Here are two good practice problems for testing your personal bogometer:
It's not P=NP but two similar problems that attract a lot of quackery.
It's not P=NP but two similar problems that attract a lot of quackery.
In any case, I demand a public apology from you, here in the public square, where you committed your cowardly wrongdoing. We'll see what you're made of in how you handle that. If you have enough conscience and consciousness to apologize for your behavior, then we can move to you explaining your specific criticism of my work, which is now incumbent on you, is it not?..since I've called out your unwarranted and thoughtless attack. This will be a lot of mental work for you, since, thus far, you demonstrate no knowledge of Sparsey, or any of the relevant computational principles. But again, that will take you a long time. So, I recommend you apologize first, rather than keep your audience waiting.
What exactly is it you want me to apologize for? Have I said anything here that is factually incorrect?
> You even begin your comment with a disclaimer to the effect of "This is really not relevant to this thread"
No, I ended my comment with the following statement:
"It's not P=NP but two similar problems that attract a lot of quackery."
And I stand by that.
> you demonstrate no knowledge of Sparsey
I don't claim to have any. My opinion of you and your work is based entirely on the fact that you claimed to have achieved quantum supremacy on a classical computer, an extraordinary claim which you failed to back up with any evidence. When I suggested how you could provide that evidence (implement Shor's algorithm) you balked. That is the sum total of the evidence on which my opinion of you and your work is based. It's quite possible that Sparsey has merit, I don't know. What I do know is that whatever merit it may have, it isn't what you told me it was.
(The whole situation is very peculiar because in our private correspondence you really emphasized the quantum stuff, but your web site doesn't mention it at all. This leaves me nonplussed.)
> you disparage my work in the public square, and in a cowardly fashion, behind my back
If something is done in public it is, by definition, not behind anyone's back, so I think you should apologize to me for accusing me of behaving in a cowardly fashion. If you can demonstrate that something I said was wrong, I will absolutely apologize for it. But if you want me to change my opinion you will have to do with with evidence and persuasive argument, not insults and demands for apologies.
But before this spins too wildly out of control, let's start by publicly clearing up the factual question at the heart of our disagreement: do you stand by the claim you made to me in private (but which I cannot find any record of you having made in public) of having achieved quantum supremacy on a classical computer?
You introduced the term "quantum supremacy" to our discussion. I don't usually think in such hype-y / marketing terminology, but I assented because I take it to be synonymous with there being a classical realization of quantum superposition. There is. Sparsey is an example. But you haven't taken the time, again by your own admission, to study or understand that.
For the (apparently hundreds, based on hits on my website) of others watching this thread, my short 2012 paper, "Quantum Computation via Sparse Distributed Representation", clearly describes how the probabilities of multiple states can simultaneously, physically be active, in superposition if those states are represented as sparse distributed representations (SDRs). And, contrary Ron Garret's claim, my 2017 paper, "A Radically New Theory of how the Brain Represents and Computes with Probabilities", provides empirical evidence that the probabilities of all states (hypotheses) represented (stored) in an SDR coding field are updated (in a way that is semantically consistent with the learned dynamics of the input space) with a number of steps that remains constant as the number of stored states increases. That capability is minimally necessary for any quantum algorithm. I don't know of any other model that has shown this ability.
I think the mental block that has existed thus far in quantum computing (QC), and in quantum theory (QT) more generally, is that theorists have been thinking in terms of the probability amplitudes of states as being represented in localist fashion A localist representation is one in which any single entity is represented by a single representational unit and that unit represents only that entity. Any localist simulation of QC on a classical machine will require exponential memory. But, an SDR coding space has a storage capacity of combinatorial order. This opens the possibility that an exponential order number of states can be represented in the polynomial number of units comprising the SDR coding field.
I'm sorry, I didn't realize I was under any obligation to apprise you. You certainly never asked me to, and if there's some kind of universal social obligation that applies here that's news to me. (BTW, did it even occur to you that the reason I did this is not because I'm a coward or a drunkard, but simply because I might be clueless?)
> the harm you inflicted would have been done
What harm could I possibly inflict? I have no influence in the quantum computing community. I've never published anything in that field, and I've been retired from research in general for nearly 20 years. No one who matters has any reason to pay attention to what I say. The only way my critique could possibly harm you is if it has merit. (Does it?)
> What if I went around in public forums and said, "Hey there's this crackpot out there, Ron Garret, check him out on your bogometer"?
You mean like when Lubos Motl called me a "category 5 loon"?
https://motls.blogspot.com/2012/10/evading-quantum-mechanics...
I was actually flattered that someone of Lubos's stature would take notice of my work and put forth the effort to criticize it. I think it said more about Lubos than it did about me that he chose to lob an ad-hominem at me, but whatever, it's a win-win for me: either Lubos is right, and I learn something, or he's wrong, and I get to be smug about having taught him something. :-) It's certainly not worth getting upset over.
> You introduced the term "quantum supremacy" to our discussion.
That's true, but I introduced it in the context of a question: are you claiming that you have achieved quantum supremacy in a classical computer? To which you answered yes.
> I don't usually think in such hype-y / marketing terminology,
Quantum supremacy is not "hype-y / marketing terminology", it's a technical term, whose definition is common knowledge in the quantum computing community:
https://en.wikipedia.org/wiki/Quantum_supremacy
I assumed you knew that, but apparently I was wrong. (In retrospect, I guess the fact that you had to look Shor's algorithm up on Wikipedia should have been a clue.)
So... now that you do know it, I am going to ask you again: are you claiming to have achieved quantum supremacy on a classical computer?
From your comments its clear you like the attention regardless of the implications. The math has been stated,the resources to understand the algorithm are all there and not particularly hard to understand if you take the time. If you have questions on implementing the algorithm on your own you can ask. Otherwise there will eventually be public code available for people to see themselves and you can just wait for it. Have a nice night.
When Gerard originally approached me, he gave me what was essentially answer to that question that I interpreted as claiming that he had achieved quantum supremacy on a classical computer. When I asked him to confirm this, he did. But the problem is, it turns out that Gerard didn't know what quantum supremacy was. He thought it was "hype-y / marketing terminology". It isn't. It's a technical term with a precise meaning [2].
So this whole mess is, AFAICT, just a misunderstanding.
So I think your claims of quackery are a bit unfounded, but I'd still like to know your thoughts. If there are people out there who think it is impossible or lacks credibility then we need to address that. I hope it's not just the website look xD. He has been saying he needs to change the website because it's old but he hasn't done so yet.
Now, I happen to believe that it is quackery because Gerard in private correspondence claimed that Sparsey achieves quantum supremacy on classical computers. That is an extraordinary claim but he did not provide any evidence to back it up, and when I suggested a simple demonstration he could do to back it up he demurred.
It's possible that Sparsey does something interesting, but it's (almost certainly) not what Gerard told me. That's why I think that evaluating it makes a good exercise.
http://sparsey.com/SDR_and_QC.html
You may have not directly called it quackery but you suggested and alluded to it requiring a bogosity meter and that quacks are attracted to it. It's pretty clear what you meant, to be honest.
Sparsey in general though is a separate subject really. What it is doing is capturing all the features it has ever seen and building temporal and spatial relations between those features and then storing those as a code that represents an input in a particular spatial and temporal context. So if I have the words CAT and BAT the A in CAT and the A in BAT will have different codes because of the letter that came before it, likewise with T because A's code is different for both C and B therefore so is T's in both separate cases. So I think that is where the whole quantum part comes into his ideas. Every sample in a sequence has many possible "realities" or codes that can represent it and by representing it in a certain way you are also representing an entire sequence that came before it and a branching point of possible future representations. I have near zero experience with quantum mechanics so I won't comment on how closely it is or isn't to that, because I have no clue.
More importantly though, even if it doesn't represent an alternative to mathematically tit for tat exactness for quantum computing, what it can do is reduce computation times dramatically. The idea is that for an equation you could give a sufficiently sized sparsey model a number of inputs. Sparsey naturally represents those inputs as some representation and you can assign the answer to that representation, and you do that with all or as many inputs as possible, each getting its own unique representation. So basically what you are doing is making a giant lookup table of answers. So when you see a set of inputs you know the answer via code lookup rather than calculating it. Sparsey is a bit memory heavy so you would only really want to do that with large hard problems. But that's where the speed comes in, you know instantly the entire state, its past, and its likely conclusion (because it probabilistic), or in fact all of its possible conclusions.
It's similar to Q Learning in some ways, in the sense that you are storing many many states you have encountered. Except sparsey doesn't store the states brute force like qlearning. Similar states, spatially, temporally or both, are stored with more code overlap, dissimilar ones have less overlap.
But that's the gist of it. He has some results posted. We've talked about getting some sort of public repository up and running for people to play around with and maybe a community forum or something, but that's one of his concerns lately, letting people see the results for themselves rather than taking his word for it. Which, I have one project I need to finish right now, but I will be working on that full time sometime in May.
That is quite possible.
> It's pretty clear what you meant, to be honest.
What I meant was exactly what I said: evaluating Sparsey makes a good exercise. I believe it is bogus, but I could be wrong. I am basing my judgement on very sparse (no pun intended) data.
> I'll let him comment on what you guys were talking about
Here are the relevant quotes from our correspondence:
Gerard: I just watched your 2011 "Quantum Conspiracy.." talk and thought you might be interested in an idea I've had concerning quantum theory and how information is represented. The core idea is briefly described in my 2012 paper. "Quantum Computation via Sparse Distributed Representation". I have to related Medium articles as well, "Quantum Computing in the Offing" and "Not Copenhagen, not Everett, a New Interpretation of Quantum Reality". The essential idea is this. To my knowledge, in all quantum theory (QT) and quantum computing (QC) models published to date, prob. amplitudes (PAs) are represented in localist fashion. That is, each PA is represented in its own physically unique memory location (e.g., 64-bits, 32 for real, 32 for imag.) that is physically disjoint from every other PA. In principal, the number of PAs needed to describe a system is of exponential order, so a localist representation requires exponential memory. BUT, if instead, PAs are represented as sparse distributed representations (SDRs), i.e., small subsets of binary units chosen from a much larger field, then they can all exist (be physically stored) in physical (and classical superposition): each PA is just a different subset of the units and the subsets can overlap. Thus, the exponential number of PAs can be represented in sub-exponential (polynomial) memory, Basically, we're leveraging a third fundamental resource, "code space", which is formally orthogonal to the other two, time and (physical) space (i.e., amount of memory). That is, the code space of an SDR coding field is of combinatorial (exponential) order.
Me: Thanks for bringing this intriguing work to my attention. You say "the exponential number of PAs can be represented in sub-exponential (polynomial) memory”. This makes it sound as if you claiming that SDRs allow you to achieve quantum supremacy in a classical machine. Am I reading you correctly? Because if that’s what you’re claiming, then you are almost certainly wrong. Since you don’t seem like a quack, it seems more likely that I’m just misinterpreting what you’re saying, so I thought I should get that cleared up before going any further.
Gerard: Actually, you are interpreting correctly and thank you for being open to the possibility that I'm not a quack :) [Lots of details about Sparsey elided.]
Me: OK, but in the interests of full disclosure you should know that my Bayesian prior on this [not being a quack] is still not very high :-) So let me ask you another “top-level filter” type question: can Sparsey run Schor’s algorithm? If so, have you done it?
Gerard: No I've never tried to run Shor's algorithm. [Lots more details elided.] What d'ya think?
Me: I think you should try to run Shor’s algorithm, because if you succeed that will be a slam-dunk proof that your claims are correct.
Gerard: Well, I don't really understand Shor's algorithm (from wikipedia). I'd have to study it. I never really have so far.
I never responded to that because at that point I was pretty convinced he was a quack. It's possible I'm wrong about that. But I'm very confident that he doesn't understand quantum supremacy, and that his claims about it are wrong.
The code space he is talking about is utilizing time. So say I have 10 bits, and one of them is on, then I can represent 10 things. Now if my system also takes into account the code of the last time step I can represent 100 different things. 10 things in 10 different temporal contexts each, is 100 unique spatio-temporal states, and that goes on for however many layers you make. Hence exponential. Now I don't know what quantum supremacy is, but you brought it up it seems, he didn't. In general I know he thinks specialized hardware is not needed, and he has stated to me he thinks SDR or sparsey based systems tailored to do quantum computing problems could be a reality. So I think he might have thought you meant quantum supremacy in that sense. As he has told me. But I won't comment too much on that because he'll be responding tomorrow most likely. It's past midnight where he is at.
But that's what I know Sparsey does, whether or not its better than quantum computing, I couldn't say. I've tried to understand what he is getting at but I just don't know enough quantum mechanics to come to any aha moments. What I do know is that the state representation in sparsey is massive and can represent many many high order concepts from low order concepts/features and find correlation between concepts both spatially and temporally. The math is very simple for that and not really hard to wrap your head around. You should start with the 2010 paper and 2014 paper of his if you are actually going to do the mental exercise you suggested. I don't know how you can claim someone is a quack if you don't even know the basic principles of the idea.
As for his theory on the quantum part of it, I know he says its a developing theory of his and he just hasn't had the time to work on the idea like he has wanted, and he has always been 100% with me on that. Which is fair enough considering he has been working on Artificial General Intelligence and in that category alone there are so many things we haven't been able to try yet. Which is one of the reasons we are in the works of making this a profitable company rather than a research company. Getting more minds on the ideas. But I would wager that if you did more than 3 responses to him being a bit more clear on what you thought was wrong with the theory he probably could have given his thoughts to you about those specific concepts/critiques. But you really didn't say anything, at least in the responses I can see. But again I have practically no insight to your guys' full conversation so I'll let him clarify tomorrow.
This will be my last response tonight, I have to be in a meeting in an hour with a colleague overseas then I'm crashing but I might respond tomorrow and Rinkus probably will as well.
https://en.wikipedia.org/wiki/Quantum_supremacy
> whether or not its better than quantum computing, I couldn't say
It doesn't have to be better, merely as good as would already be a major revolution in both math and physics.
> I would wager that if you did more than 3 responses to him being a bit more clear on what you thought was wrong with the theory
This was my last message to Rod:
> I think you should try to run Shor’s algorithm, because if you succeed that will be a slam-dunk proof that your claims are correct. You will win a Fields medal, a Nobel prize in Physics, and become fabulously wealthy beyond the wildest dreams of Croesus. I think it’s much more likely that the attempt will reveal some flaw in your scheme, but either way, it seems like a worthwhile endeavor. Worst case, you will learn something.
I really don't know how I could have been any clearer.
This was his response:
> Well, I don't really understand Shor's algorithm (from wikipedia). I'd have to study it. I never really have so far. But, I'm confident that my conjecture that QC reduces to the two properties I mentioned in last email is true.
In other words: he doesn't understand one of the most basic and fundamental results in quantum computing, but he's confident that he has made a discovery that will revolutionize the field. If it quacks like a quack...
That's true, I don't, and I've never said otherwise. I do, however, understand information theory, quantum computation and quantum supremacy. So I don't know if you're a quack or not. What I do know is that your claim of having achieved quantum supremacy on a classical computer is almost certainly false. I don't need to understand the details of your algorithm to know this, I only have to understand what "quantum supremacy" means and why it matters.
I also know that I have looked at your site and read some of your papers and I can't make heads nor tails of them. It's possible that this is because I'm an idiot and I'm simply not capable of grasping the brilliant idea embodied in Sparsey, or that I just haven't put enough effort into it. Or it's possible that this is because it's all bullshit. I honestly don't know.
But I do know one more thing, and that is that you're getting awfully defensive about all this. Why would you do that if you were confident in the merits of your position? What threat could a lazy idiot like me possibly pose in that situation?
That's a decidedly non answer from a non expert and provides no insight to the person being critiqued. That would be like if Alcubierre came to you with some ideas for his warp drive and you said, well if it can travel faster than the speed of light then you've cracked it, but I think its more likely that you will fail. All of that being said without any knowledge or attempt to run the mental exercise you purport to be fond of. It's still a theory and the math is what I said it was, its clear in the papers, its clear in the classification experiments, its clear in the raw math of it.
What's also clear from your comments is that you won't and can't actually say anything about the algorithm because you don't know anything about it. From my experience, people like you who can only see things and explain things from a high level and never actually critique the problem head on, don't actually know what they are talking about, thinking Theranos level. If it weren't the case, you'd understand that a quick glance, which I already did when you mentioned it, at wikipedia isn't going to make me understand anything if I don't have the fundamentals under my belt. Which, believe me if it was possible, I'd be glued to this computer on wikipedia becoming an expert on everything as fast as I could read.
I've tried to be a bit civil with this but it's clear you don't care and you'll probably never care. If it trolls like a troll...
As for the look and feel of the website, which in reality has no merit on whether the research is legitimate or not, I have joked that he just needs to market himself better and he has laughingly agreed. But you can hit him or I up if the research is to hard to understand from a glance. We are trying to practice how to pitch it to people better so it's easier to understand. But once every thing clicks, its actually a really simple concept, it's just that its a highly recurrent system. I had to program it out piece by piece before things started making sense.