Harnessing excess human cycles.
So how long will it be until we are served with banner ads entreating us to attend the ‘can’t-miss’, ‘must-attend’ event for the human computing industry? ‘Hgrid ‘08′ in (…of course, where else could it be held?) …Las Vegas!
In an excellent Google Talk presentation entitled “Human Computation”, Luis Von Ahn (a computer science professor at Carnegie Mellon) explains how by turning tasks that are easy for humans to process (but hard for computers) into games, he has been able to harness thousands and thousands of hours of ‘human computing cycles.’
One of the games that they have developed is called peekaboom, and having tried it out myself I am somewhat embarrassed to say that it is strangely addictive. Try it, I dare you. (I hate how even when you know about behavioural conditioning and cognitive candy, it is still damn near impossible to resist.) Now the folks in Vegas and elsewhere figured out long ago how much people enjoy a good game, so it doesn’t take a genius to figure out this is how to make a Mechanical Turk really rock & roll.
Much of the theme of this blog is predicated on the fact that far to much of the financial services business still relies on (inefficient) human cycles and that there are great opportunities for those that embrace ‘digital markets’, however I suspect there are great opportunities also to intelligently harness idle or under-optimized human cycles for many (complimentary) tasks that machines just can’t get right. (At least not yet.) There is nothing (interesting) that immediately springs to mind as an example, but I’ll definitely be mulling this over in the months ahead.
July 27th, 2007 at 3:24 am
How about rating prospective borrowers? prosper.com is an example. What I wonder about on that site are the people who borrow with the intention of plowing the money back into higher-risk, higher-interest loans on the same site. Are they making money?
Now imagine something like prosper.com, only speeded up, not to the speed of slots, but the speed of Keno anyway, with each microloan split up into tranches. The safer tranches would be held by funds or infrequent players, and the high-risk ones would be held by people betting minute to minute on “will Joe make his payment this week?”
July 30th, 2007 at 8:45 am
Interesting. I wonder if ‘crowd-rating’ would be more accurate than algorithmic rating. One of the biggest problems with credit scoring is ‘garbage-in/garbage-out’ - the state of the art (in algorithmic scoring) is actually very good, however the issue usually lies in incomplete and false input data (incomes, etc.)
Your question as to whether prosper (or zopa) users partake in credit arbitrage is interesting. I don’t know…I wouldn’t suspect their is much as the returns would unlikely to be great: even if you could net 300-400bp on the turn, the volumes needed (for this to generate a reasonable income) are quite large and the risk/reward on a lumpy portfolio would be very poor (ie you need 1000s of obligors to mitigate idiosyncratic default risk.) Funny thing is what you have described is basically the strategy of 80% of the credit hedge funds set up in the last few years. Works like a charm in a bull market…less well when things are less sunny (cf. sub-prime mortgage market!)
Forget about tranches - what about a proper ‘direct’ market in the probability that Joe makes his payment this week? Clearly some moral hazard issues (market could easily be bigger than underlying - Joe could game it, esp. depending on consequences - or not - of missing payment) but interesting to think about… of course not in the US because this of course would be betting. Bad. Bad.