Introduction to Orion document March 28, 2007
Posted by Geordie in Applications, QC-Related Posts, Superconducting Processors, World Domination.trackback
I’ve put together a document describing the Orion system, together with some background information useful to know related to quantum computing.
Follow this link to the document.
To see it just go to the bottom of the page and log in.
Feedback as always appreciated!
Do I get to add my questions?
(Does that count as a question? Doh, that’s another one.)
What is the strength of the couplings between and on your qubits?
Do you really have an analog computer since you’d have to have infinite precision control? Can I see the knob with infinite precision, because that would be a cool knob?
Adiabatic quantum computers, it seems to me, are often formulated in an analog manner, just as the niave quantum circuit model is formulated in an analog manner. But in practice, and especially when it comes to the effect of error, the digital nature of the model will reveal itself. Similarly slide rules are considered by some to be analog computers, but there is a finite precision to which you can read those damn tick marks on the slide rule. Mostly this is just a pet peive of mine because while I know what your saying, I think calling your processor an analog computer isn’t useful. The problem, I think, is that the word analog computer is often used interchangely with the word real computer where real=real numbers.
Where does the power of classical computers come from? (I think what you really want to ask is why are quantum computers more powerful than classical computers.
)
Are the couplings you use in your adiabatic quantum computer universal?
For the algorithms you demostrated, how many times do you have to run the quantum computer to get the right answer? I.e. what is you failure probability. Have you checked how this scales as a function of the number of qubits involved in the algorithm? What is this scaling? Do you understand why it is scaling in the manner it is scaling? If you add this scaling into your projection for when your quantum comptuer will outperform a classical computer do you still get 2008?
How did you determine that quantum effects are responsible for the operation of your processor?
“ORION WILL FIRST BE AVAILABLE AS AN ONLINE SERVICE THAT
PERFORMS OPTIMIZATION COMPUTATIONS ON THE HARDWARE
AT A D-WAVE FACILITY. WE WILL CHARGE FOR USE OF THE
ONLINE SYSTEM BASED ON EITHER TIME USED OR
SUBSCRIPTIONS.”
I thought that you were going to allow free access at start. Has this changed?
Man you’d think I’m curious or something
But us academics, you know, we’re pretty useless, so feel free to ignore the questions. Oh which leads me to the last question: what benefit do you obtain for alienating the academic community? Are you that confident in your path that you can’t forsee another funding round? Are you worried that if you did have to go for another round that the amount of bad mouthing you might get from the academic community would cause problems (or are academics just such nonplayers that what they say doesn’t matter?)
Dave: Sure I would add questions if it would make the doc better. Also if the answers can be made better I would do that too!
Here’s some answers to your questions:
“What is the strength of the couplings between and on your qubits?”
Forgive my choice of units, but this gives the relative sizes: Between: from -5 to +5 milliPhi_0; biases -10 to +10 milliPhi_0; bits of precision on each currently about 5 (32 steps between max points)
“Do you really have an analog computer since you’d have to have infinite precision control?”
What I mean by analog computer is the other (better) definition of what an analog computer is:
…analog computation is quite different from the commonly used binary-digital computation. In the digital approach, you first devise an algorithm (a set of instructions for finding the solution to a problem), then execute each step of the algorithm in the manner of Boolean operation. In contrast, analog computation is concerned with no symbolic Boolean operation; instead it utilizes the properties of a physical system to perform the mathematical operations required for the solution.
I think that in this light a classical analog computer has the infinite precision issue, but an analog computer of the AQC sort has naturally digital components. Kind of like an analog digital computer.
“Where does the power of classical computers come from? (I think what you really want to ask is why are quantum computers more powerful than classical computers.
)”
Yes that is right. Also your first question is an interesting one in its own right.
“Are the couplings you use in your adiabatic quantum computer universal?”
No.
“For the algorithms you demostrated, how many times do you have to run the quantum computer to get the right answer?”
Generally the success rates are around 90% for 4-vertex MIS problems and around 85% for 6-vertex MIS problems.
“Have you checked how this scales as a function of the number of qubits involved in the algorithm?”
Yes, but for the 16-qubit system we can’t do problems big enough to draw any significant conclusions.
“What is this scaling?”
See above–too early to tell.
“Do you understand why it is scaling in the manner it is scaling?”
The main issue now is probably calibration of the machine language numbers (the coupler and qubit bias numbers), although we’re going to have to build significantly bigger systems to get enough information to chop into this issue.
“If you add this scaling into your projection for when your quantum comptuer will outperform a classical computer do you still get 2008?”
We don’t have enough information to know for sure. But it’s our target!
“How did you determine that quantum effects are responsible for the operation of your processor?”
I wouldn’t characterize them as “responsible”. They are definitely involved, in the sense that their presence is changing the behaviour of the machine vs. the case where these weren’t there. One of the characterization experiments is the macroscopic resonant tunneling (MRT) experiments discussed at the APS meeting earlier this month. There are a few others.
“I thought that you were going to allow free access at start. Has this changed?”
I think I can get you some free time if you want.
“what benefit do you obtain for alienating the academic community?”
We’re not trying to alienate the academic community. I love science and think the process by which it’s done is vital and important. However what we are doing is orthogonal in many ways to that process. Our objectives are very different than those of the academic community.
“Are you that confident in your path that you can’t forsee another funding round?”
No we definitely will need at least one more funding round.
“Are you worried that if you did have to go for another round that the amount of bad mouthing you might get from the academic community would cause problems (or are academics just such nonplayers that what they say doesn’t matter?)”
It depends on what the bad-mouthing is saying. If there were problems underlying what we’re working on, we want to know what those are worse than anybody. We don’t want to fail. So if there were something like this that arose during due diligence, it might hurt us in the short term, but in the long term it is better that those things come out early as a sanity check.
In regards to the last part of your question: Academic opinions do matter and are important in the specific area of expertise of the academic. In QC though often people have opinions outside of their area of expertise. For example, if a mathematician has an opinion on fab yields they probably will be ignored. I believe you coined the extralusory intelligence phrase–QC experts tend to be guilty of this alot.
@Dave
I liked your question about precision in the couplings, but it seems to me that with the quantum (i.e. discrete) nature of stuff that your infinite precision knob would still have ticks. Doesn’t that make it finite? Or are we talking about going to INF in the plus and minus directions? I do, however, see your point that the ticks on the Orion knob in comparison to the ticks on the “infinite” precision knob matter in terms of scaling.
To my knowledge (admittedly my knowledge is limited to your blog, scott aaronson’s blog, this one, and every damned article I can find on D-Wave) D-Wave has done loads of collaboration with universities. Does “alienating the academic community” mean taking publicly stating an opinion that quantum computing has a better chance of success in a start-up vs at a university? I don’t believe D-Wave has ever stated that current academic research is foolish or useless. I believe I have heard Geordie say publicly that D-Wave’s approach may not even be the best approach. However, they believe it is the right approach because they can scale up as quick as possible. Is a scaling limit less correct because it was found by actually reaching it than by theoretically determining it? Wouldn’t better individual qubit control retroactively benefit an approach that has sought to quantity vs quality in qubits? Again, I’ve had limited resources on which to form my conclusions, but if there have been other public statements by D-Wave stating that academics are useless could you link them. I’d like to read them.
I’m not as ‘in the loop’ as some other people around here; but I don’t really feel like D-Wave is alienating the academic community. Me and the other people in this office where curious about Orion and D-Wave. Geordie was kind enough to answer our questions in an email interview. Ok, so the interview wasn’t entirely serious, but I we found out what we wanted to know.
It might be nice to see some more scientific output from D-Wave, but we should understand that scientific output isn’t the goal of the company.
Cool, thanks for the answers!
So running the orders of magnitude in my head, the ratio of you energy couplings to your temperature is something like a factor of 100-500? So if your gap is running like one over the number of qubits squared, then from the numbers you have for the success probability it seems reasonable that the main cause of this problem is thermal excitations when the gap is at a minimum, right? So Orion II must either work at a lower temperature or increase the coupling or….?
I believe you coined the extralusory intelligence phrase–QC experts tend to be guilty of this alot.
One of the challenges of quantum computing is that it is an extremely large field, stretching from the hard core fabrication and technology needed to build the gadgets to the experimental physicists probing the limits of quantum theory to theoretical physicist studying the theory behind the gadgets to theoretical computer scientists and information theory folk studying the implications of quantum computing for computation and information processing. I don’t know of ANY person who has a grasp of the entire field (although there are some people who come very close and are amazing, like, for instance, David DiVincenzo or Seth Lloyd.) So, while I’d be skeptical of a mathematician discussing fab issues, I’d equal be skeptical of someone from the experimental side venturing into the lad of computational complexity. This isn’t to say that this is an absolute truth, there are a remarkable number of people in quantum computing who can move across the divides, perhaps more so than in any other field today. But it also means that you have to be very careful with what you say…you never know when you’ll be a victim of extraillusionary intelligence yourself! (me I’m still looking for normal intelligence out of my brain!)
So, for example, when I read things like
a statement which from the comptuer science perspective is pretty total bunk, you can imagine that the computer scientists will accuse YOU of extraillusionary intelligence. Then, when there is backlash against this from computer scientists, things like
certainly don’t help. Me, I certainly value VCs and startups a ton, but I also value academics whose rule is very different in this whole equation. So I don’t see the point of statements like the above which do alienate academics by basically talking about how much better VC funded startups are.
@Chris
The alienation comes not from what they are doing or even, in my opinion the way they are doing it (I don’t really give a rats ass about publishing their results so much as whether they are actually succeeding) but from their statements which overhype the power of quantum computers. I personally just don’t see the point in this overhyping…I think what they are doing is valid enough to survive without the hype (or in other words the hype behind what their machine might eventually realistically achieve seems sufficient to me, enough.)
As for specific statements about the academic community, the statement above (lined here) is definitely not so nice.
In the long run it doesn’t matter, of course, but I can’t understand the logic behind saying something like that.
@zevans
I don’t think scientific output is the issue. The real issue is statements overhyping the power of quantum computers. Overpromising is a dangerous path to go down, not just for D-wave, but for everyone in quantum computing, and there is a real danger of a backlash against the field from this. I want nothing but success for D-wave because I do believe deeply in quantum computation, but I guess I’m young and idealistic (okay maybe the second and not the first) and think that you can achieve this goal without pressing the scientific overhype button.
We know, you’ve been bought off
Oh man, this is just way to serious of a discussion for me in the morning. But thanks for answering my questions Geordie!
Yo Bacon, we are brainstorming a viral marketing campaign that would be right up your alley. So don’t feel left out.
So maybe D-Wave have been pressing the scientific overhype button a little too much, but they aren’t the only ones. I guess I expect - or at least hope - that the people who are actually selling the details of the system to investors are a little more realistic in private about the prospects of D-Wave style computation. Certainly, from what I have read, Geordie has been more circumspect when asked direct questions about the limitations of the system, the kind of questions that I would think that potential investors would be likely to ask of D-Wave.
In any case, the point is that D-Wave are selling; they have a responsibility to secure investment, not a responsibility to the scientific community. I guess this is complicated in a sense as some within D-Wave perhaps don’t care about the opinions of those in the academic QC community, whereas others obviously do. The problem for those that do care is how to make truckloads of cash and retain credibility at the same time - like Britney.
But, like you, I think that the work is interesting and doesn’t need hype to justify its continuation.
This has been a message brought to you by the D-Wave subliminal advertising department. buy DWAVE now this stock is GOING TO TAKE OFF!!!
What are you trying run a bump and run scheme?
TINY, 12.92, Trade Time: 4:00PM ET Change: -0.21 (-1.60%)
Didn’t work today.
BTW, mostly I’m bitter because I myself want to build the first commerical quantum computer. But, like that kid in that movie about that Bobby Fisher guy, I’m stuck in that “don’t move until you see it” loop.
Viral marketing is good just as long as it doesn’t involve litebrights and Bostonians.
Dave you too can get in on this sweet AQC action…we need a good theorist. Why don’t you come join the posse. Feel the power of the dark side of the force.
@Dave
Thanks a bunch for your response. I do understand the concern for the field of quantum computing because of the hype that D-Wave and their demo has generated (i.e. if failing to meet expectations by D-Wave is viewed as a failure of the field as a whole).
As for Herb Martin’s comments… I think I tend to brush over anything said by him about the tech of the company, because Geordie obviously has a firmer grasp of the tech goings-on at D-Wave. That being said, others will view these lesser informed comments about D-Wave as comments by D-Wave.
I agree, the tone of that 2nd quote wasn’t that great =( . I, however, can’t express an opinion on the validity of the comment as I haven’t been under the scrutiny of VC investors or government grant writers.
[...] good news for D-Wave’s adiabatic Orion processor: arXiv: quant-ph/0512170 [ps, pdf, other] : Title: Error correcting codes for [...]