jump to navigation

The Machines Run Wall Street January 13, 2011

Posted by Metabiological in Synthetic Intelligence.
Tags: , ,
comments closed

There’s an excellent interview up on NPR concerning the use of high speed computers on Wall Street.  You can listen to it right here.

A few things I take away:

Firstly, despite what the commentator alludes this isn’t SI really.  These algorithms don’t display sentience and certainly don’t mimic human thought process.  But then they’re not being asked to think like us, they’re being asked to do things we can’t.  Their advantage over a human trader is in being able to sift large amounts of data and data sets and determine what variables exert the greatest affect on the market.  This isn’t an example of synthetic intelligence but it is a further example of the effects that automation will have on our society and economy.

Many people have been under the false pretense (or should I say illusion) that automation would only effect the industrial and manufacturing industries by eliminating blue-collar jobs and anything that requires manual labor.  This is flat out wrong.  While I could point out how we cling to our collective delusion that there is something special about human intelligence that can never be replicated that’s not really what’s going on here.  These algorithms aren’t simulating human thought because human thought isn’t required, only the ability to sift data and recognize patterns.  Now stop for a moment and think of all the jobs that really can be boiled down to those activities.  Quite a few right?  This is the next frontier of automation and it’s going to take us for a ride (for a really good read on the subject, look up The Lights in the Tunnel.)

Secondly, I’m amazed that so much of our prosperity rest on a system that even insiders claim to not really understand.  I can sympathize with those working the market to a degree since I fully understand the difficulty involved in predicting the actions of complex systems (I am an ecologist after all) but it scares me to think of how little we know about something that exerts such a great influence on all of us.  The commentator mentioned one company in particular, Volion, that doesn’t even know what their trading on.  Putting aside the fact of how that’s even possible I wonder… no screw that, HOW IS THAT EVEN POSSIBLE!?  How have we built the driver of our economic prosperity on quicksand?

Finally, the major problem I see here is the major problem that confronts any use of computers to make our decisions for us: how to we know they’ll make the right ones.  This isn’t a question of computers going rogue and taking over the world a la Skynet.  This is a simple question of priorities.  How do we know that computers will have the same priorities as we do?

Imagine for a moment that we build a computer, the most powerful computer ever, blessed with freedom and ability to solve any problem we put to it.  We ask this computer to solve a seemingly impossible math problem.  The computer, following our instructions, proceeds to convert all matter on earth, including us, into hardware to increase it’s computational ability and solve the problem.  Sound crazy?  It isn’t.  These algorithms do not think like we do and cannot be predicted to behave as we do.

As we hand more and more power over to machines we can’t predict working within a system we don’t understand, what do you think the odds are something will go wrong?