Why Sean Carroll Thinks Many-Worlds Is the Cleanest Version of Quantum Mechanics

Host Lex Fridman begins with the textbook definition of quantum mechanics, which separates what is observed from what is real, and asks what that means. Carroll says his personal view is that observers and measurement will not play a fundamental role in the ultimate laws of physics. He admits it would be "infinitely cool" if observation did, but he sees no evidence for it.

His favorite alternative to the textbook picture is the many-worlds interpretation, which makes two claims. First, you are a bunch of atoms with a wavefunction obeying the Schrödinger equation like everything else. Second, when you think you are measuring an electron, you become entangled with it, so the wavefunction contains one branch where the electron was here and you saw it here, another where it was there and you saw it there, and so on.

Once those branches exist, Carroll says, they no longer interact, as if they were separate worlds. This was the idea of Hugh Everett III, a Princeton graduate student in the 1950s, who argued you do not need extra rules about looking at things and should simply trust the Schrödinger equation. Carroll jokes that Everett offered therapy as much as physics, with the cost being a whole lot of extra worlds.

Worlds appear whenever a system in superposition becomes entangled with the outside world, with or without an observer. Carroll defines the "outside world" as all the degrees of freedom you are not keeping track of, as with a bottle of water where you track volume rather than every molecule. The full theory is just a wavefunction of the universe obeying the Schrödinger equation, and the hard work is mapping that onto what we observe.

Asked how many worlds there are, Carroll says nobody knows: possibly infinite, possibly finite, but certainly a very large number. Fridman raises the idea that splitting worlds might reach some equilibrium. Carroll replies that whether Hilbert space is finite or infinite dimensional is secretly connected to gravity and cosmology, a part that physicists are still working to understand.

The link comes from the 1998 discovery that the universe's expansion is accelerating. If that continues, there is a horizon around us beyond which things recede faster than light, so only a finite number of things can happen inside it. Carroll says the guess for the dimensionality is 10 to the power of 10 to the power of 122, dwarfing the roughly 10 to the 88th particles in the universe.

If that story is right, branching must reach a limit because you eventually run out of room in Hilbert space. Roughly speaking, Carroll says, that corresponds to the universe expanding, emptying out and cooling into empty space forever.

Fridman asks whether splitting and copying differ in any physical sense. Carroll prefers to say the universe splits, because talk of duplication invites a reasonable worry about where the energy for the extra universe comes from. Thinking of one pre-existing universe dividing into two skinnier ones avoids that confusion.

He gives a geometric analogy: a vector of length one at 45 degrees can be written as the sum of two vectors along the x and y axes, each of length one over the square root of two. You have described the same arrow differently without changing anything. The wavefunction of the universe is a big vector, and branching works the same way.

Carroll is emphatic that there is "zero question" that many-worlds does not violate energy conservation, in contrast to genuinely open questions such as where probability comes from. All the effort goes into translating the unambiguous equation into plain English, and the picture of a universe with a thickness dividing into thinner pieces, with the total conserved, is a reasonable one.

Why does the idea provoke discomfort? Carroll distinguishes two kinds of simplicity in a physical theory: simplicity of the theory itself, and how easily it maps onto the world. Newtonian physics maps easily, general relativity less so, and quantum mechanics is hard. He says many-worlds is the version where that mapping is hardest, so the lack of simplicity lies in how the theory is applied, not in the theory.

He considers it perfectly legitimate to be bugged by that distance from experience. Still, he argues that skepticism should eventually give way if no alternatives work and the theory keeps succeeding. For now, he notes, people are developing alternatives that sit closer to what we observe directly.

On the field of contenders, Carroll says a debate could easily field twelve candidates but would have three frontrunners. Besides Everett's many-worlds, hidden-variable theories say the wavefunction is real but not everything, with actual particles pushed around by it, which becomes much harder with quantum field theory or gravity. Collapse theories say each particle has a tiny chance per second of spontaneously collapsing, and a macroscopic object full of entangled particles collapses together, while Roger Penrose proposes collapse when gravitational differences become too large.

Carroll summarizes the three frontrunners: many-worlds has only the Schrödinger equation, hidden variables add new variables, and collapse theories sometimes follow the Schrödinger equation and sometimes collapse. Fridman asks whether the wavefunction is fundamentally probabilistic. Carroll adds a fourth family, epistemic interpretations, which treat the wavefunction as merely a prediction tool, with different observers possibly holding different wavefunctions for the same system.

He notes that the problem with epistemic views is that they decline to say what is being predicted. By contrast, ontic interpretations treat the wavefunction as real. Carroll points to a World Science Festival debate moderated by Brian Greene, with David Albert for spontaneous collapse, Sheldon Goldstein for hidden variables and Rüdiger Schack for epistemic approaches.

Carroll finds many-worlds compelling for two reasons. It is the most austere version of quantum mechanics, and he would rather work hard on mapping than complicate the formalism. He also argues that other versions carry classical baggage such as locations in space, which becomes a barrier for quantum gravity and the emergence of space-time.

Other interpretations must be rebuilt for each new system, since you have to ask what the hidden variables are or how collapse works. Many-worlds is plug-and-play: give him the theory and he can give you its many-worlds version. Where the classical description breaks down dramatically, he says, you should start from the most quantum theory you have.