1 – Artificial-Intelligence System Surfs Web to Improve Its Performance
Information extraction involves classifying data items that are stored in plain text, and is a major area of research for machine learning scientists. Last week, a research team from MIT introduced a new approach to information extraction for machine learning systems at the Association for Computational Linguistics’ Conference on Empirical Methods on Natural Language Processing, and won a best-paper award. Instead of feeding their system as much data as possible, the team’s winning approach takes a different route and focuses on a much smaller data set, a similar process used by human beings – if you’re reading a paper that you don’t understand, you’re likely to do a search on the web and find articles that you are able to understand. This new system approach does something similar; if the system’s confidence score is low in assessing a particular text, it will query for more information, pulling up a handful of new articles from the web that correlate with a specific set of terms. In future, this model could be applied to sparse data and save much time in reviewing databases.
(Read the full article at MIT News)
2 – RiskIQ Gets $30.5M to Apply Machine Learning to Security Risks
San Francisco-based digital risk management startup RiskIQ announced that it’s raised another $30.5 million Series C in a deal led by Georgian Partners, and including Summit Ventures, MassMutual Ventures, and Battery Ventures, putting its total funds raised at $65.5 million since 2009. RiskIQ AI-based services help large companies search and find sites and apps that may bear the company’s name but are run by criminals attempting to steal consumers’ information or spread malware. The company’s total bookings grew 80 percent in the first part of 2016, with a current total of 200 enterprise customers and 13,000 security analysts that includes Facebook, Under Armour, and others. Georgian Partners principal Steve Leightell will also join RiskIQ’s board of directors
(Read the full article at Silicon Valley Business Journal)
3 – First Carnegie Colloquium Focuses On Artificial Intelligence In Military, Data Privacy
Carnegie Mellon held the first of a two-part colloquium, which addressed considerations around AI on data privacy and military operations, for global policy experts at Carnegie Endowment for International Peace (CEIP) headquarters in Washington D.C. The second part will address internet governance and cyber deterrence, on December 2 at CMU’s Cohon University Center in Pittsburgh. CyLab Director David Brumley, who opened a second panel discussion for autonomous technology, said:
“Countries around the world, including the U.S., Russia, Israel, China and India, are increasingly deploying and investing in artificial intelligence and autonomy technology in their operations. Autonomy is going to be huge, and it’s absolutely critical we get it right.”
Jim Garrett, dean of CMU’s College of Engineering, emphasized that such forums are of vital importance for exchanging ideas and cultivating acceptance for a wide variety of of views on issues that have the potential to profoundly impact the global community.
(Read the full press release at Carnegie Mellon News)
4 – Oxford Researchers Develop Computer Program that Can Read Lips with Superhuman Accuracy
Researchers at Oxford have pioneered a lip-reading AI program that can read lips with 93.4 percent accuracy – far surpassing the average 52.3 percent accuracy for hearing-impaired students. Named “LipNet”, the software was built in collaboration with Google’s DeepMind, which trained it on 30,000 videos of test subjects. The system processed sentences (as opposed to individual words) and was able to place words in context. Though not yet ready for the diversity of languages, accents, and broken speech of the real world, the program has the potential to both help society – improve hearing aids, allow for conversation in noisy places, etc. – as well as harm – allow for individuals or groups to pick up on private conversations or conduct illegal mass surveillance.
5 – Machine-Learning Algorithm Quantifies Gender Bias in Astronomy
A paper by researchers from the Swiss Institute of Technology in Zurich and released on the arXiv server used machine learning to estimate gender bias in citations of academic papers in astronomy. Though not yet peer-reviewed, experts in the field have commented on what appears to be a valid methodology. Cassidy Sugimoto, an informaticist at Indiana University Bloomington, stated:
“The novelty of this paper is in dispelling the myth that gender disparity in citation can be attributed to specifics of the paper, rather than to gender.”
The algorithm was trained on 200,000 papers in 5 journals from 1950 to 2015. Results showed that papers with female authors listed first received around 6 percent fewer citations than those with a primary male author; the algorithm also predicted that those papers with female authors should have received 4 percent more citations than those authored by the males. In academics, less citations usually means fewer grants, recommendation letters, and other recognitions, says Meg Urry, director of Yale Center for Astronomy and Astrophysics. The paper also notes, however, that women publish 19 percent fewer articles than men in the 7 years after their first published paper, a critical time for contributing to academia. This may also play a factor in women securing more permanent positions.
(Read the full article at Scientific American)
Image credit: Tek-Think