Welcome to the June 2, 2017 edition of ACM TechNews, providing timely information for IT professionals three times a week.

ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).
Your Smart Home Is Trying to Reprogram You
The Conversation
Murray Goulden
May 31, 2017

As people increasingly adopt smart home devices and other Internet of Things technologies, a much broader movement is underway in which an entire class of technologies is seeking to remake the fundamentals of people's everyday lives, according to Murray Goulden, a senior research fellow at the University of Nottingham in the U.K. However, he says smart home technologies' designs for ubiquity and seamless control over both physical and virtual systems is not compatible with social realities. Goulden cites as one major drawback the technologies' failure to recognize social boundaries that humans take for granted. "Smart home technologies excel at creating data that doesn't fit into the neat, personalized boxes offered by consumer technologies," Goulden notes. "This interpersonal data concerns groups, not individuals, and smart technologies are currently very stupid when it comes to managing it." Goulden also points to the technologies' propensity for error, which could have potentially catastrophic results.

Full Article

Man sitting with his prosthetic leg crossed over his other leg Haptic Feedback Will Let Users Control Prosthetic Devices Better, Study Says
International Business Times
Rishabh Jain
May 31, 2017

Researchers at Rice University, the Italian Institute of Technology, and the University of Pisa in Italy have found that haptic feedback could help in providing patients the ability to discern the size and feel of objects without looking at them. The research will be presented at the World Haptics 2017 conference at Fürstenfeldbruck, Germany next Wednesday using 18 able-bodied subjects, who will be blindfolded and asked to feel and decipher objects using a skin-stretch upper arm device known as the Rice Haptic Rocker. It is a rubber pad fitted on the skin of the arm of a subject that does not stretch when the prosthetic hand is open, and stretches more as the hand closes. "We're using the tactile sensation on the skin as a replacement for information the brain would normally get from the muscles about hand position," says Rice University's Janelle Clark. The researchers hypothesize even non-invasive prosthetics could replicate the natural "muscle sense."

Full Article

An hourglass with blue sand Artificial Intelligence Predicts Patient Lifespans
University of Adelaide
David Ellis
June 1, 2017

Researchers at the University of Adelaide in Australia used artificial intelligence to analyze the medical imaging of 48 patients' chests and predict which patients would die within five years, with 69-percent accuracy. Although the researchers say they could not identify exactly what the computer system was seeing in the images to make its predictions, the most confident predictions were made for patients with severe chronic diseases. "Our research suggests that the computer has learned to recognize the complex imaging appearances of diseases, something that requires extensive training for human experts," says Adelaide researcher Luke Oakden-Rayner. He believes the research could lead to automated systems that can predict medical outcomes in ways that doctors are not trained to do by incorporating large volumes of data and detecting subtle patterns. "Predicting the future of a patient is useful because it may enable doctors to tailor treatments to the individual," Oakden-Rayner says.

Full Article
Physicists Uncover Similarities Between Classical and Quantum Machine Learning
Lisa Zyga
May 31, 2017

A team of European researchers has found the structure of certain types of quantum-learning algorithms is very similar to their classical counterparts, a breakthrough they say will help scientists further the development of the quantum versions. The research "shows that the potentially very complex operations involved in an optimal quantum setup can be dropped in favor of a much simpler operational scheme, which is analogous to the one used in classical algorithms, and no performance is lost in the process," says Gael Sentis at the University of the Basque Country in Spain. The researchers focused on a specific type of machine learning called inductive supervised learning, in which the algorithm is given training instances from which it extracts general rules, and then applies them to a variety of tests. The researchers demonstrated that both classical and quantum inductive supervised learning algorithms must have a training phase and a test phase that are distinct and independent.

Full Article

A pair of sheep Researchers Design AI System to Assess Pain Levels in Sheep
University of Cambridge
Sarah Collins
June 1, 2017

Researchers at the University of Cambridge in the U.K. have developed an artificial intelligence (AI) system that can detect pain levels in sheep, a breakthrough they say could aid in early diagnosis and treatment of common, but painful, conditions in other animals. The team notes the AI system uses five different facial expressions to recognize whether a sheep is in pain, and to estimate the severity of that pain. The system can detect the distinct parts of a sheep's face and compare it with a standardized measurement tool developed by veterinarians for diagnosing pain. The researchers trained the model using a small dataset consisting of about 500 photographs of sheep, which had been compiled by veterinarians in the course of providing treatment. The model is able to estimate pain levels with about 80-percent accuracy. The researchers now want to train the system to detect and recognize sheep faces from moving images.

Full Article
Machine Learning on Stampede2 Supercomputer to Bolster Brain Research
The Next Platform
Donna Loveland
May 31, 2017

A proof of concept published in 2016 suggests machine-learning algorithms running on supercomputers to classify neuroimaging data may yield the most reliable insights on the human brain to date. The researchers analyzed brain data from a group of depression sufferers and healthy controls using the Stampede system at the Texas Advanced Computing Center to predict major depressive disorder with 75-percent accuracy. The team taught a Support Vector Machine Learning algorithm by feeding it sets of data examples from both healthy and depressed subjects, tagging the features they designated meaningful. The resulting model scanned subsequent input, classifying the new examples as belonging to either the healthy or depressed category. The more advanced Stampede2 supercomputing system, scheduled for full deployment this summer, will function as "a strategic national resource to provide high-performance computing capabilities for thousands of researchers across the U.S.," according to the U.S. National Science Foundation.

Full Article
Experts Predict When Artificial Intelligence Will Exceed Human Performance
Technology Review
May 31, 2017

Katja Grace and colleagues from the University of Oxford's Future of Humanity Institute in the U.K. surveyed leading academic and industry artificial intelligence (AI) experts to speculate on when AI will outperform humans. The general consensus was AI will best humans at language translation, writing high school essays, driving trucks, and similar tasks within the next decade. However, the respondents do not expect AI to outperform people in retail until 2031, while machine surgeons will not come into their own until 2053. The experts also predicted a 50-percent likelihood that AI will be better than humans at virtually any task within approximately 45 years. Grace says a 40-year forecast should be tempered by the fact that it is the length of most people's working lives, and any anticipated technological change further out than that means the change will occur beyond the working lifetime of everyone employed today.

Full Article

A man standing on a curb with a cane Wearable System Helps Visually Impaired Users Navigate
MIT News
Larry Hardesty
May 31, 2017

Researchers from the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory have developed a new system employing a three-dimensional camera, a belt with five separately controlled vibrational motors, and an electronically refreshable Braille interface to help visually impaired users navigate their environments. The camera is worn in a pouch hung around the neck, while a processing unit runs the researchers' proprietary algorithms. One algorithm quickly identifies surfaces and their orientations from camera data, grouping pixels into clusters of three, each of which determines a plane. Should the orientations of the planes defined by five nearby clusters be within 10 degrees of each other, the system concludes that it has pinpointed a surface. To identify chairs, the system must complete three individual surface identifications in the same general area. The surfaces must be approximately parallel to the ground, and fall within a prescribed range of heights.

Full Article
Random Numbers: Hard Times Ahead for Hackers
University of Geneva
May 31, 2017

Researchers at the University of Geneva (UNIGE) in Switzerland have developed a new random-number generator based on the principles of quantum physics. The researchers note powerful quantum random-number generators currently are available commercially, but it is impossible for the user to independently verify the numbers generated are genuinely random. The user must trust the device to function correctly, even after years of use. "We wanted to create a device which can be continuously tested to ensure it functions correctly at all times and thus guarantee that the random numbers generated are reliable," says UNIGE professor Nicolas Brunner. The researchers say they achieved this by developing a self-testing quantum random-number generator, which enables users to verify in real time that the device delivers unbiased random numbers. The team believes the technology will provide a new layer of security for passwords and cryptographic protocols.

Full Article

Two students building a robot. Robot Design for Dummies
Carnegie Mellon University
Byron Spice
May 26, 2017

Researchers at Carnegie Mellon University's (CMU) Robotics Institute have developed a new interactive design tool that enables both novices and experts to build customized legged or wheeled robots using three-dimensionally-printed components and commercially available actuators. The team says the system includes a familiar drag-and-drop interface, enabling users to choose from a library of components and place them into the design. In addition, the tool suggests components that are interoperable with each other, as well as offering potential placements of actuators, and it can automatically generate structural components to connect those actuators. After the design is complete, the tool provides a physical simulation environment to test the robot before fabricating it. "The system makes it easy to experiment with different body proportions and motor configurations, and see how these decisions affect the robot's ability to do certain tasks," says CMU's Ruta Desai.

Full Article
Can We Quantify Machine Consciousness?
IEEE Spectrum
Christof Koch; Giulio Tononi
May 25, 2017

The argument that consciousness is computable--and instillable within machines--cannot be verified or refuted until a measurable theory of consciousness becomes available, and integrated information theory (IIT) could hold the answer. IIT identifies intrinsic existence, structure, integration, definition, and specificity as essential properties of all experiences of consciousness. IIT also dictates that the overall level of consciousness is based on its internal architecture, one that has a maximum of intrinsic cause-effect power. Since IIT is expressed mathematically, it can be representative of any physical system, but a major drawback in terms of generating conscious machines is its stipulation that the computation or simulation of intrinsic causal power is impossible--it must be architected into the system's physics. IIT insists computers imbued with human-level intelligence and behaviors will not be truly conscious, but machines based on less conventional architectures, such as neuromorphic hardware schemes, could in principle exhibit significant conscious experience.

Full Article
Learning About Nutrition From 'Food Porn' and Online Quizzes
Harvard University
Leah Burrows
May 24, 2017

Researchers at Harvard and Columbia universities conducted a study to determine whether popular online quizzes and food imagery could be leveraged for nutrition education. Using Harvard professor Krzysztof Gajos' LabIntheWild behavioral research platform, the researchers sought to evaluate volunteers' learning about nutrition in the context of an online social quiz. Participants compared photos of meals and gauged their nutritional value as part of the six-month test. Although participants who received feedback explaining the correctness of their answer performed better on the quiz and gained more knowledge than those who received no feedback or explanations, the difference between explanations generated by experts and explanations from peers was insignificant. "We can corral the wisdom of the crowd to help people make more informed decisions to improve their health," says Columbia professor Lena Mamykina. The work was presented last month at the ACM SIGCHI Conference on Human Factors in Computing Systems (CHI 2017) in Denver, CO.

Full Article
Pioneering Computer Scientist Calls for National Algorithm Safety Board
Thomas Macaulay
May 31, 2017

University of Maryland professor Ben Shneiderman has proposed creating a National Algorithm Safety Board to ensure that software is held accountable and to restore trust in algorithms damaged by recent scandals concerning their use to spread biases, misinformation, and inequality. Shneiderman, who received the ACM SIGCHI Lifetime Achievement Award in 2001, says the board would oversee planning, continuous monitoring, and retrospective analysis independently, which together would provide a foundation to ensure the correct system is chosen and then supervised, with the resulting knowledge applied toward better future algorithm production. Shneiderman's oversight model would entail open adversarial reviews, advisory boards, planning commissions, and internal and external audits, with transparency delivered by exposing the inner mechanisms of algorithms. Accountability would be implemented via open failure reporting, and liability by terminating agreements that protect parties from liabilities that arise during the contract. Shneiderman cites the U.S. Federal Reserve Board as a template for his proposal, and also calls for overseers' authorization to subpoena information and enforce recommendations.

Full Article
MIT Advances in Imaging
ACM Books

Association for Computing Machinery

2 Penn Plaza, Suite 701
New York, NY 10121-0701

ACM Media Sales

If you are interested in advertising in ACM TechNews or other ACM publications, please contact ACM Media Sales or (212) 626-0686, or visit ACM Media for more information.

To submit feedback about ACM TechNews, contact: technews@hq.acm.org