This week’s list of data news highlights covers October 6-12, 2018, and includes articles about Kansas City’s smart sewer system and an initiative to give AI common sense.
Chinese technology company Tencent and London-based medical firm Medopad are developing an AI system to diagnose patients with Parkinson’s disease faster than traditional methods. The system relies on a smartphone camera to film patients’ hands as they move and uses AI to analyze the state of their motor functions. Motor function assessments normally takes over half an hour, but Tencent and Medopad are training their system on footage of diagnosed Parkinson’s patients to reduce this time to three minutes and allow patients to take the assessment from home.
Researchers at Facebook have developed an AI system that can analyze a picture of food and determine the ingredients used to make it. The researchers trained their system on images of dishes paired with the recipe and in tests on just images of dishes, the system can generate what it thinks the appropriate recipe is.
Fish farming company Cermaq Group AS has developed a system called iFarm that uses facial recognition to identify individual salmon so it can better track and prevent the spread of epidemics such as sea lice, which infect hundreds of millions of farmed fish per year. iFarm uses a 3D scanner to identify salmon based on distinct patterns of spots on their faces when they surface and its sensors could screen for lice or other abnormalities, alerting the comment to quarantine them if necessary. Tracking the health of individual fish and treating diseases early could reduce fish mortality by 50 to 75 percent.
Google has developed a feature for its intelligent phone bot Duplex that can screen potential spam calls. When users suspect a call is spam, they can press a button to have Duplex ask callers their name and reason for calling. Duplex will transcribe this into an on-screen notification to allow users to decide whether to take the call.
Kansas City, Missouri, has developed the world’s largest smart sewer and stormwater management system, which it expects to save nearly $1 billion by 2035 and prevent approximately 5 billion gallons of sewer overflow per year. The system uses nearly 300 sensors embedded in 2,800 miles of sewer pipes feeding data into decision support systems that can help municipal workers better control the flow of water and prevent sewage from spilling into the MIssouri River. Kansas City has spent $505 million to prevent sewer overflows since 2010, and expected to spend $2.5 billion to comply with the requirements of the Clean Water Act by 2035, however the new system is expected to reduce this cost by up to $1 billion.
Instagram has announced it will use AI to detect bullying in users’ photos and captions and proactively send content to human moderators to review. Instagram developed a meaning learning tool to detect offensive language in comments last year, and has expanded the tool to detect content indicating bullying in both text and photos. The tool, which Instagram is also adding to live videos, is designed to detect harassment that targets of bullying often do not report.
Engineers at the University of Washington have developed a system for 3D printing devices that can track and store information about their use without the use of electronics. The system relies on backscatter, a method for sharing information by reflecting signals directed at an object from an antenna. The 3D printed devices are plastic and embedded with gears and springs that can respond to specific movements in a way that produces a unique signal via backscatter, which can then be interpreted by monitoring software. The engineers demonstrated how this approach could be used to track the amount of laundry detergent left in a bottle, as well as how many times a pill bottle has been opened, without the need for the bottles to be powered or use electronics.
The nonprofit Global Emancipation Network and analytics firm Splunk have developed an analytics platform named Minerva to help identify human traffickers and their victims. Minerva can analyze text and images on the web and deep web and helps organizations such as the National Center for Missing and Exploited Children identify useful data in large, difficult-to-search data sets, such as identifying to whom a phone number that appears in an online sex advertisement is registered.
Researchers at the Naval Medical Center San Diego and Google have developed an AI system that can evaluate lymph node biopsies and detect metastatic breast cancer more accurately than human experts. 90 percent of breast cancer deaths are the result of metastasis, however studies have shown that breast cancer patients often experience inadequate examinations and diagnostic tests. The system, named LYNA (Lymph Node Assistant), achieved 99.3 percent accuracy in tests, while human pathologists, who often miss small metastases, can be as low as 62 percent accurate.
The U.S. Defense Advanced Projects Research Agency (DARPA) has launched a new initiative called the Machine Common Sense (MCS) program to develop AI systems capable of common sense reasoning. Though AI researchers have made significant progress in natural language processing, it is still challenging for AI systems to make inferences from text that rely on common knowledge about the real world. The MCS program will create computational models designed to learn from experience and mimic human cognition, as well as develop a “common sense knowledge repository” capable of answering questions that require common sense by analyzing information from the Internet.
Image: Brian Hillegas.