Scott Aaronson

Found him through zcash and zk-SNARKs (2-1e Everything is connected in a complex way)

Aaronson inspired memos

• Hidden-variable theories <> epistemology (fallibilism)
• interactive proofs are faster <> reading makes learning faster and especially finding two-steps aheaders easier
  • Read efficiently by using others to filter what to read. Efficiency is not about speed per se. Solving problems also the same (i.e., solve what matters).
• Local uncertainty becomes (almost) certain globally due to similar process as BPP?
  • <> past and future are other universes means we are already using other universes when we multi-process computations?revisit
    • <> self-locating belief (20260102)
• the P vs NP problem—what happens if this gets solved? (source)
  • NP would’ve been P if you can access the multiverse?
  • faster way to solve sudoku can cure cancers faster (maybe)
• Figuring the Conplexty Zoo is about epistemology in disguise, since knowledge creation is a form of computation.
• Easy to verify (NP) <> how idea exist in person’s mind (P) (kind of) <> bucket theory <> goes from private public (kind of, not necessarily)
• vs deutsch (in particular, vs MWI)
  • what exactly is happening with Shor’s algorithm???
    • Aaronson can keep QM and QC without committing to MWI
      • Hilbert space is mathematics; whether it corresponds to “real universes” is metaphysicsrevisit
        • I guess the reason one cannot say QC is performed across the multiverse is simply because that QC is performed in your universe, correct? The twist is that since Hilbert space cannot be measured (in the sense of observing what’s going on inside), and because each possible state corresponds to different state, it is somewhat like each state being in its own universe.
          • You can only access amplitudes, interference patterns, global properties
            • Similarity: NP gives you a promise that a solution exists but not the distribution or structure of all possible solutions — a bit like QM giving you amplitude structure but not access to all branches.
            • Both NP and QM rule out omniscience about all underlying possibilities
              • NP is computational limitation
              • QM is physical limitation
                • But reality doesn’t differentiate computational vs physicalrevisit
                  • You cannot get full hidden-variable distributions (QM).
                  • You cannot get full solution space distributions (NP).
                  • You cannot get full branch-by-branch outcomes (QC).
                  • You cannot simulate full ZK transcript distributions if the protocol isn’t zero-knowledge.
              • How are the following related?
                • NP’s uncheckable witnesses come from combinatorial explosion.
                • Hilbert spaces’ inaccessible amplitudes come from measurement law.
                • SZK’s indistinguishability comes from distribution symmetry.
                  • Our universe contains rich structures (witnesses, amplitudes, branches), but the laws of physics limit our ability to probe those structures fully. This is exactly what distinguishes NP from BPP, CZK from SZK, and Hilbert space from classical universes.
              • Nobody explained why such limitations exist
                • Or maybe taking MWI seriously would do it?
                  • It does—says Deutsch—because it’d explain them
    • Deutsch is not saying quantum computers perform computations across parallel universes. He is saying quantum computers are possible because the physical world has the structure that MWI describes.
      • Quantum computation is evidence that reality has the structure described by MWI
• If you can distinguish skill from luck, you must be the skillful one (and this is different from verifying vs solving NP problems)
• symbols accelerate computation
  • <> 10-2g2c1d
    • <> only accelerate what’s worth accelerating <> 2-1b2b1.1 If it’s not worth doing well, it’s not worth doing at all
• exhaustive entity → no market (because market needs redundancy) → no communication which means no computation
  • to compute, you need others?
• fungibility
  • for Deutsch this implies the reality of many-worlds
  • for Aaronson this implies irrelevance of many-worlds
• Focusing on time instead gives us space, because you cannot see every space (inspired dream where I realize how one perceives his surroundings not visually but via imagination) <> self-locating belief <> Tachyon (why we need clock)
• Behavioral test doesn’t work, maybe because of NCT (what needs to be replicated is mind, and that’s equivalent to predicting mind, and that’s probably forbidden by NCT, at least according to my understanding of Aaronson)
• Computation per se doesn’t need MWI. But evolution and epistemology need MWI. And computation needs evolution?
  • E.g., see 1-2 We are fallible
• 1-2i → culture >>> rules (>>> decisions)
  • <> faster computation <> and computation is nothing more than an aid to communication
• 1-1a4b2
  • abduction is faster than pure reason—this is why test matters—it’s about faster computation (i.e., complexity theory)
    • deutsch <> aaronson
      • <> 9-2a1.1 (execution is speed <> computability <> complexity)
• on aaronson-deutsch (or zcash-multiverse)
  • e.g., 6-3a2.1 <> 1-2g2t4 (or just search with fungib)

Democritus

• p. 76 (?) → reading a lot of math proofs doesn’t mean you’ll be good at math, just like how reading a lot of books doesn’t translate into making you smarter
• p. 110 → we don’t really need experiments because we are Turing machine <> but often times being in contact with external reality gets you to reality faster <> see from different perspectives <> seeing the front
• ch. 13 → understanding, if it’s not about mere computation, might then be about ability to create counterfactual explanations
• p. 291 → NP problem just admits: “There is an answer”
  - Black Swan presumes that one presumed some probability distribution, and because NP problem doesn’t deal with probability distribution, it has nothing to do with it

Ghost

• p. 6: the second laws of thermodynamics forbids perpetual motion machine, and Aaronson is looking for whether anything forbids complete understanding, and two (thermodynamics and quantum theory) are related
• p. 9 <> you don’t have to evoke the multiverse to argue for free will (unlike Deutsch)?
  • <> 9-4b2b1: Knowledge is unpredictable because of NCT
    • Then the multiverse per se has nothing to do with freedom (e.g., 9-4b2b2)
• p. 36: does the multiverse explain why there’s no cloning theorem?
• p. 39: assuming godlike ability to know the quantum state is similar to assuming that you can know the multiverse (the very assumption of course negates the concept of multiverse) ~ the morphology is similar (what does this mean?)
• p. 67 <> Deutsch argues even if the probabilities are known, there is freedom <> but aaronson isn’t interested in such freedom (aaronson is saying one cannot likely predict what happens even within one’s universe—i.e., you don’t have to evoke the multiverse to argue for free will)? <> what does aaronson say about MWI?
• p. 68 → aaronson argues that there are things which cannot be predicted probabilistically in our universe, and deutsch basically supplants his argument by saying that is so because there exists the multiverserevisit
  • aaronson meets deutsch

Philosopher

• p. 3 → why hasn’t it? <> sometimes the key is in the question, not the answer
  - If the Turing principle was about computability, then Aaronson takes Turing further, and thereby takes Deutsch further…revisit
• p. 28: culture as computation complexity concept
• p. 35 → maybe aaronson is taking quantum computation theory seriously than deutsch?
• p. 36 → on the multiverse: if they are decohered, they don’t interfere anymore; if they interfere, they are the same
  • <> samenessrevisit
  • <> differencerevisit
• p. 40 → time <> space <> computation
  • more specifically, what does the simple fact that one can think, which is a form of computation, imply about the nature of time and space?revisit