David Chalmers in his book: : Reality+: Virtual Worlds and the Problems of Philosophy, has a discussion on information and reality. He identifies different types of information: semantic, structural, and symbolic.
Semantic information is what we colloquially think of as information, it’s the patterns that tell someone or something about reality. A map of a city is semantic information, at least for an observer who isn’t blind and can understand the language of the labels. On the other hand, a sheet of paper with random patterns isn’t semantic.
But to a physicist, the random patterns are still information. I’ve often called these kinds of patterns physical information to distinguish it from the semantic variety. I like this name because it emphasizes that all information is physical. In this view, semantic information is a subset of physical information.
Chalmers introduces the concept of structural information, which I initially thought was equivalent to physical information. But physical information, as I conceive it, is agnostic on the ultimate ontology of its patterns. Chalmers’ conception of structural information is specifically patterns of bits.
Symbolic information is structural information that encodes semantic information. In a way, it could be seen as the relation between structural and semantic information.
Chalmers’ concept of structural information is related to an ontological view he calls it-from-bit, the idea that reality, at its base level, is digital. The idea is that reality, if we go down far enough, perhaps down to the Planck scale or lower, is binary in nature. So elementary particles, as well as spacetime, are ultimately composed of bits.
(Chalmers notes that this is somewhat distinct from John Wheeler’s original idea of it-from-bit, which involves a participatory view of reality being constructed based on what yes-no questions we ask it. Digital physics keeps the yes/no part, but ditches the semi-idealist aspects.)
Chalmers admits that every bit we currently know of is actually a bit-from-it arrangement. The bits in commercial computers are transistors with their own components. Even in quantum computers, the qubits are typically the spin of particles which have a lot of other characteristics. He also admits the idea that reality is digital isn’t particularly favored in physics. It isn’t indicated in quantum field theory, or even in more speculative theories like loop quantum gravity or string theory. Although the idea has been explored by some physicists, in some cases in more of the it-from-qubit type form.
But Chalmers argues that for his purposes in considering the simulation hypothesis, it’s only relevant that it hasn’t been ruled out by fundamental theories. (This feels a bit weak to me, since lots of things haven’t been ruled out that we have little reason to think are reality.)
In the later parts of the book, Chalmers ties this in with computation in physical systems. He notes that computation isn’t just mathematical, it’s physical, requiring the right kind of causal structure, including counter-factual constraints. It isn’t enough that the system have a sequence of states, but those states need to be just one of many that could have happened had the initial conditions been different. Put another way, there had to have been a necessity to that sequence.
Chalmers sees computers as causation machines. And that causal structure, he notes, isn’t cheap. Which isn’t to say it can’t be implemented in different substrates, just that the substrate can’t be something arbitrary, like a rock. This is an argument against the idea of unlimited pancomputationalism, of a type that makes any talk of computation trivial, but not of the more limited variety that leaves the idea of a simulation viable.
Chalmers argues that this counter-factual requirement also applies to scientific theories.
Structural realism (which I’ve discussed before), is the view that what is real in reliable scientific theories about fundamental reality are the relational structures described by the mathematics of those theories. This sits on the epistemically cautious side of scientific realism, which argues that reliable theories represent reality, as opposed to anti-realism or instrumentalism, which argue that theories are only prediction mechanisms.
Anti-realists often point out that reliable scientific theories are often replaced with newer more reliable theories, sometimes with radically different views of reality. Therefore, the pessimistic induction is that we can’t count on theories representing reality. Realists point out that each successive theory gets us to a closer approximation of reality, and that it would be a miracle if a theory made reliable predictions without approximating reality in some fashion.
Structural realism seems like a compromise between these positions, albeit one that slightly favors realism. (It also sits well with my own suspicion that the distinction between reliably predicting observations and representing reality may have little if any meaning.)
A criticism of structural realism is that it seems to be saying that the mathematics itself is reality. But Chalmers argues that these are causal structures. Similar to the fact that computation must happen in a causal system, these structures must themselves have a causal aspect, one involving counter-factuals. In his view, this makes the pure mathematics view untenable. (I’m not sure this stands up to scrutiny since causality is itself arguably a relation, not that I’m in the math-is-fundamental camp myself.)
Structural realism has two sub-groups: OSR (ontic-structural realism) and ESR (epistemic structural realism). OSR makes a stronger statement, that reality just are these relational structures. ESR makes the weaker statement that the relational structures are all we can know about reality.
Chalmers muses that OSR could be viewed as a pure and uncompromising it-from-structure view. ESR, on the other hand, could be seen as an it-from-structure-from-it view, or more speculatively an it-from-structure-from-bit scenario. In other words, he sees ESR more compatible with the possibility of a digital reality, and the simulation hypothesis. Although it could be the case that OSR is true for base reality, while ESR is the case for simulated ones. Which might leave us with a more it-from-bit-from-it view.
My own understanding of OSR is it only excludes intrinsic non-relational properties. In that sense, it seems compatible with everything Chalmers describes ESR as being compatible with. But this is complex stuff and I may well be missing something.
But Chalmers’ overall point is that if these structures can be instantiated in a simulation, then for all intents in purposes, they’re as real as the structures outside of that simulation.
Okay, I think that’s enough for this post. What do you think of Chalmers’ reasoning? Are there reasons he misses to dismiss the idea of digital physics? Or anything that makes it more likely? Is merely instantiating the structures of structural realism sufficient for reality? Or is something else required?