Comment by psakkaris | Hacker News Reader

Comment by psakkaris | Hacker News Reader

Parent Full thread

Thank you for the criticisms, It will really help with revision so I can explain the work better:

1. The work is "easy" or "trivial"

I disagree. It does not scale the way we want yet for coherent control gates but we were able to scale the terminal QFT accurately by using Graph Theory on a classical computer vs matrix mechanics. That is an alternate algorithm at the least so not very easy. For example, if someone told you go implement the terminal QFT on a classical computer using graph theory how quickly do you think you could have done it? Or even better, what if I tell you now "implement an alternate scalable terminal QFT algorithm on a classical computer without using QuDot Nets" how easy do you view that question?

2. Maybe I did not explain the IF BPP=BQP argument well

IF BPP=BQP, and that is a BIG IF, then it is very likely that a different deterministic theory could reproduce the randomness in quantum mechanics. We are not saying that quick sort isn't random, we are saying that quick sort can be replaced by a non random algorithm. Why? because there is much evidence that P=BPP which implies that randomized algorithms can have equivalent non-random algorithms. Similar to how the primality testing algorithm was shown to have a non-random equivalent. "Quantum computing from Democritus" does a nice job of explaining this.

Lastly, it could be that this graph algorithm of quantum computing will never be able to scale coherent control gates. But at the very least we came up with an alternate algorithm to do quantum computing on a classical device. Not completely worthless right? What else are you going to do on a Saturday night? :-)

Thanks again for feedback, even though it was negative

For 1: I didn't mean to imply anything about the work being easy or hard, by the way. I was just contrasting against the difficulty of proving BQP=BPP. (e.g. I might refer to Terence Tao's work on Navier-Stokes [1] as 'solving an easier version', but certainly no connotation about the difficulty of the work would be intended).

But I do think you should avoid making such large claims based on what was actually done. It comes off as naive; unaware of how much work has gone into trying to make these things work. Of course people have tried hierarchical models!

For 2: You're still confusing "I have an efficient deterministic algorithm to make predictions about X" with "X must be deterministic". The efficiency of the algorithm is totally independent of whether or not the other system is "really" random or not! Some random processes are downright trivial to make predictions about, even with really truly literally random inputs. And some are arbitrarily hard to predict, even when we replace the randomness with a cryptographic pseudo-random number generator (but don't tell you the seed).

Consider that we already have deterministic algorithms that predict quantum mechanics. The fact that the algorithms take a long time to run has little bearing on the predictions they make. Finding a more efficient algorithm for quantum won't turn it into a deterministic process. You'll still be able to pass Bell tests, and the algorithm had better predict that too or its demonstrably wrong!

1: https://terrytao.wordpress.com/2014/02/04/finite-time-blowup...

It is 12:00am. I just read your "implement an alternate scalable terminal QFT algorithm" note.

So, because of the restrictions on earlier operations, our input is a bunch of separable qubits. That helps a lot. At first I was thinking of turning phase gradients on one side into increments on the other, but there's a much easier way. Just measure each qubit after it gets Hadamarded. All the phase steps can then be conditioned classically.

Now it's 12:10 am.

3 more minutes to make an example circuit in Quirk... and done: http://algorithmicassertions.com/quirk#circuit=%7B%22cols%22...

Reply on news.ycombinator.com