Editorial

Introduction to the Special Articles in This Issue

Artificial Intelligence on Mobile Devices

Qiang Yang, Feng Zhao

n This issue of AI Magazine contains several n the film 2001: A Space Odyssey, a bulky computer system guest-edited articles devoted to some exemplary known as HAL captivated a generation of people with the works of AI on mobile devices. It includes four concept of artificial intelligence (AI) and intelligent works that range from mobile activity recogni - I machines. Today, as the world moves into one in which every - tion and air-quality detection to machine trans - lation and image compression one owns at least one mobile device, be it a smartphone, a tablet, or other handheld device, applications on the devices are increasingly more intelligent as well. Estimates put the global revenue for mobile apps at $150 billion dollars, 1 and mobile speech-recognition platforms will grow 68 percent through 2017 through cloud-based solutions, according to ABI Research. 2 We will see more and more applications of AI on the mobile devices. This special issue of AI Magazine is devoted to some exempla - ry works of AI on mobile devices. We include four works that range from mobile activity recognition and air-quality detection to machine translation and image compression. These works were chosen from a variety of sources, including the Interna - tional Joint Conference on Artificial Intelligence 2011 Special Track on Integrated and Embedded AI Systems, held in Barcelona, Spain, in July 2011. In “User-Centric Indoor Air-Quality Monitoring on Mobile Devices,” written by Yifei Jiang, Kun Li, Ricardo Piedrahita, Yun Xiang, Lei Tian, Omkar Mansata, Qin Lv, Robert P. Dick, Michael Hannigan, and Li Shang, the authors develop a novel and important technique for portable indoor air quality (IAQ) detec -

Copyright © 2013, Association for the Advancement of Artificial Intelligence. All rights reserved. ISSN 0738-4602 SUMMER 2013 9 Editorial tion. They present their system MAQS, ing, which leads to a deliberation-intu - and Technology. His research interests are which uses a mobile-phone-based ition architecture to balance accuracy data mining and artificial intelligence indoor location tracking system to and energy efficiency. Focusing on trip including machine learning, planning, and determine the whereabouts of users. detection, the deliberation phase uses activity recognition. He is a Fellow of IEEE, IAPR, and AAAS. He received his Ph.D. from MAQS uses air-quality sensing devices sensors that provide information such the Computer Science Department of the as GPS and Wi-Fi data in a cell-ID-based to report on the quality measurements University of Maryland, College Park in for an individual or a group of people learning system, and the intuition 1989. He previously served as assistant, then in real time. This system demonstrates phase relies on the cell ID patterns to associate professor at the University of an integrated ability to locate users predict the start and end of a trip. The Waterloo from 1989–1995, and a professor based on Wi-Fi signals and to apply an system was tested on real-life mobile and NSERC Industrial Research Chair at air-exchange-rate-based IAQ method phone data over a long period of time, Simon Fraser University in Canada from 1995–2001. He was elected as a vice chair of using CO 2 sensors and a zone-based showing a large amount of energy sav - proximity detection method for ings and high accuracy and timeliness ACM SIGART in July 2010 and is the found - pulling the data from different sensors in intuitive trip detection. ing editor in chief of ACM Transactions on Intelligent Systems and Technology. Yang is on and phones together. The MAQS sys - In the final article, “Learning Com - the editorial board of IEEE Intelligent Systems pact Visual Descriptors for Low Bit Rate tem in particular tackles the important and has served as a program comittee issue of energy savings by allowing the Mobile Landmark Search,” Ling-Yu cochair and general cochair of several inter - different devices to share information. Duan, Jie Chen, Rongrong Ji, Tiejun national conferences. He serves on the IJCAI In “Speaking Louder than Words Huang, and Wen Gao consider a very Board of Trustees and will be the IJCAI pro - with Pictures Across Languages,” important problem of mobile-based gram chair for IJCAI 2015. Andrew Finch, Wei Song, Kumiko systems: how to transmit large images Tanaka-Ishii, and Eiichiro Sumita using low-bit mobile-based image Feng Zhao is an assistant managing director explore an iPad-based system for cross- search. The method that they employ at Microsoft Research Asia, where he over - language communication using pic - is clever: once a mobile user enters a sees multiple research areas including hard - tures. Their system allows users of one region, the server transmits a down - ware computing, mobile and sensing, soft - ware analytics, systems, and wireless and language to express their thoughts in a stream supervision to “teach” the networking. His own research has focused mobile device by linearly projecting sequence of picture icons, based on on networked embedded systems such as which they generate natural language the original high-dimensional image sensor networks, energy and resource man - in the native languages of the users of into a compact code book. Then, users’ agement of distributed systems as in data the system for verification of the user’s queries based on a high-dimensional centers and cloud computing, and mobile intention. Then, the system can trans - code-word histogram can be converted devices and mobility. He has also done work late the user’s sentences to another lan - to a compact histogram, which is trans - on parallel computing, fast N-body algo - guage based on the meaning of the pic - mitted for search and query. rithms, machine recognition, and diagnos - ture icons, completing the machine- The four articles in this special issue tics. Prior to Microsoft Research Asia, Zhao was a principal researcher at Microsoft translation process. This second chan - represent a cross section of work in the increasingly popular area of AI on Research Redmond, where he founded and nel of communication makes the managed the Networked Embedded Com - mobile devices. We can see that the AI machine-translation result more puting Group. The group has developed the robust. In addition, as the authors systems on mobile phones must con - MSR sensor mote, tiny Web Service, claim, the process of pointing and tap - sider a number of new issues and con - SenseWeb and SensorMap, Data Center ping on icons is a natural way of straints, including limited energy Genome, JouleMeter, and GAMPS data com - inputting on mobile devices. They resources, constrained communication pression. With the help of some of these implemented their idea in a system bandwidth, relatively small display technologies, today the Microsoft data cen - known as picoTrans, which is able to area, and emphasis on user interaction ters are among the most densely instru - predict the user’s intended expression models. In addition, it is necessary to mented and monitored cloud computing infrastructures in the world. Zhao received from the icon sequence most of the consider a number of other issues for AI systems on mobile phones, such as his Ph.D. in electrical engineering and com - time. This system not only shows the puter science from MIT, has taught at Stan - geographical information, location possibility of machine translation on ford University and Ohio State University, mobile devices, but also points at a new techniques, user studies, and comput - and currently also serves as an affiliate fac - way of natural interaction between er-human interaction. ulty member at the University of Washing - users and mobile devices. ton and an adjunct professor at Hong Kong University of Science and Technology. Zhao In “Thinking Fast and Slow: An Notes Approach to Energy-Efficient Human was a principal scientist at Xerox PARC and Activity Recognition on Mobile 1. See Strategy Analytics (www.strategyana - founded PARC’s research effort in sensor Devices,” Yifei Jiang, Du Li, and Qin Lv lytics.com). networks and distributed diagnostics. He served as the founding editor-in-chief of propose a two-phase approach to activ - 2. See ABI Research (www.abiresearch.com). ACM Transactions on Sensor Networks. An ity recognition for trip detection using Qiang Yang is the head of Huawei Noah’s IEEE Fellow and ACM Distinguished Engi - a mobile phone. The two-phase Ark Research Lab and a professor in the neer, he received a Sloan Research Fellow - approach applies a well-known theory Department of Computer Science and Engi - ship and NSF and ONR Young Investigator in psychology of human decision mak - neering, Hong Kong University of Science Awards.

10 AI MAGAZINE