What this covers
This May 2021 Ask Me Anything episode from Sean Carroll's Mindscape podcast draws on submitted questions from Patreon supporters to explore puzzles across physics and philosophy. Carroll's approach throughout is diagnostic: he argues that many apparent paradoxes or mysteries arise from conflating distinct concepts or privileging intuition over formal mathematics. The episode ranges across foundational topics—the arrow of time, quantum mechanics, determinism and free will, consciousness, artificial intelligence—while also taking on narrower technical questions and broader methodological issues about how to evaluate competing claims and interpret experimental evidence.
On the physics side, Carroll clarifies entropy and thermodynamics: the arrow of time stems from low past entropy, not from quantum interpretation; the cosmological constant's non-zero value reflects our incomplete understanding rather than any flaw in nature. He distinguishes hidden variables from locality, showing that John Stewart Bell's theorem rules out local hidden variables but not non-local variants. On quantum mechanics, he treats the uncertainty principle as a consequence of Fourier analysis rather than measurement disturbance, and defends the coherence of eternalism—though carefully, stressing that simultaneity across moments is incoherent. For consciousness, he resists binary framings, arguing instead that sentience comprises distinct capacities acquired separately across evolution. On artificial intelligence, he separates genuine safety concerns from what he sees as the premature and conceptually confused alignment problem.
Throughout, Carroll emphasizes Bayesian reasoning as the proper framework for managing uncertainty and updating beliefs, and he models intellectual honesty by acknowledging legitimate counterarguments before rebutting them—a criterion he applies when assessing expert credibility.
Carroll argues that many apparent puzzles in physics and philosophy dissolve once you separate distinct concepts (time vs. arrow of time, locality vs. hidden variables, subjective vs. objective entropy) and trust the mathematics over intuition, while treating free will as a non-problem resolved by compatibilism.
- The arrow of time comes from low past entropy, not from quantum interpretation or from time itself
- Bell's theorem rules out local hidden variables, not hidden variables per se
- Compatibilism vs incompatibilism does not depend on whether physics is deterministic, only that there are laws
Advanced civilizations would not require faster-than-light travel because solving aging is tractable biology while FTL would violate physics; they could simply live millennia or sleep through journeys.
- Even if UFO sightings were legitimate, advanced civilizations would not need faster-than-light travel or universe-hopping because solving aging is merely a tractable biology problem, whereas FTL travel would require violating the laws of physics; one could simply live for millennia or sleep through long journeys.
“Solving aging is just a biology problem. Faster than light travel is a how do you violate the laws of physics problem.”
In many-worlds with continuous variables, you ask for the probability of being in a set of branches meeting a criterion, not which exact branch you occupy, using calculus to handle infinite branching.
- In many-worlds with continuous variables (like the time a nucleus decays), the apparent infinity of branches is handled the same way calculus handles continua: you do not ask which exact world you are in, but the probability of being in the set of worlds where the nucleus decayed between time T and T plus delta T.
“you should think about the set of worlds in which the nucleus decayed between time T and T plus delta T, that's a sensible thing to think about”
Time moves at one second per second in its proper reference frame; Big Bang cosmology has a well-defined rest frame (galaxies and microwave background), so inflation timescales are not altered by density.
- It is not true that time flows differently or at a different rate near the Big Bang just because matter is dense; time moves at one second per second in its proper reference frame, and Big Bang cosmology has a well-defined rest frame (the rest frame of galaxies and the microwave background) in which inflation timescales are measured.
“Just the fact that things are dense doesn't mean that time moves any differently, time moves at one second per second.”
Democratic world government fails because people lack shared interests and voluntary buy-in across borders; world dictatorship is undesirable, making global problems hard to solve through such institutions.
- Democratic governments only work when people under them feel they have common interests and buy in voluntarily; since people in the year 2021 do not feel that way across national boundaries, an all-encompassing democratic world government would not work, and a world dictatorship is undesirable, making global problems hard to solve through such institutions.
“democracy is an experiment that requires buy-in, you can't impose it”
Theoretical physicists do not write proofs but derive equation implications, then convince colleagues through talks, collaborations, and peer review in a bottom-up process with no central authority.
- Theoretical physicists generally do not do proofs; instead they write equations on paper and derive implications of those equations (predicting signals an experiment should see), then convince colleagues socially through talks, collaborations, and peer review, in a bottom-up process with no central authority.
“we don't do proofs in theoretical physics, usually... mostly you sort of derive implications of your equations”
New particle physics is unlikely to have practical effects beyond knowledge because producing particles requires giant accelerators unsuitable for devices; seek physics for discovery, not technological application.
- New particle physics is very unlikely to have either good or bad practical effects for humanity beyond knowledge, because producing new particles requires giant accelerators that cannot be shrunk into usable devices; therefore one should look to particle physics for discovery, which is an unalloyed good, not for technological applications, contrary to worries that new physics is more likely to harm than help.
“In order to make new particle physics you need a giant particle accelerator. These are not the things that you can shrink down into the size of a phone”
Psi-epistemic interpretations face an unresolved puzzle: if the wave function is merely predictive, not real, it cannot explain how probability distributions produce interference patterns in the double-slit experiment.
- Psi-epistemic interpretations hold that the wave function is just a tool for predicting measurement outcomes rather than a real thing, but this creates an unresolved puzzle in the double-slit experiment: probability distributions do not interfere with themselves, yet interference patterns appear, and Carroll admits he does not know how psi-epistemic accounts explain this.
“Probability distributions don't interfere with themselves, 'cause they're not real... so what's going on? I don't know what is going on”
Bell's theorem does not eliminate hidden variables; it proves quantum correlations exceed local predictions, but non-local hidden variable theories like Bohmian mechanics remain fully compatible with observation.
- Bell's theorem does not eliminate hidden variables; it proves that quantum mechanical correlations between measurements on separated systems can exceed what any purely local theory could reproduce, where 'local' means both that things exist only at local points in space and that signals cannot travel faster than light. Non-local hidden variable theories like Bohmian mechanics remain fully compatible with observation.
“what Bell's theorem rules out is the possibility that Einstein did want... a hidden variable theory that was local”
Physics-based work on political and multi-agent social systems exists abundantly but lacks the label 'political physics,' creating the false impression no one works on it.
- While econophysics is an established named field with textbooks, there is a large body of physics-based work on political and multi-agent social systems that simply is not labeled as 'political physics,' so it appears no one works on it when in fact many do, just under different names.
“there are no political physics books, and so you might get the impression that no one works on it, but lots of people work on it, they just don't call it that”
Without expertise, judge credibility by track record and how speakers address counterarguments; those dismissing critics as pawns are less credible than those presenting strong opposing views before rebutting them.
- When you lack expertise to evaluate a claim, you must judge the credibility of the person making it; beyond track record and recognition, the most informative criterion is how they relate their claim to opposing arguments: someone who dismisses critics as establishment pawns is less credible, while someone who fairly presents the strongest counterarguments and then rebuts them demonstrates appreciation of how they might be wrong and is more likely to be right.
“A person who demonstrates that they appreciate why they might be wrong is much more likely, in my mind, to be right at the end of the day.”
Utilitarian summing of experiences into total utility is incoherent because there is no sensible way to add and compare experiences across different beings, making the enterprise itself unsound.
- The utilitarian project of summing experiences into a single 'total utility' number to maximize is misguided and leads to the repugnant conclusion (that it is always better to add more people no matter how miserable); Carroll argues there is no coherent way to add up and compare experiences across beings (such as a cow's and a human's), so the whole enterprise is not merely difficult but not a sensible thing to attempt.
“I don't think there's a number called utility that you can get by adding things up of different experiences at different moments of time in any way.”
Gödel's incompleteness theorem has no clear bearing on physics because physics does not prove theorems from axioms but constructs models of the world.
- Gödel's incompleteness theorem (that a formal axiomatic system cannot prove its own consistency, implying unprovable true statements) has no clear bearing on physics, because physics does not work by starting from axioms and proving theorems; physics constructs models that model the world, so there is no identified precise limitation Gödel's theorem imposes on physics.
“That's not how physics works, physics doesn't go around starting with axioms and proving things”
Avi Loeb is not a charlatan regarding ʻOumuamua; Carroll invites podcast guests only if he thinks they are worth listening to in good faith, regardless of his agreement with them.
- Avi Loeb is not a charlatan selling snake oil regarding ʻOumuamua; more broadly, Carroll never invites guests onto his podcast merely to debunk them, so any guest can be taken as someone he thinks is worth listening to in good faith, whether or not he agrees with them.
“I will never have someone on the podcast just to debunk them.”
AI safety is a genuine problem arising long before true AI, because we already have complex systems we do not fully understand; the real challenge is managing risks in systems beyond human comprehension.
- AI safety is a genuine and important problem that arises long before true artificial intelligence, because we already have computers, networks, and deep-learning algorithms too complex for any human to fully grasp, doing things for reasons no human knows; the real challenge is quantifying and protecting against risks in complex systems we do not fully understand.
“it's just complex systems, risk analysis... it does things and no human being knows exactly why it's doing what it is doing”
The correct approach to random-transporter scenarios is Bayesian updating: assign priors to hypotheses and update via Bayes' rule with observations; credence should never reach 0 or 1.
- In a scenario where you are told a transporter is random but you always appear on pad 1, the correct approach is Bayesian: you assign prior credences to hypotheses (random, always pad 1, etc.) and update via Bayes' formula with each observation; no matter how high your prior in randomness, enough pad-1 observations shift you toward 'always pad 1,' but your credence should never reach 0 or 1 since future data could always change your mind.
“it should never go to 0 to 1, because you always have more data that you can collect in the future that might change your mind”
The best evidence for dark matter comes from the cosmic microwave background, not galaxy rotation curves; alternative explanations must account for CMB observations, not just galactic rotation.
- The best evidence for dark matter comes from the cosmic microwave background, not from galaxy rotation curves; therefore any proposed alternative explanation that modifies gravity or invokes external forces must explain the cosmic microwave background, not just galactic rotation.
“the best evidence for dark matter comes from the cosmic microwave background, not from galaxies”
The arrow of time is entirely explained by low early entropy; no quantum interpretation explains why the early universe had low entropy, which must be assumed independently.
- The arrow of time is explained entirely by the fact that entropy was low near the Big Bang; no quantum mechanical interpretation explains why the early universe had low entropy. Even with intrinsically time-asymmetric collapse models, a universe started in thermal equilibrium would stay high entropy, so low initial entropy must be assumed independently of quantum interpretation.
“The arrow of time of our universe is explained by the fact that entropy was low in the past, entropy was low near the Bing Bang”
Discoveries most likely to disappoint Carroll represent retreats from intelligibility, such as anthropic explanations or objective collapse, which are scientifically acceptable but ugly if true.
- The scientific discoveries that would most disappoint Carroll are those representing a retreat from the intelligibility of the universe; he illustrates with the anthropic explanation of the cosmological constant (which makes a once-and-for-all predictive formula less likely) and with dynamical objective collapse of wave functions, both of which he would find ugly though scientifically acceptable if true.
“the kinds of scientific discoveries that would disappoint me would be ones that somehow represented a retreat from intelligibility in the universe”
Donald Trump was the worst US president due to disrespect for democratic institutions, corruption, incompetence, and refusal to accept electoral losses, causing incalculable damage to democracy.
- Donald Trump was the worst US president not because of his policies but because of his lack of respect for democratic institutions, corruption, incompetence, cult of personality, and refusal to accept losing the election, which has done incalculable damage to democracy; democracy cannot function when substantial numbers of people will never accept electoral outcomes against them.
“it's not because of his policies... it is because of the complete lack of respect for what you call democracy and the institutions of government”
Thinking like a physicist comprises characteristic paradigms, questions, toy systems, and analytical methods that surface unconsciously across unrelated contexts beyond explicit physics.
- There is something real called thinking like a physicist, comprising paradigms, characteristic questions, toy systems and spherical cows, and ways of analyzing complex systems, which may run as unconscious sub-routines that surface across unrelated contexts even when one is not consciously doing physics.
“there is something called thinking like a physicist, and it's not a single thing, there are ways that physicists think”
To change loved ones' belief in astrology or spiritualism, show by example that you can abandon such beliefs while leading a happy life, since people adopt them for non-rational reasons.
- When trying to change a loved one's belief in things like astrology or New Age spiritualism, the effective strategy is often not to deploy your strongest rational debating arguments, but to show by example that one can abandon those beliefs and still lead a happy, fulfilling life, since people often adopt such beliefs for reasons other than pure rationality and can be reasoned out of them for non-rational reasons too.
“the strategy is not to bring out your biggest debating guns... the strategy is to by example show them that they can get rid of those beliefs and still lead a happy, fulfilling life”
Increasing entropy explains why processes do not occur equally forward and backward; human purpose depends on treating the past as fixed and future as influenceable, both consequences of entropy.
- Increasing entropy is at the core of why anything that does not happen equally well forward and backward in time occurs; in particular, human purpose and meaning depend on treating the past as fixed and the future as influenceable, which is itself a consequence of entropy increasing, since we have causal leverage over the future but not the past.
“It's because entropy is increasing that we think of the past as being fixed and the future as being influenceable.”
Response to evil depends on whether the person was controlled or truly could have acted otherwise rationally; feeling sorry suits the former, legitimate anger the latter.
- How to feel about people who commit evil acts depends on the best description of them: if they were addicted, controlled, or mentally incapable, the appropriate response is to feel sorry for them, but if the best description (given what we know about them, not the microscopic state of the universe) is that they could have done otherwise rationally, then anger at them is perfectly legitimate.
“If you think that the best description we have of this person is as someone who could have done otherwise given what we know about them... then you can be angry at them.”
AdS/CFT is a powerful working model but does not describe our universe (not anti-de Sitter space) and struggles with small-scale bulk physics; focusing on it may not optimally advance real-world holography.
- AdS/CFT (Maldacena's correspondence between a gravity theory in anti-de Sitter bulk and a conformal field theory on its boundary) is a powerful working model of holography but does not describe our universe (we do not live in anti-de Sitter space) and struggles to describe small-scale bulk physics; Carroll's tentative view is that holography in the real world will be subtle enough that over-focusing on AdS/CFT is not helping as much as hoped.
“my feeling is that AdS/CFT has misled us a little bit”
A single photon cannot decay into two photons because conservation of momentum and angular momentum forbid it; if possible, we would have observed it long ago.
- A single photon cannot spontaneously decay into two lower-energy photons because conservation of momentum (a directional quantity) leaves zero phase space for the process, and conservation of angular momentum forbids one spin-1 particle decaying into two spin-1 particles moving in the same direction; if it could happen we would have observed it long ago.
“once you take all of the conservation laws into account, no, one photon cannot spontaneously decay into two”
A missing middle in physics communication exists between equation-free accounts and professional textbooks: explaining what equations mean and their conceptual framework without technical computation.
- There is a missing middle ground in physics communication between equation-free popular accounts and full professional textbooks: you can explain what equations mean (what a derivative or integral is) without teaching how to manipulate them, conveying the conceptual framework without the technical computation, which is what Carroll attempted in his Biggest Ideas in the Universe videos.
“there's a middle ground where I can just tell you what the equations mean, without telling you how to manipulate them”
If the universe has finite degrees of freedom, quantum field theory would be a low-energy emergent approximation like describing molecules as continuous fluid.
- If the universe has a finite number of degrees of freedom (a finite-dimensional Hilbert space), it cannot be exactly described by quantum field theory, which assigns a field value at every point in space; rather, QFT would be a low-energy emergent approximation, analogous to describing finitely many air molecules as a continuous fluid.
“If the universe has a finite number of degrees of freedom, then it is not described by a quantum field theory. It might be approximately described by a quantum field theory.”
Using physical determinism to alleviate regret is self-deception because at the human level we do not know our own micro-state, so it offers no genuine psychological help.
- Using the thought that one's actions are ultimately the result of physical laws to alleviate regret is a cop-out akin to self-deception, because the fact that actions follow from physical laws is irrelevant at the human level where we do not know our own micro-state, so it offers no genuine psychological help.
“I think that is a cop-out, that's just like self-deception, because the fact that... it just doesn't matter that the actions are ultimately the result of physical laws.”
A black hole is the maximum entropy for a region of given size; Hawking radiation does not violate entropy because emitted photons' total entropy slightly exceeds the black hole's.
- A black hole is not a maximum entropy object absolutely, but the maximum entropy you can fit into a region of a given size; therefore Hawking radiation does not decrease total entropy, because the emitted photons spread over an enormous distance and their total entropy is slightly higher than the black hole's, with no contradiction since that entropy is no longer confined to the region.
“A black hole is the maximum entropy that you can fit into a region of prescribed size.”
Consciousness comprises degrees of sentience (humans more than cats, cats more than worms) with different aspects acquired evolutionarily, not a single-valued spectrum or on-off switch.
- Consciousness is neither an on-off switch nor a single-valued spectrum; there are degrees of sentience (humans more than cats, cats more than earthworms, etc.), but consciousness comprises different aspects acquired along the evolutionary tree, such as the ability to imagine the future, which some organisms have and others lack entirely rather than possessing partially.
“I don't think it's either on or off, nor is it uni-valued.”
Quantum crossing symmetry implies any particle or force affecting human matter could be produced in colliders; finding nothing relevant implies everyday life is fully understood, establishing the core-theory claim.
- Quantum field theory's crossing symmetry implies that any particle or field able to affect the matter we are made of could inevitably be produced by colliding that same matter together; since particle accelerators do exactly this and have found nothing relevant, there can be no undiscovered force or particle relevant to everyday human-scale life unless quantum field theory is dramatically wrong, which is the basis of the core-theory claim that everyday life is fully understood.
“If you have a field or a particle that can affect the fields and particles of which we are made, then it is inevitable that you can make that particle or make that field just by colliding together the particles and fields of which we are made.”
String theory's appeal is not mere beauty but beautiful explanation of known real features; it has not derived our universe, so it may remain a mathematics resource despite not describing our world.
- It is misleading to present string theory's selling point as mere beauty; what makes it compelling is that it provides a beautiful explanation for features we know are real (spacetime, gauge theories, gravity, curvature). However string theory has not derived our specific 3+1 dimensional world with SU(3)xSU(2)xU(1) forces, so it remains possible string theory ends up done in math departments while still being a fruitful source of physics ideas like black hole information and gauge-gravity dualities.
“It's not just beautiful. Who cares if something is just beautiful but has no connection to the world we observe?”
Infinite space does not guarantee every possible matter arrangement occurs; an infinite space could repeat one pattern forever, so actual outcomes depend on how matter is distributed.
- An infinitely large space does not guarantee that every possible arrangement of matter occurs somewhere; an infinite space could repeat the same pattern forever (like an infinite checkerboard never producing a green square), so whether all configurations appear depends on how matter is actually arranged, specifically whether it is infinitely fluctuating.
“if space is simply infinitely big, that doesn't guarantee that everything happens somewhere”
Simple physical phenomena like swirls and sine waves appear ubiquitously because they are toy models of basic behavior; seeing them in everyday life does not indicate deep connections between systems.
- Simple, paradigmatic physical phenomena like swirls and sine waves appear ubiquitously in wildly unrelated systems precisely because they are toy models of basic behavior, so noticing such a pattern in everyday life (coffee spirals resembling galaxies, jogging speed resembling a quantum oscillator) does not indicate any deep or fundamental connection between those systems.
“these are going to appear all over the place it in contexts that really don't have any deep, fundamental, true connection with each other at all”
In a closed universe, net charge, angular momentum, and energy must be zero topologically because field lines cannot escape to infinity; in open universes these need not be zero.
- In a closed universe where space is a closed manifold (sphere or torus), the net charge, angular momentum, and even energy must be zero as a topological fact, because field lines cannot escape to infinity and must close back on themselves; but we do not know whether our universe is closed, and if it is open, net charge or angular momentum are unproblematic.
“in a closed universe... the net charge and angular momentum, and even the net energy will be zero, that's just a topological fact”
In a universe with positive cosmological constant, high entropy is empty de Sitter space, not hot dense states; the cosmic no-hair theorem and second law establish that high-entropy early states remain empty.
- Because gravity couples spacetime curvature to matter, a truly high-entropy state in a universe with positive cosmological constant is empty de Sitter space, not a hot dense expanding state; Carroll and Chatwin-Davies proved an equivalence between the cosmic no-hair theorem and the second law, so a high-entropy early universe would simply remain empty space and never collapse into stars.
“what you mean by high entropy in a universe with a positive cosmological constant is empty space, de Sitter space”
The most likely Fermi paradox solution is that intelligent space-faring civilizations are simply absent, due to extremely small probability of life or rare later hurdles like intelligence.
- The most likely solution to the Fermi paradox is that intelligent space-faring civilizations simply are not out there, which could be because the probability of life arising is extremely small, or because some later hurdle (multicellular life, eukaryotes, intelligence, or technological capability like opposable thumbs) is rare, any of which would make us effectively alone.
“the most likely solution is that the intelligence space-faring civilizations just aren't there for some reason or another”
A cannonball fired straight up lands at the same spot because velocity is relative and there is no absolute frame of reference, as Galileo used to argue the Earth could rotate undetected.
- A cannon ball fired straight up from a surface moving with the Earth lands back at the same spot, the same reasoning Galileo used: a ball dropped from a ship's mast falls straight down whether the ship is stationary or moving at constant velocity, because there is no absolute frame of reference and velocity is only relative, which Galileo used to argue the Earth could be rotating unnoticed.
“if the ship is moving at a constant velocity, you will also expect the ball fall straight down, because the ship and the ball are moving with each other and there is no absolute frame of reference”
Newton's laws can be non-deterministic in carefully constructed cases like Norton's dome, but such situations have measure zero among initial conditions, so determinism holds practically.
- Newton's laws can be non-deterministic in carefully constructed cases like Norton's dome, but such situations form a set of measure zero among all possible initial conditions, so for almost all initial conditions in the precise mathematical sense everything is perfectly deterministic and this is not a practical worry.
“It is a set of measures zero of situations you can find yourself in where there's ambiguity in what Newton's laws say.”
Sounding confident need not mean certainty; it comes from quantifying uncertainty well, having clear credences and reasons, so even 'I don't know' can be delivered confidently.
- Sounding confident need not mean being certain; it can come from being comfortable quantifying your own uncertainty, having a good handle on the relative credences you assign to different beliefs and being clear about your reasons, so that even saying 'I don't know' is delivered without feeling bad and thus comes across as confident.
“maybe there is a way to sound confident just by being comfortable about quantifying your uncertainty”
Rather than letting ultra-wealthy individuals decide space travel, tax them heavily and collectively spend that money, since no one becomes a multi-billionaire without an enabling system benefiting all.
- Rather than letting a handful of ultra-wealthy individuals decide how space travel happens, it would be better to tax them heavily so they remain merely super-rich, and then collectively decide how to spend that money, because no one becomes a multi-billionaire without a system set up to enable it, so the system's participants should benefit.
“no one becomes a multi-billionaire without help”
There is no accepted model for large-scale cosmic magnetic fields; generation requires new physics beyond the standard model, as magnetohydrodynamics is an intractable non-linear problem.
- There is no accepted model for the origin of large-scale cosmic magnetic fields; while moving electric charges can generate them via Maxwell's equations and galaxies may act as batteries amplifying seed fields, the magnetohydrodynamics is an intractable non-linear mess, and generating primordial seed fields in the early universe likely requires new physics beyond the standard model.
“I think that right now the problem is just too hard and there's no consensus.”
Free will compatibility with determinism depends only on whether laws exist, not whether they are deterministic; the distinction is whether agent-level descriptions are compatible with particles obeying laws.
- Whether you are a compatibilist or incompatibilist about free will has nothing to do with whether the laws of physics are deterministic; what matters is only that there are laws. Calling oneself a 'hard determinist' is a mistake because one would remain anti-free-will under indeterministic laws like spontaneous collapse theories; the real distinction is whether describing humans as choice-making agents is compatible with their being made of law-obeying particles.
“whether you're a compatibilist or an in compatibilist has nothing at all to do with whether the laws of physics are deterministic”
Brian Greene's proposal to switch off gravity in the early universe merely smuggles in a special initial condition rather than solving the low-entropy problem, and misinterprets the Higgs mechanism.
- Brian Greene's proposal that gravity was switched off in the early universe (letting matter equilibrate to a black-body high-entropy state before gravity turns on) does not robustly solve the low-entropy problem because it is not robust: it merely smuggles in a very special initial condition, which is exposed by reversing the problem (a collapsing version would not smooth out), and it has nothing to do with the Higgs mechanism since gravity depends on Newton's constant not particle mass.
“They cheat, in other words, by putting the universe into a very, very special initial condition, and you can show that by just reversing the problem.”
Eternalism correctly states 'all moments are equally real' not 'all moments are equally real now,' since simultaneous existence of different moments contradicts itself; this refines Augustine's memory-based idea.
- An eternalist holds that every moment of time is equally real, but the correct formulation is 'all moments of time are equally real' rather than the tempting but contradictory 'all moments are equally real now' or co-existing simultaneously, since giving different moments existence at the same time is a contradiction; this refines St. Augustine's idea that the past is present memory.
“What is very tempting to say, but is wrong, is all moments of time are equally real now or co-existing”
The observable universe is almost entirely matter, not antimatter, because matter-antimatter annihilation at meeting regions would produce gamma rays far exceeding observed backgrounds.
- We can be confident the observable universe is almost all matter rather than antimatter because the universe is not as empty as it appears: matter and antimatter regions would constantly meet across interstellar and intergalactic media and annihilate into gamma rays, and predicted gamma-ray backgrounds for a half-and-half universe far exceed observations, especially given the denser early universe.
“the matter and antimatter would constantly be hitting each other and annihilating, giving off gamma rays”
Laplace's demon would need to be as complex as the universe itself to model it, so it could never exist; treating it as 'a dude who knows stuff' is deeply misleading.
- Laplace's demon is a thought experiment of a vast intelligence knowing every particle's position and velocity plus all laws, to which past and future are equally clear, but it crucially does not exist and could never exist; to model the universe it would need to be as complex as the universe itself, so taking it literally as 'a dude who knows a lot of stuff' is deeply misleading.
“It's crucially important, by the way, Laplace's demon doesn't exist. It's a thought experiment”
AI safety is important before true AI, but the alignment problem is premature and wrong-headed because 'values' may not apply to AI, and genuinely intelligent systems might modify values unpredictably.
- AI safety (managing risks of complex systems we do not fully understand) is a genuinely important problem even before true AI, but the AI alignment problem of aligning an AI's values to human values is wrong-headed and premature, because it assumes 'values' is even a sensible concept for AI and that AI thinking resembles human thinking; a genuinely intelligent AI might modify its own values unpredictably, and there are likely countless non-human ways to be intelligent.
“It seems highly anthropomorphic to me to think that we can talk about the values of an AI in more or less the same way that we can talk about the values of human beings”
There is no universal list of physics subjects everyone should master; broad education helps on average but outcomes are hard to predict, so learn what you need and what interests you.
- There is no universal list of physics subjects everyone should master; while a broad education makes a better physicist on average, it is hard to predict which skills people will use, and some thrive being narrow while others should be broad in directions (biology, information theory, computer science) different from what their elders prefer, so one should learn what one needs and what tickles one's curiosity.
“I don't have a list of things that everyone should know, and I think that you should know the things you need to know to get ahead and tickle your curiosity”
Closed timelike curve solutions do not support eternalism as strongly as many-worlds branching because CTC solutions in hypothetical universes (Tipler cylinders, Gödel space) are not realized in ours.
- The existence of closed timelike curve solutions to Einstein's well-tested equations does not give high credence to eternalism in the way many-worlds branching does, because branching is predicted to actually happen in our universe whenever quantum systems entangle, whereas CTC solutions that are not our universe (Tipler cylinders, Gödel universe, wormholes) provide no argument against presentism as long as they are not realized in our actual spacetime.
“The fact that there's an equation that has other solutions that are not like our universe doesn't matter if it's not our universe”
Quantum uncertainty follows from Fourier analysis like classical waves, but is distinctly quantum because position and momentum are observables yielding specific answers in quantum mechanics.
- The quantum uncertainty principle follows from Fourier analysis (a wave localized in position is spread out in momentum and vice versa) just like classical waves, but its physical force is distinctively quantum because in quantum mechanics both position and momentum are observables yielding specific answers, whereas for a classical water wave you simply observe the wave itself.
“it's that relationship between the fact that you cannot localize both in position and momentum and the fact that position, momentum are both things you can observe that gives the uncertainty principle its oomph”
Space and time are not quantized; in non-relativistic quantum mechanics position is continuous and time is smooth; Planck time may limit measurement precision but does not make time discrete.
- As far as we know, neither space nor time is quantized; in non-relativistic quantum mechanics position is a continuum and the Schrödinger equation treats time as a smooth continuous variable. The Planck time is not a quantum of time as commonly claimed; it may represent a fundamental measurement precision limit but does not make time discrete.
“People will say the Planck time... is somehow a quantum of time. That's just false.”
In deterministic physics, information is conserved, so the entire past is quantitatively contained in the present, giving rigor to Augustine's memory idea; which interpretation is correct remains open.
- To the extent the laws of physics are deterministic, information is conserved from moment to moment, so in a precise sense the entire past is contained in the present moment, which gives quantitative rigor to St. Augustine's claim that the past is present memory; whether all of today's specific information was implicit in the early universe depends on which quantum interpretation (deterministic hidden variables, stochastic collapse, or many-worlds) is correct.
“The entire information is conserved from moment to moment to the extent the laws of physics are deterministic, so in some sense, the past is contained in the present moment.”
Entropy is both subjective and objective by different definitions: Gibbs entropy is subjective, von Neumann entanglement entropy is objective, and Boltzmann entropy is objective given subjective coarse-graining.
- Whether the entropy of a physical system is subjective or objective is 'yes' because the word has multiple definitions: the Gibbs/ignorance entropy from a probability distribution over microstates is fully subjective; the von Neumann entanglement entropy of a quantum subsystem is fully objective; and the Boltzmann entropy is objective given a coarse-graining but the choice of coarse-graining is subjective.
“the word entropy has different definitions, there are different ideas to which we attach the word entropy”
Describing humans as choice-making agents is justified because it is the best theory at the human level and everyone's working theory; even free-will deniers must live as if people make choices.
- Describing human beings as agents who make choices is justified because it is the best theory we have of people at the human level of description, and it is in fact everyone's working theory; even those who deny free will cannot actually live or talk as if people do not make choices, illustrated by the Alice and Bob navigation example where attributing a choice is the only sane description.
“The reason why compatibilists think that it's sensible to talk about human beings as agents that make choices is because that's the best theory we have of people, and it literally is everyone's theory.”
Describing someone as a rational agent becomes less useful under compulsion, addiction, or brain damage; how to treat people with impaired rational agency is the genuinely interesting practical question.
- It becomes less and less useful to describe a person as a rational agent making choices when they are under compulsion, addiction, or brain damage, so the practical question of how to treat people who suffer impairments to rational agency is the genuinely interesting discussion, in contrast to the boring labeling dispute over the term 'free will.'
“it becomes less and less useful to describe that person as an agent making rational choices”
Advanced civilizations would more likely send durable space probes than electromagnetic signals, since probes can wait in solar systems for billions of years rather than requiring exact-time reception.
- It is not necessarily true that we would detect extraterrestrials via electromagnetic radiation before encountering their matter; electromagnetic signals zip by and require the target to be listening at exactly the right time, whereas a truly advanced civilization wanting to contact others would more sensibly send space probes that can wait in a solar system for billions of years, integrating over time in a way radio cannot, and without wasting enormous energy.
“If you really want to sort of spread through the galaxy and talk to everyone... you should send space probes, you should send matter.”
In many-worlds branching visualized with time flowing upward, one time variable in the Schrödinger equation brings all branches into existence; it is multiple spacetimes, not multiple times.
- When visualizing many-worlds branching with time flowing upward, it is the same single time variable clicking off in the Schrödinger equation that brings all branches into existence; it is multiple spacetimes, not multiple times, and the only misleading part is imagining the drawn space between branches actually exists.
“It is the same time, it's not multiple times, it's multiple spacetimes”
The argument 'you're sometimes wrong, therefore doubt this belief now' is logically invalid; one can be generally sometimes wrong yet highly confident in particular cases via Bayesian reasoning.
- It is easy to say one should be critical of one's own beliefs but hard to know which beliefs to doubt; the argument 'you admit you're sometimes wrong, therefore you should doubt this belief now' is logically invalid, since one can be sometimes wrong in general yet highly confident in a particular case. The aspiration is to be a good Bayesian assigning credences proportional to the reasons for them.
“it's easy to say you should be critical of your own beliefs, it's harder to know which of your beliefs are the ones that might turn out to be wrong”
Brains model environment probabilistically despite atoms not doing so because brains are information-processing collections with memory and incomplete representations, unlike individual atoms.
- It is not paradoxical that brains model their environment in terms of probability distributions while individual atoms do not, because atoms have no internal memory or information-processing registers, whereas brains are collections of many atoms arranged to represent the outside world; since those representations are necessarily incomplete, modeling the environment probabilistically is the sensible response to permanent ignorance of the universe's micro-state.
“because those images are incomplete, because we have incomplete information, it makes perfect sense to me that we model our environment as probability distributions”
Calling the universe preposterous reflects mismatch between theoretical expectations and experimental findings, naming ourselves not the universe, reminding us understanding remains incomplete.
- Calling the universe preposterous (as in the discovery of a non-zero cosmological constant) is not insulting the universe, which by definition cannot be unnatural; it is naming ourselves, reflecting the mismatch between our theoretical expectations (e.g. naturalness arguments that the cosmological constant should be zero) and what experiments reveal, reminding us we are not done understanding it.
“calling the universe preposterous is not calling the universe names, it's calling us names”
The standard Boltzmann brain scenario relies on misinterpreting quantum mechanics; the true vacuum is static and does not dynamically fluctuate, so Boltzmann brains never arise.
- The standard scenario where Boltzmann brains fluctuate into existence from an eternally-lasting vacuum relies on a misinterpretation of quantum mechanics: the true vacuum state is static and does not dynamically fluctuate when unobserved, so it never produces Boltzmann brains, regardless of the infinite time available.
“It's just not true that empty space fluctuates in the vacuum, even if it's at a non-zero temperature.”