Join 3,424 readers in helping fund MetaFilter (Hide)

Bitcoin-like cryptocash which is tradeable for computation
February 19, 2012 6:45 PM   Subscribe

Bitcoin-like cryptocash which is tradeable for computation
There are huge computational resources devoted to the proof of work which authenticates bitcoin transactions, and I've been thinking about ways to turn all those resources to more broadly useful computations. I've come up with a design for a bitcoin-like cryptocurrency which in which computation in the proof of work can be purchased using the currency itself. The design's described in the PDF I linked at the project URL.
Role: I'm doing the whole thing, so far
posted by Coventry (3 comments total) 1 user marked this as a favorite

I've just recently decided to start cracking the kinds of books necessary to understand this kind of thing, because I think novel applications of cryptographic technology will be a powerful force, good or bad, in my future. So, please, forgive me if these are naive questions:

1) How useful is the ability to submit these kinds of problems (refering to "embarassingly parallel" monte carlo simulations)? Obviously, if this is already the way that these problems are being handled in already in this particular field, NooShare is immediately useful. More precisely, I want to know these things: Can all arbitrary algorithms be reduced to NooShare-calculable ones? Can all of them? Is there a way to do this programatically?

2)Can this be implemented on the back of something that already has wide deployment? Could it have shared mining with bitcoin and namecoin (I presume not, because these projects both use SHA256)? Could SETI@home push this out with the next update of their client? Basically, could this bootstrap itself with the infrastructure of exisiting distributed computing schemes?

3) In the paper, you mention a possible exploit wherein malicious users could willfully submit programs that would never stop running, or run very slowly. Is this a real world threat, as you see it? Is there a way to prevent it (Again, I presume not, because if I understand correctly, this would essentially require solving the halting problem)?
posted by LiteOpera at 5:06 PM on February 21, 2012


Thanks for your interest, LiteOpera.
  1. Anything could be run in this framework in principal, the important question is what will run efficiently. Roughly speaking, the best problems for NooShare are those which can be solved by quickly experimenting with lots of memory-small solutions and looking for the best ones. At the moment, my main question is whether this style of computation would be useful/cost-effective for people. Someone started a similar market for a similar computational model in 2006, and it worries me that it didn't take off. But I think things have changed about since then. Lots of people are running massively parallel computations with limited storage and communication on GPUs, so more people will be used to thinking this way. And I think the economic structure of NooShare gives it an advantage over the earlier market. The main thing I'm thinking about at the moment is cheap ways to check these opinions. (If anyone with experience in GPU computation has an opinion, please weigh in!)
  2. No, bitcoin/namecoin/seti@Home are committed to specific calculations. The principal goal of NooShare is enable largely arbitrary calculations in the proof of work.
  3. Yes, the defense against that exploit is described in section 2.2.2. Basically, there is a default proof of work calculation which miners can fail over to if the calculation purchased in the current block starts to take too long.

posted by Coventry at 10:43 AM on February 22, 2012


Oops, "changed about" -> "changed a lot."
posted by Coventry at 10:45 AM on February 22, 2012


« Older The Linux Setup...   |   Gyms Near You!... Newer »


You are not currently logged in. Log in or create a new account to post comments.