1 – DARPA Challenge Tests AI as Cybersecurity Defenders

In August, seven teams will compete in DARPA’s Cyber Grand Challenge, pitting autonomous AI systems against threats and attacks from other network servers and patching up proprietary software errors in the process. The U.S. military agency hopes to promote programs that can work side by side with humans examining crash reports and assessing network vulnerabilities. Network protection can be implemented in the form of either ‘binary armor’ or through the finding and correcting of software malfunctions. In addition to a $2 million prize, the winning team will compete against human hackers at the DEF CON 2016 conference taking place in Las Vegas at the same time, marking the first time that a completely automated system will go ‘head to head’ with human experts.

(Read the full article on IEEE Spectrum)

2 – eBay Acquires Predictive Analytics Startup SalesPredict to Boost its Machine Learning

On Monday, eBay announced its plans to acquire the Israel-based company SalesPredict, which uses predictive analytics to assess customer behavior and sales conversions. The purchase will augment and support eBay’s machine learning and data science goals. Amit Menipaz, vice president and general manager of Structured Data, commented:

“For our buyers, it will help us better understand the price differentiating attributes of our products, and, for our sellers, it will help us build out the predictive models that can define the probability of selling a given product, at a given price over time.”

SalesPredict’s technology will specifically add the data processing and inferencing portion of eBay’s structured data initiative cycle. An acquisition price was not disclosed, but some of SalesPredict’s employees will join eBay’s structured data team in its Israeli Development Center in Netanya.

(Read the full article on eBay News)

3 – Researchers Want to Achieve Machine Translation of the 24 Languages of the EU

Two EU-funded research projects aim to use the same machine learning methods as DeepMind’s Go program to develop systems able to translate between the 24 languages of the EU. The approach uses pattern recognition based on large quantities of data as opposed to ‘feeding’ the computer an array of often complex grammatical rules and syntactic contexts. Saarbrücken Computer Linguist Josef van Genabith, who is leading the projects, explained:

“This machine learning strategy has nothing to do with natural intelligence, but it does have similarities with the processes that occur in the human brain when we control the muscles in our bodies. Children have to learn to pick up their feet when walking in the woods so as not to trip over roots or stones. In adults, this sort of mental process runs automatically in the background, as the brain has learnt how their feet have to be placed.”

At present, Josef and his research teams are identifying all possible “language resources” from European government ministries that cover relevant areas, such as texts and translations in finance, foreign affairs, and economics. Though computers may be soon able to translate vast quantities of information more quickly, van Genabith notes that human translators may still be needed for some time to edit the ‘imperfect’ texts.

(Read the full article on EurekAlert)

4 – Machine Learning Puts New Lens on Autism Screening and Diagnostics

Research teams from the University of Southern California (USC), along with leading autism researchers in the field, recent published a paper on their research in using machine learning to help screen for autism and provide parent/caretaker guidance. The study’s authors used two recognized industry tests (ADI-R and SRS) and used machine learning methods to analyze parents’ responses. From this data set, researchers identified five ADI-R questions that seemed most capable of maintaining 95 percent of the instrument’s performance. In addition to reducing administrative costs, the approach could result in more personalized interview question sets and help prevent misdiagnoses of individuals.

(Read the full article on USCViterbi)

5 – Duke Robotics Research Team Creates Chip to Expand Robots’ Motions

Researchers at Duke University have developed a prototype chip that helps robots master the complex calculations needed in choosing the most efficient path of motion.  According to Daniel Sorin, a W.H. Garnder Jr. professor of electrical and computer engineering, the chip overcomes typical ‘set motion’ paths and allows robots to make faster and more intelligence decisions about its next moves. The chip, which uses an algorithm that runs thousands of potential motions at once, has been successfully tested on a robotic arm and the team has visions of its use in a range of applications that include autonomous vehicles. The team is currently working on taking the chip from prototype to a more specialized version for use in industry.

(Read the full article on Duke’s The Chronicle)

Image credit: iStockphoto