Blog: Finding gainful employment for your computer
- 06 September, 2007 12:01
A common fantasy -- I'm sure you have had it -- is to make your computer earn real money for you automatically and autonomously. They do so much now; how hard could it be?
One approach is to sell Google space for ads on our websites, and many do just that. However the proceeds are usually not enough to retire on, leaving ample room for more extravagant fantasies. A more ambitious dream is using artificial intelligence to build an entrepreneurial program that would work just like a real businessperson, recognizing and exploiting opportunities as they come up, doing all the paperwork, and just sending you the profits, together with a 1099-DIV at tax time. This is obviously uphill.
A third approach relies on the commonplace that almost all of us buy far more computing resources -- processing, storage, and bandwidth -- than we need. We buy not just for peak use but for our fantasies of our peak use. Visionaries have been wondering for years if a market could be made over the internet for these unused resources (occasionally some of us need more resources than we have locally, even with all our overprovisioning; these would be the buyers).
In theory participation in such a market could be made entirely autonomous. The machines could run usage models to assess their needs over the next unit of time and buy or sell accordingly, setting prices with auctions, etc. You would only be aware of the process when deposits (or withdrawals) were made to your account. From time to time UPS might deliver hardware upgrades that your system had ordered, when and as it calculated that the upgrade could earn its way.
However, while systems have sprung up for distributing excess resources, these have not been organized as markets, and certainly not as markets based on real dollars. The problem is that to achieve any serious scale such markets would have to be organized on a peer-to- peer basis, and financial transaction environments not supervised by central authorities invite hanky-panky. It's a hard problem.
As a result today excess processing cycles are usually given away, the most famous example of the case being SETI.com. Current practice for distributing excess bandwidth, as in systems like Bit Torrent, is a bit more sophisticated, often relying on a simple form of barter called tit-for-tat, in which a given system A will upload preferentially to systems that have allowed A access to its archives in the past. (Programs supporting the distribution of excess storage have certainly been written but for whatever reason have not caught on, perhaps because people don't want to run the risk of having child pornography found on their disk.) Distributed sensing -- data for dollars -- is in its infancy. All in all, peer to peer transactions with real dollars seem far off.
Recently two computer scientists, David Parkes of Harvard and Johan Pouwelse of the Delft Institute of Technology in the Netherlands, brought those old dreams a bit nearer earth by pushing the BitTorrent model a step closer to true currency. Their program generalizes the tit-for-tat protocol used in bandwidth sharing programs to a notion of "tribe," where the members of a group are all linked by reciprocity accounts. That is, when system A helps out B, B sends the news of A's helpfulness to his or her "friends", who then stand willing to help A out themselves, depending on how close they were to B. "This unravels the bilateral tit- for-tat slightly, allowing a longer barter cycle," says Parkes. The security problems raised by this change are easier to handle in the context of a tribe. Research is underway to see just how long these barter cycles can be extended -- how large a tribe can be.
The name of the program executing this interesting experiment is Tribler, from "tribe". It can be downloaded at http://tv.seas.harvard.edu. You might want to take a look at it, though note that it comes preloaded with content from BitTorrent, etc., some of which is a bit raw.
While this research is not done with the interests of corporations in mind, it is likely to occur to anyone with a CIO perspective that the employees of a corporation comprise a natural tribe for the purposes of this program.