... That's a fact about empty space itself, and that fact is the vacuum energy, aka the cosmological constant.”
Spacetime does not need to 'know' the distance between two masses to impute the right gravity; spacetime is not empty but contains a metric tensor field at every point that encodes its geometry and obeys Einstein's equation, so the metric stretches between objects locally, which is why there is no spooky action at a distance in general relativity.
“spacetime contains fields at every point, in particular in this case, it contains the metric tensor field... That's why there's not any question of spooky action at a distance in general relativity.”
Genetic and cross-cultural data showing that moral qualities like kinship and heroism are encoded in our genes are all facts (ises) and provide zero evidence that we can derive ought from is, because 'ought' is a logically different property than 'is' and cannot be derived from it, just as adding even numbers can never yield an odd number no matter how many trials you run.
“the inability to derive ought from is, is simply a logical impossibility... Ought is a different property than is. It can't be derived from it.”
Contrary to the prominent view (held by Frank Wilczek and others) that quantum calculations suffice and interpretation is just semantics, the interpretive questions — whether the wave function represents reality or is a calculational tool, whether there are hidden variables, objective collapses, or branching universes — are direct physical questions about the fundamental nature of reality, not semantics.
“I completely, wildly, dramatically disagree that the interpretation is just semantics. There are physical questions. Is the wave function a representation of reality? Or is it just a calculational tool”
In a world with gravity, any region of space likely has a finite-dimensional Hilbert space, because trying to pack too much energy into a region creates a black hole, black holes have finite entropy (per Hawking), and entropy relates to the dimensionality of the Hilbert space; the total dimensionality remains unknown because we don't know how many local regions of space exist.
“the Hilbert space of quantum gravity is locally finite dimensional. And the reason why is because... when you try to put a lot of energy into a quantum field in a world with gravity... you create a black hole”
Carroll is a compatibilist: the laws of physics are not overcome by willpower (no libertarian free will), but it remains overwhelmingly useful to describe human beings as decision-making agents who respond to reasons, so giving advice to take initiative is consistent with denying libertarian free will; he suspects nearly everyone is implicitly a compatibilist since even free-will skeptics try to convince people and use the language of choice and morality.
“A compatibilist is someone who recognizes that the laws of physics are the laws of physics, I cannot overcome them, and yet there is an absolutely clear sense of volition and decision-making and free will that we have at the emergent level of being human beings.”
Penrose's Gödel-based argument that consciousness is non-computational fails because it assumes humans can 'just see' that a formal system is consistent, but we cannot prove that consistency; the human brain is physical and sloppy, often attributing unwarranted certainty to its thoughts, so the appearance of reaching conclusions a computer cannot is just giving ourselves too much credit.
“I think that we are sloppy thinkers, our human brains, and we can often attribute certainty to some of our thoughts that are in fact don't deserve to have certainty attributed to them... But I think that's just giving us too much credit, honestly.”
The free will question is entirely about whether you are governed by laws, not whether those laws are deterministic; quantum indeterminism is irrelevant to libertarian free will because it is not your will making the random outcome occur, so anyone framing free will in terms of determinism versus indeterminism should not be taken seriously.
“free will has nothing whatsoever at all, even a little bit to do with determinism... what matters is not whether the laws of physics are deterministic or not, what matters is just that there are laws.”
While heterodox public intellectuals should exist, one must be wary of a bias whereby people elevate the views of speakers (e.g., black heterodox intellectuals downplaying systemic racism) precisely because those views are what the listener wants to hear; it is more important to face seriously the orthodox views one does not want to hear, and to value being correct rather than valorizing heterodoxy for its own sake.
“there's a huge bias that you can fall into in saying, 'Oh, you're a good black person, you're saying what I wanna hear, let me talk to you and elevate your opinions.'”
Practical questions about punishing wrongdoers or treating addictions and mental disorders are largely orthogonal to the free will debate; these are consequentialist questions about how to use biology, psychology, pharmacology and the legal system to achieve agreed-upon better outcomes, and the discussion would be far better if no one used the words 'free will' in it at all.
“To me, these questions are largely orthogonal to the free will question. These are all questions about which I am very much a consequentialist.”
Going to higher collision energies is the most reliable way to find new particle phenomena because it is almost guaranteed to eventually reveal something new, whereas clever low-energy experiments (like axion light-shining-through-walls) are highly model-dependent and any single one is unlikely to be the right answer.
“if you just go to high energies and smash particles together, eventually you're guaranteed almost to find something new.”
Mathematical structures are real in some sense but not the same sense that tables and chairs are real; this question of what is real in mathematics is the most fundamental and unresolved issue in the philosophy of mathematics that Carroll personally struggles with.
“there are senses in which mathematical things are real, but not the same sense as the tables and chairs are real.”
Scientific discoveries in physics or biology (such as which species feel pain) can affect how we behave given our moral principles, but they do not affect meta-ethics — the justification of those moral principles — because the discovery of a fact never changes how one justifies a value.
“physics can absolutely have implications for how we behave in the world given what our moral principles are... But these discoveries in neither physics nor biology directly affect what I take to be the moral rules that I live my life by.”
Modern fundamental physics is stuck because its highly successful models (Standard Model, general relativity, standard cosmology) fit all experiments, so progress toward better theories must come from guessing rather than from experimental surprises, which is why beyond-the-standard-model work is highly speculative.
“our guidance to finding better answers is not coming from experimental surprises, as it usually does in physics. So we're reduced to a situation where in order to get better answers, we kind of have to guess”
Unlike science, which requires genuine expertise to practice, art should be democratized so that more people play instruments and paint regardless of talent, because modern technologies enable participation and the value lies in doing it and enjoying it rather than doing it well.
“I wished I lived in a world where we didn't try to confine the practice of art to experts. Science is a different thing, because the practice of science requires a certain amount of expertise, but in art, that's the difference between doing it and doing it well”
A good university should not devote all faculty positions to people writing only for refereed journals; it is valuable to devote some fraction of resources to hugely influential public-facing intellectuals doing interdisciplinary work and serving as role models, because a diverse intellectual ecosystem is important, even though you would not want every professor to be like that.
“a good university should not devote all of its faculty positions purely 100% to people whose noses are to the grindstone in the academic treadmill, writing papers for referee journals... There are other things that are also important.”
In Carroll's version of many worlds, branches occur only when a quantum system decoheres by becoming entangled with its environment; an electron passing through both slits is still a single branch with a spread-out wave function, which is precisely why interference is observable, and a spin pointing along X is simultaneously a superposition along Z while remaining one branch.
“branches happen when quantum mechanical systems decohere, that is to say when they become entangled with their environment... The electron going through two different slits is still one branch of the wave function.”
Claims of metaphysical truths attainable only through long discipline and inexpressible except by metaphor are possible in principle for truths about oneself and how to live (though not for physics or math), but Carroll is skeptical because the heavy personal investment predisposes one to self-deception ('the easiest person to fool is yourself'), and he sees no tangible wisdom difference in long-time practitioners.
“this kind of thing where, 'Oh, I have some special knowledge, but I can't tell you what it is,' man, that just seems right for fooling yourself.”
Carroll is skeptical that Oumuamua is artificial mainly because of priors — most objects in the galaxy are natural, not artificial, so it would take strong evidence to shift his view — and because the specific solar-sail scenario is incoherent (unclear aiming, no clear artificial signal); but he agrees the possibility deserves serious investigation, and Avi Loeb shifted his credence toward taking the search for extraterrestrial life more seriously.
“I think that most objects flying through the solar system are much more likely to be natural than artificial, so I would need really good evidence to change my mind”
The shift from Trump-Pence to Biden-Harris feels different not because polarization decreased but because the political 'temperature' lowered — by analogy to gas molecules or poker table dynamics — so the discourse style changed from personal grievance-driven decisions to technical problem-solving, even while Republican commitment to oppose the Democratic agenda remains.
“the switch from Trump Pence to Biden Harris has lowered our political temperature. That doesn't mean it's lowered the polarization... But the discourse is just a fundamentally different kind of style”
The core difference between physicists and mathematicians is one of personality: mathematicians are not content to stop when they are pretty sure something is right and need to prove it, whereas a physicist like Carroll, once convinced something is pretty much right, wants to move on; physics is driven by caring about the real world while pure math puzzles disconnected from the world hold no interest for him.
“mathematicians are not content to stop when they're pretty sure something is right, they need to prove it... If I think that something is pretty much right, I wanna move on to do something else.”
Carroll's interviewing philosophy is to make guests look good and let them present their best ideas rather than to debate them or confront them with his own disagreements, because over 130+ episodes he has spoken far more than any guest, so a productive disagreement that lets him learn is the goal rather than scoring points on contested topics like the foundations of quantum mechanics.
“I don't bring people in the podcast to debate them, I certainly do not bringing people on the podcast to ride my hobby horses and confront them with my opinions about things about which we might disagree.”
Carroll can leave Caltech and pursue exactly the research he wants because podcast, book, and talk income — substantially from Patreon supporters — covers his salary, freeing him from grant and institutional obligations, though he notes a future academic job offering the right freedom plus colleagues could still be better.
“I can pay my salary through a combination of the podcast, writing books, giving talks, things like that. So I don't need to be at Caltech anymore.”
Carroll records using an Electro-Voice RE320 XLR microphone into a Sound Devices MixPre-3 mixer, records guests locally via Zencastr (which produces higher-quality audio than internet transmission), edits in Audacity, and mails guests a Blue Yeti USB microphone to keep.
“my microphone is what it's called an Electro-Voice RE320, it's a very nice voice microphone.”