System Bits: May 17

AI drives Toyota websites
An innovation in artificial intelligence described in a 2001 paper by UCLA computer science professor Adnan Darwiche has found a somewhat unexpected application: helping car buyers of Toyota and Lexus customize their vehicles online.

The websites let shoppers tailor their vehicle from among a range of models, colors and accessories. The software that powers the sites, called a “product configurator,” is based on a logical form of artificial intelligence devised by Darwiche, that performs sophisticated, real-time reasoning to ensure that if a consumer wants a specific vehicle — for example, a red Camry with a tan interior and a performance package — that exact combination of options could be manufactured by the company or is available in its inventory. The websites can also reason about features that are co-dependent, such as removing a minimum number of features when a combination is not feasible or determining which features must be bought together.

Darwiche’s innovation is known as a decomposable negation normal form circuit, and it addressed a central challenge in artificial intelligence: How to efficiently reason with knowledge, which is a core task of intelligent behavior. His research on DNNF circuits is just one part of his body of work in a field called knowledge compilation, which, in essence assembles knowledge into simple forms that allow reasoning to be conducted very efficiently. He proposed the compilation of knowledge into tractable circuit representations, including DNNF circuits, leading to very simple and minimal reasoning systems.

More recently, Darwiche has started an effort to use tractable circuits to help solve a highly complex type of computer science problems known as Beyond NP problems, which he said will expand the use of computers in various applications, including medical decision making, yet current knowledge is limited on how to solve them very efficiently.

Low-tech snoopers easily identify Twitter users’ homes, workplaces
According to MIT and Oxford University researchers, the location stamps on just a handful of Twitter posts — as few as eight over the course of a single day — can be enough to disclose the addresses of the poster’s home and workplace to a relatively low-tech snooper.

The tweets themselves might be otherwise innocuous — links to funny videos, say, or comments on the news. The location information comes from geographic coordinates automatically associated with the tweets, they said.

Latanya Sweeney, professor of government and technology in residence at Harvard University and a former chief technology officer of the U.S. Federal Trade Commission said the survey shows how people can learn sensitive information from seemingly innocuous facts, and, second, people will easily share information they believe is innocuous.

Harnessing thousands of network cameras for public safety
Adding fuel to the growing vision processing market, Purdue University researchers have developed a prototype system that could allow law enforcement and public safety agencies to tap into thousands of cameras located in numerous venues including parking garages, college campuses, national parks and highways.

In addition to applications in law enforcement, the system can be used to quickly find damage, plan rescues and other operations during natural disasters such as tornadoes, hurricanes and blizzards, according to David Ebert, Purdue University’s Silicon Valley Professor of Electrical and Computer Engineering and director of Visual Analytics for Command, Control and Interoperability Environments, or VACCINE, a U.S. Department of Homeland Security center based at Purdue.

While surveillance cameras usually are operated in closed circuits commonly called CCTV and are available only to authorized personnel, in recent years many organizations have deployed cameras for a wide range of purposes, and these are accessible to the public without the need for a password. ,” said Yung-Hsiang Lu, an associate professor of electrical and computer engineering. And although the cameras are not deployed for surveillance purposes, they can be utilized to increase public safety by properly integrating with current surveillance systems.”

Purdue researchers reminded that they previously demonstrated a system that can allow law enforcement to see the locations and viewing angles of CCTVs. The new work extends the previous study by including real-time streams of public network cameras that are deployed by city and state governments along highways and at intersections, national parks, construction sites and other public venues. They expect the number of network cameras and their coverage to grow substantially in the near future.