Wednesday, July 3, 2019

Aaronson's provability is hard to grasp

I mocked Scott Aaronson's quantum computer random number generator, and he replies:
3. The entire point of my recent work, on certified randomness generation (see for example here or here), is that sampling random bits with a NISQ-era device could have a practical application. That application is … I hope you’re sitting down for this … sampling random bits! And then, more importantly and nontrivially, proving to a faraway skeptic that the bits really were randomly generated. ...

As I explicitly said in the post, the whole point of my scheme is to prove to a faraway skeptic — one who doesn’t trust your hardware — that the bits you generated are really random. If you don’t have that requirement, then generating random bits is obviously trivial with existing technology. If you do have the requirement, on the other hand, then you’ll have to do something interesting — and as far as I know, as long as it’s rooted in physics, it will either involve Bell inequality violation or quantum computation.

The weird thing is, I’ve given hour-long talks where I’ve hammered home the above idea at least 20 times (“the entire point of this scheme is to prove the bits are random to a faraway skeptic…”), and then gotten questions afterward that showed that people completely missed it anyway (“why not just use my local hardware RNG? isn’t that random enough?”). Is there something about the requirement of provability that’s particularly hard to grasp??
It is funny to see Scott complain about being misunderstood. It is a common theme on his blog. His motto on the top of his blog is just to clarify one of those misunderstandings.

And whenever he expresses a political opinion, he always has to issue a number of clarifications.

To answer his question, mathematical provability is not hard to grasp, but randomness is impossible to prove. At best he is proving something relative to the random oracle model, some assumptions about the hardware, some assumptions about entanglement, etc.

I am sure his audience keeps waiting for him to claim something useful. Instead he only claims that the useful application is to refute the QC skeptics!

He really hates to admit that QC might just be a house of cards, so he has to pretent that it is all the fault of the QC skeptics that it is a house of cards.

This is a bit like saying that Enron and Theranos were completely valid companies because no one had proved that they were frauds. Until they did.

Sorry, but we QC skeptics are not going to be refuted by some hokey random number generator that is a trillion times worse than the ones in common use.

3 comments:

  1. Great points!

    A great thing to do so as to pin down Scott Aaronson is to ask a few simple questions such as:

    What is the Aaronson Theorem, Conjecture, Principle, or Postulate? What is the Aaronson Algorithm, Subroutine, or Language? What does the Aaronson Processor look like, and how can he one build one? What is the Aaronson Proof?

    ReplyDelete
    Replies
    1. The question is why so much investment goes into fraud. Elizabeth Holmes was giving away the secret the entire time. She was a fan of Melville (ever read The Confidence-Man?) and said she was reading MOBY-DICK! Furthermore, she would play Sympathy for the Devil on the PA system.

      Delete
  2. Started writing my reply here, but it grew too big, so posted an entry at my blog: https://ajitjadhav.wordpress.com/2019/07/04/do-you-really-need-a-qc-in-order-to-have-a-really-unpredictable-stream-of-bits/

    Happy Independence Day, you all!

    --Ajit

    ReplyDelete