Driverless Cars Could Let You Choose Who Survives in a Crash

New Scientist

Researchers at the University of Bologna in Italy have developed a dial that will switch a smart car's setting from "full altruist" to "full egoist," with the middle setting being impartial. The researchers think the new "ethical knob" could work for all areas of industry that are becoming increasingly autonomous. The system tells an autonomous car the value that drivers give to their life relative to the lives of others. "The car would use this information to calculate the actions it will execute, taking into account the probability that the passengers or other parties suffer harm as a consequence of the car's decision," says the University of Bologna's Guiseppe Contissa. However, others studying this issue have reservations. For example, if everyone chooses the maximum self-protective mode, it could end in a Tragedy of the Commons type scenario. In addition, if everyone were to choose the impartial option, the ethical knob will not help with any existing dilemma.

From "Driverless Cars Could Let You Choose Who Survives in a Crash"
New Scientist (10/13/17) Abigail Beall
View Full Article

Could AI Be the Future of Fake News and Product Reviews?

Scientific American

Researchers at the University of Chicago are experimenting with artificial intelligence (AI)-based techniques for automatically generating convincing online reviews, such as bogus Yelp restaurant critiques. "We have validated the danger of someone using AI to create fake accounts that are good enough to fool current countermeasures," says University of Chicago professor Ben Zhao. The team employed a deep-learning artificial neural network to extract letter and word patterns used in millions of existing reviews, and used them produce its own reviews. Testing showed Yelp's filtering software, which also relies on machine-learning algorithms, had problems distinguishing real from fake critiques. In addition, many human evaluators also were fooled by the automated posts. Zhao's team is building algorithms designed to serve as countermeasures to identify phony reviews, and they also are considering further research into fake news detection. The researchers will present their research in November at the ACM Conference on Computer and Communications Security (CCS 2017) in Dallas, TX.

From "Could AI Be the Future of Fake News and Product Reviews?"
Scientific American (10/16/17) Larry Greenemeier
View Full Article

Addressing National Priorities and Societal Needs Symposium

CCC Blog

The Computing Community Consortium on October 23-24 will host the "Computing Research: Addressing National Priorities and Societal Needs Symposium" to focus on the current and future roles of computing in addressing societal needs. The symposium is centered around four themes, including Intelligent Infrastructure for our Cities and Communities, with highlights of current progress as well as the need for significantly more cross-disciplinary research. The theme of Artificial Intelligence (AI) and Amplifying Human Abilities will concentrate on AI's emergent role in enhancing human abilities in such fields as health, transportation, universal access, data analysis, and education. The Security and Privacy for Democracy session will detail new technology helping society manage information security and privacy risks. The final session on Data, Algorithms, and Fairness will emphasize the pressing need to make sure computer programs are not used for discriminating, promoting inequality, or undermining social justice.

From "Addressing National Priorities and Societal Needs Symposium"
CCC Blog (10/12/17) Khari Douglas
View Full Article

Internet Researchers Harnessed the Power of Algorithm to Find Hate Speech

Aalto University

Researchers in Finland have trained a machine-learning algorithm to identify hate speech by comparing computationally what differentiates text that includes hate speech from text that does not contain hate speech and developing a categorization system for hate speech. The researchers employed the algorithm on a daily basis to screen all openly available content that municipal election candidates had generated on Facebook and Twitter. The algorithm was taught using thousands of messages, which were cross-analyzed to uphold scientific validity. "When categorizing messages, the researcher has to take a stance on the language and context, and it is therefore important that several people participate in interpreting the teaching material," says the University of Helsinki's Salla-Maaria Laaksonen. She says social media services and platforms could identify hate speech if they wanted to, and in that way influence the activities of Internet users. "There is no other way to extend it to the level of individual citizens," Laaksonen notes.

From "Internet Researchers Harnessed the Power of Algorithm to Find Hate Speech"
Aalto University (10/13/17)
View Full Article

Quantum Communication Demonstrated in Real-Life City Conditions

Live Science

Researchers at the University of Ottawa in Canada for the first time have transmitted a quantum-encrypted message containing two bits of data in each photon between buildings in an actual city. An accompanying study estimates this technique can cut in half the number of photons required to send a message. Ottawa professor Ebrahim Karimi notes applying high-dimensional encoding would fortify security, further protecting the quantum communication channel from the effects of external "noise." The team wants to test the sending and receiving of high-dimensional quantum-encrypted messages at distances of up to 3.5 miles so they can employ their technique on an urban scale, which requires surmounting significant challenges, including that of turbulence as the light passes through the air. "In our experiment, we are sending a single photon, so that is really difficult," Karimi says. "You need to send it to go under a certain angle and use a complicated telescope with sophisticated electronics."

From "Quantum Communication Demonstrated in Real-Life City Conditions"
Live Science (10/10/17) Tereza Pultarova
View Full Article

Forget About It

Argonne National Laboratory

Researchers at Argonne National Laboratory led a study integrating supercomputer modeling and x-ray characterization of a material that gradually "forgets" as a potential tool for advanced bio-inspired computing. "The brain...can only function efficiently because it is able to forget," says Argonne's Subramanian Sankaranarayanan. "It's hard to create a non-living material that shows a pattern resembling a kind of forgetfulness, but the specific material we were working with can actually mimic that kind of behavior." The material, called a quantum perovskite, exhibits an adaptive response when protons are repeatedly inserted and subtracted that resembles the brain's desensitization to a recurring stimulus. The researchers found the material's electrical resistance is affected by proton doping, enabling the perovskite to be programmed like a computer. "These simulations, which quite closely match the experimental results, are inspiring whole new algorithms to train neural networks to learn," says Argonne's Mathew Cherukara. He notes this could lead to more efficient artificial intelligence.

From "Forget About It"
Argonne National Laboratory (10/10/17) Savannah Mitchem
View Full Article

New Technique Scours the Genome for Genes That Combat Disease

MIT News

Researchers at the Massachusetts Institute of Technology (MIT) have employed a modified CRISPR genome-editing system to screen for genes that protect people against certain diseases. The team adapted CRISPR to randomly activate or deactivate distinct gene sets across large cell populations to identify genes that protect cells from a protein associated with Parkinson's disease. "What we decided to do was take a completely unbiased approach where instead of targeting individual genes of interest, we would express randomized guides inside of the cell," says MIT professor Timothy Lu. "Using that approach, can we screen for guide RNAs that have unusually strong protective activities in a model of neurodegenerative disease." Once the researchers spotted disease-protecting genes in yeast cells, they tested the human equivalents in human neurons that overproduce the protein alpha synuclein. Lu notes these genes also shielded against alpha-synuclein-induced death, suggesting they could be worth evaluating as gene therapy treatments for Parkinson's disease.

From "New Technique Scours the Genome for Genes That Combat Disease"
MIT News (10/12/17) Anne Trafton
View Full Article

AI and Aerospace Models Used to Optimize Blood Flow in Veins

Imperial College London

Researchers at Imperial College London in the U.K. have used computer modeling techniques to model how unsteady the currents are in the blood flows of patients undergoing dialysis. The researchers used machine-learning algorithms to train a computer to optimize the shape of an arterio-venous fistulae (AVF) so the unsteadiness in the blood flow could be suppressed. The prototype device the researchers developed to hold the AVF in the optimal shape has undergone successful preliminary tests in pigs. The researchers plan to conduct trials with pigs for several months at a time to further test the effectiveness of the AVF device. "We routinely use computer simulations to study air flow over airplanes," says Imperial College London's Peter Vincent. "These same techniques can now be used to optimize medical devices, including AVF." The researchers say the prototype AVF technology could benefit other medical procedures to optimize the shape of blood vessel connections.

From "AI and Aerospace Models Used to Optimize Blood Flow in Veins"
Imperial College London (10/10/17) Colin Smith
View Full Article

Johns Hopkins Scientists to Build Machine Translation System for Obscure Languages


A team of Johns Hopkins University (JHU) scientists has received a $10.7-million grant from the U.S. Office of the Director of National Intelligence to create an information retrieval and translation system for obscure languages. JHU professor Philipp Koehn is leading the development of a system that can respond to queries typed in English based on documents written in "low resource" languages for which little written material exists. Koehn says the team will compile online samples of a target language that have been translated into English, and then perform a machine analysis of language patterns, including sentence structures and positions of constituent components. The researchers plan to use the data to derive algorithms for automatically translating the target language, with the system designed to respond to inquiries featuring a word or term and a topic area or "domain." Koehn says the response should tell the user how the content is pertinent to the inquiry.

From "Johns Hopkins Scientists to Build Machine Translation System for Obscure Languages"
JHU Hub (10/09/2017) Arthur Hirsch
View Full Article

Cornell to Launch Center for Data Science for Improved Decision Making With $1.5-Million Grant

Campus Technology

Researchers at Cornell University say they have received a $1.5-million grant from the U.S. National Science Foundation's (NSF) Transdisciplinary Research in Principles of Data Science program to launch the Center for Data Science and Improved Decision Making, whose goal is to ensure that data management systems are responsible and employed for the public benefit. The researchers note the center will focus on how to guarantee privacy and ensure that decision-making systems are not biased by characteristics such as race, gender, or age. In addition, the center will study the structure and processes of social networks, as well as research interventions based on user histories to offer recommendations and suggestions related to a range of issues, quantify the uncertainty of predictions, and examine deep-learning systems. A separate team of Cornell researchers received a $3-million NSF grant last month to study the ethics of big data decision making.

From "Cornell to Launch Center for Data Science for Improved Decision Making With $1.5-Million Grant"
Campus Technology (10/09/17) Joshua Bolkan
View Full Article

Intel Accelerates Its Quantum Computing Efforts With 17-Qubit Chip

IEEE Spectrum

Intel is shipping an experimental quantum computing chip to research partners in the Netherlands to demonstrate that its packaging and integration technology give the company an edge in developing practical quantum computers. The chip contains 17 superconducting quantum bits (qubits), the minimum for performing surface code error correction, an algorithm needed to scale up quantum computers to useful sizes. Intel's research partners will be testing the individual qubits' abilities and conducting surface code error correction and other algorithms. For the new chip, Intel adapted flip chip technology to work at millikelvin temperatures. Flip chip technology involves adding a dot of solder to each bond pad, flipping the chip upside down atop the circuit board, and melting the solder to bond it--resulting in a smaller, denser, and lower inductance connection. By developing the infrastructure technology along with quantum computing chips, "we're putting in the work to deliver something that looks more like a computer," says Intel's Jim Clarke.

From "Intel Accelerates Its Quantum Computing Efforts With 17-Qubit Chip"
IEEE Spectrum (10/10/17) Samuel K. Moore
View Full Article

Put Humans at the Center of AI

Technology Review

In an interview, Stanford University Artificial Intelligence (AI) Lab director Fei-Fei Li advocates for more human-centered AI and the benefits it can yield. She notes although most current AI breakthroughs in pattern recognition are significant, they lack the contextual awareness and learning flexibility of people. Making technology safer, productive, and better for humans "requires a layer of human-level communication and collaboration," Li says. She believes returning contextual understanding, knowledge abstraction, and reasoning to AI research is essential for making machines more helpful and useful. Li cites her Visual Genome image database at Stanford as "exactly the kind of project that's pushing the boundaries of language understanding and visual understanding [by AI]." Li also sees a strong economic need for more diversity in the AI workforce, which should power greater innovation and creativity while also helping to instill human morals and ethical values within AI.

From "Put Humans at the Center of AI"
Technology Review (10/09/17) Will Knight
View Full Article