AMA | December 2025
What this covers
This transcript collects listener questions posed to Sean Carroll during Mindscape's December 2025 Ask Me Anything episode. The format differs from a conventional interview: it presents the questions themselves, not Carroll's answers, so no single argument or position runs through the material. Instead, the transcript serves as a map of what conceptual terrain preoccupies the audience. Eight questioners raise problems spanning foundational physics, artificial intelligence, philosophy of mind, and metaphysics, with notable clustering around quantum mechanics and existential risk.
The heaviest concentration falls on quantum interpretation and the nature of the wave function. Several questions probe Many-Worlds interpretation, the status of ontology in physical theory, and whether mathematical models describe reality or merely predict outcomes—touching on the underdetermination of theory by evidence and the map versus territory distinction. Others examine stranger quantum scenarios: superposition of spacetime geometry near black holes, the event horizon as a possible illusion born of time dilation, and an alternative to Many-Worlds in which the wave function decoherences into a single branch upon observation. AI risk and machine consciousness form a second substantial thread, with questions about whether large language models trained on viral sequences could enable bioweapon design, how to recognize machine sentience, and whether the stakes justify taking superintelligence warnings seriously. A third line of inquiry explores emergence, the reductionist hierarchy of sciences, and whether emergent models harbor inherent limitations. Peripheral questions touch on relativistic travel, personal identity under teleportation, the Sleeping Beauty problem and self-locating probability, information persistence after death, and financial transparency in AI firms.
This is an AMA question compilation in which listeners pose physics, philosophy, and life questions to Sean Carroll; the provided transcript contains only the questions, not Carroll's answers, so no authorial argument is advanced — the central takeaway is the range of conceptual puzzles (quantum interpretation, entropy, AI risk, emergence, black holes) that the audience finds most pressing.
- Questions cluster heavily around Many-Worlds interpretation, the reality of the wave function, and quantum field theory subtleties
- Several questions probe AI risk, superintelligence, and machine consciousness
- Recurring interest in entropy, the arrow of time, emergence, and complexity
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.
Throwing a radioactive particle toward a black hole creates a branch where it decays before the horizon (with decay products escaping and carrying energy away) and a branch where it crosses intact (contributing its full energy to the black hole's mass), which would place the black hole in a superposition of different masses and therefore put spacetime curvature itself into superposition.
“Have I put a macroscopic object, the black hole, in a superposition of different masses and thus different curvatures of spacetime, therefore putting spacetime itself in a superposition?”
There is a distinction between the reality we try to describe and our modeling of that world, and taking the wave function as ontologically real may amount to confusing the mathematics (the map) for reality (the territory).
“I think there's a reality we try to describe and there's our modelling of that world. Isn't taking the wave function as real confusing the math for reality?”
The argument that we are unlikely to be in a simulation because our perceptions are 'high resolution' is flawed, because we could not know that our 21st-century experience isn't actually low-resolution by the standards of advanced simulators, since we cannot yet conceive of a higher resolution simulation than the one we might be in.
“how we would know if our current 21st century experience of the world we appear to live in ISN'T low-resolution by the advanced standards of our simulators?”
If reality is made of quantum states rather than information, then since quantum states encode information it is unclear what distinguishes 'information' from the things that are 'real', whether information is ever ontologically fundamental, and if so where the computation is happening.
“Since quantum states encode information, what exactly distinguishes "information" from the things that are "real"? Is information ever ontologically fundamental?”
A physical theory is a mathematical model plus rules linking observations to mathematical objects; since a good theory only predicts experimental results and the model itself tells us nothing about reality, the ontology one adopts (e.g., the universe is a vector in Hilbert space versus a world of objects, fields, and cats) may be a matter of personal choice while predictions remain identical.
“I can choose to believe that the universe is just a vector in a Hilbert space (whatever that would mean) or alternatively that the world around me consists of objects, fields and cats. In either case quantum mechanics continues to make the right predictions.”
An alien pilot who travels at relativistic speed but has no contact with his home world after departure would not notice any effect of relativity; lacking knowledge of relativity, he would simply assume he traveled faster than light, and all his calculations would be self-consistent from that perspective.
“Wouldn't he just assume that he traveled faster than light, and all his calculations make sense from that perspective?”
The use of large language models trained on viruses to design new bacteriophages represents a plausible mechanism by which AI could become the world-ending catastrophe some have predicted.
“I recently read an article about scientists using a LLM trained on viruses to design a new bacteriophage... This seems like a very plausible mechanism for AI to be the world ending catastrophe some people have been predicting.”
Even if one assigns only a 1% probability that the analyses of AI experts like Hinton and Bengio are correct, the extremely high stakes combined with risk aversion mean we are rationally required to take the possibility of doom from superintelligence very seriously and update our priors despite our own non-expertise.
“Even if you judge that there is only, say, 1% probability that the analysis of Hinton and Bengio is correct, it still appears that we are rationally required to take this very seriously. I wouldn't get on the plane”
If teleportation painlessly vaporizes you and constructs a perfect copy down to your last pre-vaporization thought, the question is whether identity is grounded in continuity of thought or in the chain of computations that believes itself to be you — and whether the post-teleport individual is the same person or merely a continuation of a narrative.
“do you think identity is a continuity of thought or the chain of computations that believes itself to be you?”
There is a regress of explanatory depth in which psychology is informed by biology, biology by chemistry, chemistry by physics, physics by mathematics and logic, and mathematics, logic, and physics possibly by metaphysics and ontology — raising the question of whether metaphysics and ontology are the deepest possible subjects or whether the search for a 'deepest' subject is itself pointless.
“Psychology is the most immediate raw facts of my subjective experience... But it's arguable that Biology informs Psychology. Further, Chemistry informs Biology. And Physics informs Chemistry. But Mathematics and Logic may possibly inform Physics.”
When machines achieve self-awareness it will be unethical to treat them as anything less than independent sentient beings with rights equal to our own; the crucial unsolved problem is how to tell when machines are as conscious, self-aware, and capable of free will as we are.
“when machines achieve self-awareness, it will be unethical to treat them as anything less than independent sentient beings with rights equal to our own.”
All emergent models are wrong some of the time even well within their domain of applicability — for example, modeling planetary motion by centers of mass works until the star explodes or the planet evolves people who push it away — suggesting an engineer-minded 'good-enoughness' may be an inherent limitation of emergence itself.
“model a planet's motion around its star using their centers of mass, and it will work most of the time. But after a while the star may explode, or the planet may evolve people who push it away from the star.”
Many Worlds could be modified to work in reverse — starting with many orthogonal branches in the wave function but with the wave function decohering into a single surviving branch upon observation, requiring an adjustment to the Schrödinger equation — and this collapse-like alternative is arguably as likely as Many Worlds while better according with actual observation of a single outcome.
“the wave function decoheres into a single branch upon observation or the system's entanglement with the environment... isn't this as likely as the Many Worlds theory and one that accords with actual observation?”
Because time dilation for anything approaching the event horizon relative to outside observers approaches infinity, it is unclear how anything has enough time within the universe's lifespan to actually fall in and cross the horizon relative to the rest of the universe, raising the question of whether black holes truly exist or are merely regions of very slow passing of time relative to us.
“are there actually black holes in the universe or just spaces of very slow passing of time relative to us?”
Both world-first (Halfer) and observer-first (Thirder) perspectives in the Sleeping Beauty problem may be literally true and valuable, such that our individual consciousnesses are each a distinct branch of a conscious multiverse while simultaneously coexisting in the same physical universe, motivating a foundation that presumes 'many observers' as its cornerstone.
“our individual consciousnesses are each a distinct branch of a conscious multiverse, while simultaneously coexisting in the same physical universe? Why shouldn't we build a foundation that presumes or predicts 'many observers' as its cornerstone?”
If information cannot be lost or destroyed, and our thoughts and memories are information, then when one dies that information must persist — which may imply there is something after death even without invoking god or religion.
“If information can't be lost or destroyed. Our thoughts and more so our memories are information. So when I die... What happens to THAT information?!”
Comparison is the thief of joy, yet people still struggle with imposter syndrome by judging their own accomplishments as inadequate relative to others.
“I know that comparison is the thief of joy, but regardless I find myself often struggling because I think that my own accomplishments are not good enough”
There could be additional spatial dimensions we are unaware of because they are tiny (compactified), and one might be able to devise an experiment to detect whether they exist.
“could it be that there are more spacial dimensions which we are not aware of (e.g. because they are tiny) and can we think of an experiment to find out whether they exist?”
AI firms are raising staggering amounts of capital (approaching a trillion dollars) while privately held, with zero obligation for public financial transparency; any firm capable of single-handedly triggering a global recession should be forced to follow the same financial reporting rules as a publicly traded company, yet no one in or out of government is pushing this.
“any firm that has raised sufficient capital to single handedly trigger a global recession should probably be forced to follow the same financial reporting rules as a publicly traded bicycle manufacturer, and yet I've heard exactly zero voices in or out of government pushing this angle”