Association for Computing Machinery
Welcome to the October 15, 2008 edition of ACM TechNews, providing timely information for IT professionals three times a week.


UK University Holds Artificial Intelligence Test
Associated Press (10/13/08) Satter, Raphael G.

The University of Reading recently conducted its annual Turing Test of artificial intelligence. Dozens of volunteers at split-screen terminals carried out two conversations simultaneously, one with a chat program and one with a human. After five minutes, the volunteers were asked to identify the human and the machine. The chatbot Elbot was declared the winner for fooling three out of the 12 judges assigned to evaluate the program's conversational skills, earning the Loebner Artificial Intelligence Prize's bronze medal. The contest is based on the ideas of British mathematician Alan Turing, who in 1950 argued that conversation was proof of intelligence, and if a computer talked like a human, then for all practical purposes it thought like a human. Each of the programs approached the Turing Test in slightly different ways. One program often referenced its native Odessa and "Aunt Sonya in America." Another used humor to try to fool the judges. Elbot tried to throw the judges off by humorously admitting it was a machine, saying it accidentally poured milk on its cereal instead of oil, and by trying to dominate the conversation to keep it from wandering into areas it was not properly programmed to handle. Elbot's bronze medal is awarded to the software that best mimics human conversation in text form. So far, no silver or gold medals have been awarded. A silver medal would go to a machine that could pass a longer version of the Turing Test and fool at least half the judges, and a gold medal would be awarded to a machine that could process audio and visual information in addition to text.

Probe Sees Unused Internet
Technology Review (10/15/08) Lemos, Robert

Internet addresses may not be running out as quickly as many feared, concludes a new research study. The study found that millions of Internet addresses have been assigned but remain unused. In a paper to be presented at the ACM Internet Measurement Conference, which takes place October 20-22, in Vouliagmeni, Greece, six researchers have documented what they say is the first complete census of the Internet in more than two decades. The researchers discovered a surprising number of unused addresses and predict that plenty of addresses will still be unused when the last numbers are assigned in a few years. The researchers say the main problem is that some companies and institutions are using only a small portion of the millions of addresses they have been allocated. The paper's lead author, University of Southern California professor John Heidemann, says the study indicates that there might be better ways of managing the IPv4 address space. A new map of the Internet created by the study suggests that there is room for more hosts even if addresses are running out. The map found that roughly a quarter of all blocks of network addresses are still unused. IPv4 offers about 4.3 billion addresses, while IPv6, the next-generation Internet address scheme, will offer 51 thousand trillion trillion addresses.

Revived Fervor for Smart Monitors Linked to a Server
New York Times (10/13/08) P. B7; Vance, Ashlee

Many technology companies have been working to replace individual PCs at employee desks with thin clients that connect to a server, significantly reducing the need for upgrades and fixes on individual machines. Although thin-client systems have so far failed to live up to their promise, largely because of slow connections and awkward software, the technology is making a comeback as major companies increase their investments and roll out new products. Thin-client supporters have abandoned challenging Microsoft, and in most cases they even make it easy to use Windows over the network, but they continue to push thin clients as a way to save companies money and make their systems more secure. The technology offers several benefits. Software glitches, updates, and network security are much less of a problem, as the thin-client model places the burden of managing computers on professionals monitoring servers instead of each worker with a computer. Perhaps the biggest benefit is that all user data is stored on the network, so if a machine breaks it can be swapped out for another one with no data loss. The biggest changes driving this new wave of thin-client systems are improvements in networking and software. The spread of high-speed Internet connections enables users to connect from a cubicle or from home without worrying about connection speed and reliability, and manufacturers continue to fine-tune software that controls communication between thin clients and servers, leading to smoother-running machines that can handle even video and audio applications.

Databases Should Track Who Alters Information, Says BCS Speaker (10/14/08)

Computer scientist Charlie Bachman says people who create, modify, or delete data should have their actions recorded and linked with the outcome as part of an effort to control the problem of illegal or inappropriate alterations to databases. Bachman, who will give a lecture on the evolution of database technology to the British Computer Society (BCS) Data Management Specialist Group later this month, says that when a person alters a database, their authorship should be recorded along with the data so others can observe who the author was and evaluate the authorship appropriately. Bachman also says that some transactions or edits might require a dual authorship to authorize changes. He says user IDs and password protection against unauthorized access also could be extended to protect against unauthorized updates. A recent BCS report suggested that the original collector of personal data has an ongoing responsibility for the accuracy of that data when it is shared. Combining new application systems with existing ones will continue to be the predominant problem of database management for as long as there are databases, Bachman says, with the biggest challenges being the integration of new application systems with all of the existing applications systems.

Feds Considering Changes to H-1B Application Process in Wake of Report
Computerworld (10/10/08) Thibodeau, Patrick

Evidence of fraud being used in H-1B applications will likely result in increased scrutiny of the visa petitions, concludes a report by the U.S. Citizenship and Immigration Services (USCIS). A USCIS spokesman says the agency is considering a series of reforms in the H-1B application process, including the use of independent open source data to obtain information about visa applicants or the companies that file petitions on their behalf. For example, USCIS officials could verify petitioning companies against commercially available records to ensure they are legitimate businesses. The USCIS also is considering implementing a risk assessment program for applicants that would be based on objective criteria relating to fraud indicators, enabling the agency to give greater scrutiny to H-1B petitions that raise question marks during the review process. Other possible reforms include modifying the H-1B evidentiary requirements and revising the forms that employers use when filing applications. The report says that 21 percent of 246 H-1B applications reviewed by USCIS staffers contained either outright fraud or technical violations of federal laws and regulations, meaning thousands of employers may be violating the rules.

IBM Expands Research Efforts in China
Wall Street Journal (10/14/08) P. B5; Areddy, James T.

IBM is opening a new research facility in Shanghai, China, that will develop new applications for the Internet and small businesses. China's rapid growth, huge population, and large number of private businesses are attracting research efforts from some of the world's largest technology companies. IBM has eight research and development labs worldwide, and has not opened a new research facility since 1998, when it opened two in India. The Shanghai lab will act as an extension of IBM's Beijing lab, which opened in 1995. IBM director of research John E. Kelly III says the Shanghai lab "punctuates" IBM's commitment to long-term research. The new lab, located in a technology park in eastern Shanghai, will start with about a dozen computer scientists and engineers and eventually expand to about 100. IBM did not disclose how much it plans to spend on the new center. Last year, IBM allocated $6.15 billion to research, development, and engineering, or about 6.2 percent of the company's $98.79 billion in annual revenue. IBM researchers are already preparing demonstrations of programs currently in development in China, including one that helps retailers decide where to locate stores, and one for fact-checking potential clients.
View Full Article - May Require Paid Subscription | Return to Headlines

Computers That Understand How You Feel
University of Twente (10/13/08) Wanders, Rianne

A Ph.D. candidate at the University of Twente has developed a dialog system that is capable of interpreting human emotion. Trung Bui used a mathematical technique developed in the 1960s for controlling factory processes, but took a hybrid approach by combining the Partially Observable Markov Decision Process (POMDP) with the Dynamic Decision Network (DDN) technique. POMDP works well with small-scale dialog problems, but DDN-POMDP enables dialog systems to look ahead and determine whether they have enough calculating power to handle larger problems. Bui used the hybrid strategy in a navigation system for emergency services, and a separate stress module allows it to measure the emotional level of the user during communication. The navigation system is capable of anticipating that more mistakes are likely to be made when the user has a raised stress level, and will regularly ask the user for confirmation.

Academics Sink Teeth Into Yahoo Search Service
CNet (10/10/08) Shankland, Stephen

Academics and startups can construct their own search sites around Yahoo's search engine at no charge and manipulate results as they see fit through Yahoo's Build Your Own Search Service (BOSS), and the venture could give Yahoo potentially higher standing in a market where Google reigns supreme. BOSS can be used to modify search results, as illustrated by an application used by Chengxiang Zhai and Bin Tan of the University of Illinois at Urbana-Champaign. Their application directed Yahoo's search engine along specific paths based on the data stored on the user's own computer to deduce which of several items that shared the same name a user was more likely to be searching for. "We believe the client side of personalization ... can alleviate concern over privacy and it can provide more information about user activity," Zhai says. "And it can naturally distribute computation" so a search company's machines share work with the user's own system. Another service of potentially substantial value to academics is Yahoo's search assist feature, which suggests searches based on what people have started to type into the search box. For instance, it can display the variations of a search term, its membership in diverse categories, and the probability that people are searching for the term by itself or as part of a bigger query. "That's got a lot of potential," says Stanford University natural-language processing Ph.D. candidate Dan Ramage.

The Sum of Your Facial Parts
New York Times (10/09/08) Kershaw, Sarah

Researchers at Tel Aviv University have developed "beautification engine" software that uses a mathematical formula to alter the original form of a person's face in a picture to a theoretically more attractive version. The program is based on the responses of 68 men and women who viewed pictures of male and female faces and picked the most attractive ones. Data from the survey was applied to an algorithm involving 234 measurements between facial features, including the distances between lips and chin, the forehead and the eyes, and between the eyes. The research was published in the August proceedings of ACM's SIGGRAPH. Studies have shown that there is a surprising amount of agreement over what makes a face attractive. Symmetry is key, along with youthfulness, clarity or smoothness of skin, and vivid color, such as in the eyes or hair. Tommer Leyvand, one of the developers of the software and a Microsoft researcher, says the goal was to tackle the challenge of altering a face according to agreed-upon standards of attractiveness while producing a result that left the face completely recognizable. "This tool shows in the most simple fashion how easy it is to manipulate photographs and make people more attractive," Leyvand says. "But the difference is so subtle that it just shows how insignificant it is." Leyvand says the software could have practical applications in advertising, film, and animation.

Grids Get Down to Business
ICT Results (10/13/08)

European researchers have developed a system that permits managers to design and implement business processes across a grid through the A-WARE project. The new platform supports easy access to grid resources. Different computing platforms can share resources via grids, and A-WARE researcher Claudio Cacciari says this capability is very valuable to companies that "have so many different types of computers for each department." He adds that the system was also designed to understand business processes rendered in languages such as Business Process Modeling Notation, and that it interoperates with existing enterprise application standards. This spares business experts from becoming grid experts to develop new processes on the system. The A-WARE system has three layers of functionality: The grid-layer, a Web-based portal layer that makes grid functionality and resources easily accessible, and middleware that acts as a bridge between the other layers. Cacciari says companies will benefit tremendously from the A-WARE system's ability to extend the functionality and flexibility of their enterprise systems, and to facilitate grid applications that combine power with ease of use.

How Spam Is Improving AI
Technology Review (10/14/08) Kleiner, Kurt

Online security puzzles are under attack from researchers who are using them to develop smarter, more humanlike algorithms. The most common type of puzzle, CAPTCHAs, a series of distorted letters and numbers, is increasingly being solved by smarter artificial intelligence (AI) software. Earlier this year, University of Newcastle researcher Jeff Yan unveiled a program capable of completing the textual CAPTCHAs used to protect Microsoft's Hotmail, MSN, and Windows Live services, with a 60 percent success rate. Meanwhile, Palo Alto Research Center researcher Philippe Golle has developed a program that can solve a newer, image-based CAPTCHA called Asirra, developed by Microsoft, which asks users to correctly classify images of either cats or dogs using a database of three million images. Golle's program can correctly identify Asirra's cats or dogs 87 percent of the time. Golle trained his program using 8,000 images collected from the same Web site Asirra used. The software uses a statistical analysis of color and texture, using clues such as the pink of a dog's tongue or the green in a cat's eyes. Although the program can recognize each individual picture 87 percent of the time, the full Asirra CAPTCHA requires 12 pictures to be identified simultaneously, so the attack only works 10.3 percent of the time. Golle says Microsoft could increase the difficulty for machines by adding more pictures. Yan will present details of a new program that can solve more widely used textual CAPTCHAs at ACM's Computer and Communications Security Conference, which takes place Oct. 27-31, in Alexandria, Virginia.

Breathing Second Life Into Language Teaching
EurekAlert (10/09/08) Garcia-Ruiz, Miguel A.

An international research team has developed a wireless virtual reality environment that can promote language learning and help students practice their foreign language skills. University of Colima, Mexico, researchers, working with researchers at the University for Technology in Amman, Jordan, are using virtual reality to help teach students foreign languages. "Virtual reality is today one of the new frontiers in computer-assisted language technology," the researchers say, "offering a stimuli-rich environment for language students." One virtual reality software application of particular interest to educators is the open source Distributed Interactive Virtual Environments (DIVE), which was developed by the Swedish Institute of Computer Science and can be run on a variety of operating systems. DIVE allows users to share a virtual environment over a network, either local or over the Internet, and has a three-dimensional graphical user interface that enables communication through either voice or text chat, with each user represented by an avatar. A new environment created by the researchers, Realtown, is a virtual reality environment within a DIVE installation. Realtown has a virtual supermarket, schools, a pharmacy, a bank, and other locations. Realtown allows students to simultaneously perceive and interpret visual, auditory, and physical stimuli to help incorporate their language knowledge.

CarTel Personalizes Commutes by Using WiFi to Network Cars
MIT News (10/08/08) Thomson, Elizabeth A.

The Massachusetts Institute of Technology's (MIT's) CarTel project uses automobiles to monitor the surrounding environment by sending data from an onboard computer to a Web server where data can be visualized and browsed. The cars use pre-existing Wi-Fi networks they encounter while on a trip. The resulting data, which can be accessed from the Web or on a cell phone, helps drivers track conditions specific to their cars and provides historical and real-time traffic conditions at different times of the day using combined data from all CarTel participants. MIT professor Hari Balakrishnan says the goal is to make the data behind CarTel available to the public and to help travelers plan and organize their trips. The current system, which was deployed in 50 Boston-area cars in January, monitors traffic by recording each vehicle's speed at different points during a round trip. MIT professor Samuel Madden says CarTel understands where traffic delays are and recommends alternative routes to avoid them. To make CarTel possible, researchers developed QuickWiFi, technology that can connect to a Wi-Fi network in 360 milliseconds. Researchers also had to focus on managing the huge amounts of data the system gathers. Depending on the sensors in use, CarTel can receive more than 600 data points per second, so the team developed two generations of software to synthesize all the data for use, including a series of new algorithms for traffic-aware routing.

IBM Researchers Using iPhone App Store as Test Bed
Wired News (10/07/08) Ganapati, Priya

IBM Almaden research scientists are using Apple's iPhone Applications Store to launch mobile Web research projects, including an experimental text-input system and an application to sync multiple devices. IBM researcher Shumin Zhai says the iPhone App Store provides an opportunity to "experiment in the wild." IBM Almaden researchers are thinking broadly on how to redesign the user experience for common applications such as email and calendars for desktops, laptops, and smart phones. The researchers say that iPhone's touch screen has set the standard for user interaction among smart phones and possibly among future computers. Zhai's WritingPad allows words to be "drawn" by tracing a continuous line from letter to letter on a keyboard. The system infers the word from the pattern drawn across the keyboard, enabling users to significantly increase their typing speed. Zhai says after only 24 hours on the App Store, WritingPad received 60 user reviews, and now more than 500 reviews have been written. IBM Almaden expects to release the Personal Information Environment program next, which synchronizes and manages data across multiple devices using an instant messaging-like protocol to enable users to share information and data.

Researchers Show How to Crack Popular Smart Cards
InfoWorld (10/07/08) de Winter, Brenno

Researchers at the Dutch Radboud University Nijmegen have published a cryptographic algorithm and source code that could be used to duplicate smart cards used by several major transit systems. The scientists presented their findings at the Esorics security conference in Malaga, Spain, and also published an article with cryptographic details. The research demonstrated how to circumvent the security mechanism of NXP Semiconductor's Mifare Classic RFID cards, which are widely used to provide access control to buildings and public transportation. The researchers exposed the workings of the chip by analyzing communication between the chip and the reader. A RFID-compatible device, the Ghost, was designed to work independently from a computer, which allowed the researchers to obtain the cryptographic protocol. Part of the vulnerability comes from the fact that the RFID reader has to communicate in a predictable way. Once the mechanism was exposed, the scientists were able to crack keys in less than a second using an industry standard computer with only 8MB of memory. The researchers also examined another chip, the Hitag2, to crack Mifare. Information on a Hitag2 hack is freely available online, which helped the researchers crack Mifare. Another effort by German researcher Henryk Plotz cracked the Mifare Classic by removing a Mifare chip from a card and removing layers, photographing each layer under a microscope and analyzing all the connections.

Massive Experiments, Global Networks
Christian Science Monitor (10/09/08) P. 13; Spotts, Peter N.

E-science involves huge, cross-disciplinary experiments conducted by many groups using grid computing, a process that taps the computing power of interconnected computer centers at universities and labs across the globe. Purdue University's Gerry McCartney says e-science and its computing networks are "leveling the playing field of opportunity" in science, and notes that one e-science network seeing a lot of use is PolarGrid, which researchers are employing to study images of Arctic and Antarctic ice sheets for the purpose of behavior modeling. A central goal of PolarGrid is to give field researchers real-time access to the wider grid network so they can assess images, model ice-sheet activity, and use the information to adapt their field experiments to shifting conditions while the scientists are still on the ice. Meanwhile, the European Organization for Nuclear Research recently launched the grid network it will use to analyze the data from its Large Hadron Collider by harnessing the computing muscle of more than 140 computer centers in 33 nations. The collective processing power of the TeraGrid in the United States is open to any university researcher in the country, while the Network for Computational Nanotechnology was set up "to put nanotechnology simulation tools in the hands of people who otherwise wouldn't touch them with a 10-foot pole," says Purdue professor Gerhard Klimeck.

Abstract News © Copyright 2008 INFORMATION, INC.
Powered by Information, Inc.

To submit feedback about ACM TechNews, contact:
Change your Email Address for TechNews (log into myACM)

About ACM | Contact us | Boards & Committees | Press Room | Membership | Privacy Policy | Code of Ethics | System Availability | Copyright © 2014, ACM, Inc.