Thursday, April 10, 2008

Microsoft Clearflow as an example of over-engineering

Microsoft announced plans to launch Clearflow, a web service that incorporates complex software models to help users avoid traffic jams. Read about it in various places, including the New York Times. My first thought when reading about it wasn't that it sounds like a cool technology, it was that it sounds like a ridiculously over-engineered solution to a problem.

The problem:

Commuters don't have accurate data regarding current traffic situations on main and side streets.

The artificial intelligence researchers' solution:

Form a team and spend 5 years building a predictive model for current traffic conditions based on four years of data and 16,500 trips covering 125,000 miles. The end result means that Clearflow combines live traffic data based on data from traffic sensors and combines that with predicted traffic events based on variables such as time of day, weather, current sporting events.

End result:

Probably more accurate traffic information, but largely based on intelligent guesses.

My solution (I'm not an artificial intelligence researcher):

Use real data, not predictive models. My GPS knows which road I'm on, the speed limit of that road, where the next intersection is, and how fast I'm traveling. Because my GPS knows that, the GPS satellite knows that. Because the GPS satellite knows that about me, it also knows it about other drivers on that same road or other roads. So, let's use all of that data to determine how fast traffic is actually moving on a particular road (relative to the speed limits and intersections) and therefore determine whether traffic problems exist on that road right now. Next step: automatically update the display on my GPS to suggest alternate routes that do not currently have similar traffic problems.

End result:

Extremely accurate traffic information, using actual current data and no guesses.

Perhaps I'm missing something obvious. Or perhaps this is an example of the adage, "to the man with a hammer, every problem looks like a nail". Thoughts? Counter-arguments? Anyone?

No comments: