Comments on: A great question http://dwave.wordpress.com/2007/02/01/a-great-question/ Wed, 04 Nov 2009 04:15:07 +0000 http://wordpress.com/ hourly 1 By: pupkarik http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-19524 pupkarik Sat, 03 May 2008 06:52:07 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-19524 xdvdsfvd fdgsd fdsaiuwa dfgdsgfs xdvdsfvd fdgsd
fdsaiuwa dfgdsgfs

]]> By: 11 questions for Geordie D. Wave « we don’t need no “sticking” room 408 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1449 11 questions for Geordie D. Wave « we don’t need no “sticking” room 408 Tue, 06 Mar 2007 02:22:46 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1449 [...] something so complicated. I actually talk a little bit about this in one of my blog posts (see here) although that focuses more on the big company vs. start up [...] [...] something so complicated. I actually talk a little bit about this in one of my blog posts (see here) although that focuses more on the big company vs. start up [...]

]]> By: Matteo Martini http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1438 Matteo Martini Sun, 04 Mar 2007 12:07:36 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1438 Richard George wrote: " and a quantum computer is never going to compete for that application " Can not see why.. " If I were someone at IBM or HP looking to make an extra billion of sales in five or ten years time, I’d be asking myself if I wanted to spend that $50M today on researching Quantum computing, or to use that money to try to win some of the existing market share off my competitors my improving my existing products. " I think you are missing few points: 1) $50 million today may well be worth few " extra billions " of sales in 5 or 10 years time; 2) Moore` s Law is in danger of stopping in 3-4 years, at the 22nm node, as nobody, at the moment, has a clear idea of how to proceed to an hypothetical 15nm node. Current silicon tech is probably not stretchable until 15nm, and no clear alternative has appeared so far..; 3) quantum computing does not seem 10 years away, as D-Wave is working on a 1kqbit QC in 1 1/2 year timeframe; 4) if $50M means $50M per year, D-Wave seems to have built a QC with less money than that Richard George wrote:
” and a quantum computer is never going to compete for that application ”

Can not see why..

” If I were someone at IBM or HP looking to make an extra billion of sales in five or ten years time, I’d be asking myself if I wanted to spend that $50M today on researching Quantum computing, or to use that money to try to win some of the existing market share off my competitors my improving my existing products. ”

I think you are missing few points:
1) $50 million today may well be worth few ” extra billions ” of sales in 5 or 10 years time;
2) Moore` s Law is in danger of stopping in 3-4 years, at the 22nm node, as nobody, at the moment, has a clear idea of how to proceed to an hypothetical 15nm node. Current silicon tech is probably not stretchable until 15nm, and no clear alternative has appeared so far..;
3) quantum computing does not seem 10 years away, as D-Wave is working on a 1kqbit QC in 1 1/2 year timeframe;
4) if $50M means $50M per year, D-Wave seems to have built a QC with less money than that

]]> By: Richard George http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1374 Richard George Wed, 21 Feb 2007 14:31:39 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1374 I think that IBM and HP make most of their revenue from selling machines to do really simple things with almost linear complexity (we'll say N.log(N) is about the worst case allowable)... but with very high reliability and well thought-out back-up and disaster recovery plans. eg. automating the accounting at world banks, keeping inventories for the military such that they can process 1000's of transactions per second 365/24 even if there is an earthquake or a fire at one of the record centres. This is where most of the market for large computers is today, and a quantum computer is never going to compete for that application. If I were someone at IBM or HP looking to make an extra billion of sales in five or ten years time, I'd be asking myself if I wanted to spend that $50M today on researching Quantum computing, or to use that money to try to win some of the existing market share off my competitors my improving my existing products. I think that IBM and HP make most of their revenue from selling machines to do really simple things with almost linear complexity (we’ll say N.log(N) is about the worst case allowable)… but with very high reliability and well thought-out back-up and disaster recovery plans.

eg. automating the accounting at world banks, keeping inventories for the military such that they can process 1000’s of transactions per second 365/24 even if there is an earthquake or a fire at one of the record centres. This is where most of the market for large computers is today, and a quantum computer is never going to compete for that application.

If I were someone at IBM or HP looking to make an extra billion of sales in five or ten years time, I’d be asking myself if I wanted to spend that $50M today on researching Quantum computing, or to use that money to try to win some of the existing market share off my competitors my improving my existing products.

]]> By: Mag.e http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1364 Mag.e Mon, 19 Feb 2007 17:00:49 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1364 I just ran over this patent - it seems IBM IS investing in QM and it's exactly the D-wave technology they are interested in: http://www.google.com/patents?vid=USPAT6495854&id=luYKAAAAEBAJ I just ran over this patent – it seems IBM IS investing in QM and it’s exactly the D-wave technology they are interested in:
http://www.google.com/patents?vid=USPAT6495854&id=luYKAAAAEBAJ

]]> By: Anonymous Coward http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1340 Anonymous Coward Fri, 16 Feb 2007 14:28:07 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1340 Thank you for your article. Your clarifications reminded of the following remark (paraphrased): "if we ask the customers what they want, they'd ask for a faster horse (instead of a car)". I don't know who to attribute this to. Thank you for your article. Your clarifications reminded of the following remark (paraphrased): “if we ask the customers what they want, they’d ask for a faster horse (instead of a car)”. I don’t know who to attribute this to.

]]> By: Dmitry http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1338 Dmitry Fri, 16 Feb 2007 08:20:40 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1338 Also I agree with Matteo that improving solution performance of NPC problems is definitely not a niche market. Many of graph problems are NPC. As a ph.d student who works on natural language processing, I encounter such problems at every point, since for now many language processing algorithms from parsing to semantic processing work on relatively small subgraphs. To improve semantic learning and processing approximations are good but not good enough. Another domain which is filled with NPC problems is logical inference. Essentially many useful theorems could be proved if we could use such "small" quadratic speedup. For now, logical inference slowly crawls from pure research to practical applications, since it is not easily scalable. If they succeed to find QC algorithms which are quadratically speed up some types of logical inference, there should be a huge marked for this coming from database processing, web search and renewed expert systems. Cant wait 2 years to see if you indeed succeed to scale it up :) Also I agree with Matteo that improving solution performance of NPC problems is definitely not a niche market. Many of graph problems are NPC.
As a ph.d student who works on natural language processing, I encounter such problems at every point, since for now many language processing algorithms from parsing to semantic processing work on relatively small subgraphs. To improve semantic learning and processing approximations are good but not good enough.

Another domain which is filled with NPC problems is logical inference. Essentially many useful theorems could be proved if we could use such “small” quadratic speedup. For now, logical inference slowly crawls from pure research to practical applications, since it is not easily scalable. If they succeed to find QC algorithms which are quadratically speed up some types of logical inference, there should be a huge marked for this coming from database processing, web search and renewed expert systems.

Cant wait 2 years to see if you indeed succeed to scale it up :)

]]> By: Dmitry http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1335 Dmitry Fri, 16 Feb 2007 04:44:53 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1335 > All QC’s could do is solve the problem a bit faster assuming that > QC’s ran at the same clock speed Maybe quadratic speedup is not great for pure theoreticans but it is not a BIT faster, it allows solving problems PRINCIPIALLY unsolvable by any linear-time computers. Consider 2^35=34*10^9 - this one easily enumeraged by today's computer. Now consider 2^70=12*10^20. Even if you multiply current clock speed by 100 and run it on 1000 processors it will not help. And the difference between 35 and 70 states in NPC problem is extremly significant for many real NPC graph problems. > All QC’s could do is solve the problem a bit faster assuming that
> QC’s ran at the same clock speed

Maybe quadratic speedup is not great for pure theoreticans but it is not a BIT faster, it allows solving problems PRINCIPIALLY unsolvable by any linear-time computers.
Consider 2^35=34*10^9 – this one easily enumeraged by today’s computer.
Now consider 2^70=12*10^20.
Even if you multiply current clock speed by 100 and run it on 1000 processors it will not help.
And the difference between 35 and 70 states in NPC problem is extremly significant for many real NPC graph problems.

]]> By: Jeremy http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1333 Jeremy Thu, 15 Feb 2007 23:13:06 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1333 Your Celera example isn't a very appropriate one. Celera produced a draft Whole Genome Shotgun of the human genome. There wasn't any confusion at the time about how long it would take to sequence the genome. The public effort's goals and timelines were related to producing an accurate and complete genome sequence. Celera's goal was to produce some amount of coverage over most of the genome. There were few technological or scientific challenges that need to be overcome to produce a draft genome sequence (even at the time) because by the nature of the product, you ignore the hard or confusing stuff. For a more current example you can look at the mouse sequencing project which produced a WGS and assembly just over 4 years ago. Where is the complete mouse sequence? It is just about here now. The WGS approach (Celera) only changes the tyoe of collection to be broad covergae across the genome, rather than deep coverage of a specific portion of the genome in the BAC by BAC approach. Any let me tell you, if you want a complete genome after you've spent the money producing a WGS draft, it is 10-50x the work of producing a complete sequence from a BAC by BAC approach in the first place. This doesn't mean that WGS isn't applicable to a wide range of projects and because of the lower percieved cost it is used for many projects today, it just means that WGS is another tool that has its appropriate uses, and the human genome sequence really shouldn't have been one of them. Your Celera example isn’t a very appropriate one. Celera produced a draft Whole Genome Shotgun of the human genome. There wasn’t any confusion at the time about how long it would take to sequence the genome. The public effort’s goals and timelines were related to producing an accurate and complete genome sequence. Celera’s goal was to produce some amount of coverage over most of the genome. There were few technological or scientific challenges that need to be overcome to produce a draft genome sequence (even at the time) because by the nature of the product, you ignore the hard or confusing stuff. For a more current example you can look at the mouse sequencing project which produced a WGS and assembly just over 4 years ago. Where is the complete mouse sequence? It is just about here now. The WGS approach (Celera) only changes the tyoe of collection to be broad covergae across the genome, rather than deep coverage of a specific portion of the genome in the BAC by BAC approach. Any let me tell you, if you want a complete genome after you’ve spent the money producing a WGS draft, it is 10-50x the work of producing a complete sequence from a BAC by BAC approach in the first place. This doesn’t mean that WGS isn’t applicable to a wide range of projects and because of the lower percieved cost it is used for many projects today, it just means that WGS is another tool that has its appropriate uses, and the human genome sequence really shouldn’t have been one of them.

]]> By: ValleyProofs » Quantum Computing Demo at the Computer History Museum http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1328 ValleyProofs » Quantum Computing Demo at the Computer History Museum Thu, 15 Feb 2007 08:26:13 +0000 http://dwave.wordpress.com/2007/02/01/a-great-question/#comment-1328 [...] Here’s also a link to dwave. So why is not IBM or Intel doing this? Good question read here for a take on [...] [...] Here’s also a link to dwave. So why is not IBM or Intel doing this? Good question read here for a take on [...]

]]>