We are talking manhole explosions.
Those terrifying eruptions can blast a 300-pound manhole cover high above the street, trailed by a hideous cloud of almost-blinding flame and rust-colored gunk. Rudin, now a professor of computer science and engineering at Duke University, led a project that used machine learning to predict which Con Edison manholes were most at risk of blowing up.
For that research and other work that applied artificial intelligence to the everyday world — including the basis of a program used by the Police Department to identify crimes in different precincts that might be related — Rudin has just won a $1 million prize. The award, the Squirrel AI Award for Artificial Intelligence for the Benefit of Humanity, was announced by the Association for the Advancement of Artificial Intelligence.
Rudin, who has also collaborated on a formula that predicts the likelihood of seizures in critically ill hospital patients who have brain injuries, became involved with the manholes when Con Edison hired researchers from Columbia University. The assignment was massive: There are more than 50,000 manholes in Manhattan alone. They serve as entrances to underground conduits housing enough electrical cable to stretch almost all the way around the world at the Equator.
The cables are a particular worry in winter, when melting snow, along with salt and chemicals from road de-icing, can seep into manholes and cause short-circuiting if the insulation on the cables deteriorates. Rudin said that even a few sparks in the tight confines of a manhole can cause combustion.
Rudin’s team assembled information about every manhole on every block — a huge trove of data. The algorithm that Rudin and her team developed searched for patterns behind manhole explosions. In a test by Con Edison, the top 2 percent of “trouble holes” in Manhattan — manholes with problems — that the algorithm ranked as vulnerable included 11 percent that had recently had fires or explosions. Con Edison said it continues to use similar system analysis to predict what it calls “manhole events.”
Rudin did not need an algorithm to predict what she would do with the $1 million.
“The first thing I’m going to do is allocate 40 percent to pay taxes,” she said. “I feel that’s a really important thing to do, to support our society.”
“You hear about these multimillionaires and the interesting things they do with their money,” she continued. “I’ve never heard anyone brag about taxes, but I’m so proud of the public schools my children go to and our police officers and our firefighters and people who tend our parks and all of those whose job it is to keep society going.”