Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Joshua New
New York Marathon

This week’s list of data news highlights covers November 4-10, 2017, and includes articles about an AI system that can read research papers to aid materials fabrication and a program to make quantum computers more reliable.

1. Spotting Ovarian Cancer with AI

Researchers at the Dana-Farber Cancer Institute and Brigham and Women’s Hospital have developed a non-invasive test that uses genetic sequencing and machine learning to diagnose ovarian cancer with 91.3 percent accuracy. Ovarian cancer is traditionally difficult to detect before symptoms appear, but at that point the cancer can already be in its latter stages, making ovarian cancer particularly dangerous. The researchers’ test analyzes genetic defects caused by many ovarian cancers called microRNAs, which can appear in blood before symptoms develop, and then uses a machine learning algorithm to detect the presence of specific microRNAs in a patient’s blood.

2. Making Trucking Safer with a Hat

Ford has developed a hat designed for truck drivers called SafeCap that uses sensors to track wearers’ head movements and alerts wearers if it appears they are falling asleep while driving. SafeCap contains a gyroscope that can detect movement patterns indicating wearers are fatigued, and uses sound, light, and vibration to warn drivers and call their attention back to driving.

3. Finding Recipes for Materials Fabrication

A group of researchers from the Massachusetts Institute of Technology, the University of Massachusetts at Amherst, and the University of California at Berkeley have developed an AI system that can analyze research papers and extract information about how to produce particular materials. The system identifies paragraphs in research papers containing materials recipes and classifies key words in the recipe, including the kinds of equipment involved, quantities, and operating conditions. In tests, the system could identify recipe paragraphs with 99 percent accuracy and correctly classify the parts of recipes with 86 percent accuracy.

4. Running the New York Marathon Blind

Blind athlete Simon Wheatcroft successfully completed the New York Marathon with assistance from wearable technology from startup WearWorks that provided Wheatcroft with detailed navigation cues through haptic feedback. WearWorks developed a bracelet called the Wayband that links with a user’s smartphone and uses its GPS to create a model of the race route. When Wayband detects the wearer is veering off course, it will vibrate in different patterns to signal which corrective action to take, such as two long vibrations to signal a right turn. WearWorks also provided Wheatcroft with a wearable device called the Tortoise that uses ultrasonic sound waves to detect nearby objects, such as another runner, and vibrates to provide warning when these objects get too close.

5. Predicting the Risk of Side Effects from Radiotherapy

Researchers at the Institute of Cancer Research, London, have developed an analytical technique that uses AI to predict how sensitive prostate cancer patients will be to the side effects of prostate radiotherapy. Because radiotherapy can affect people differently, patients regularly receive either too high a dose, which increases the risk of side effects, or too low a dose, which avoids side effects but reduces the effectiveness of the treatment in the process. By better understanding the likelihood a patient will experience side effects, this technique could eventually lead to more personalized radiotherapy treatments that minimize the risk of side effects and are more likely to be effective.  

6. Using Wearables to Develop Precision Medicine

The U.S. National Institutes of Health’s (NIH) precision medicine initiative, called the All of Us Research Program, will trial the use of wearables made by FitBit to study how to best incorporate the biometric data it collects into its planned million-person search cohort. NIH will distribute 10,000 FitBit devices to a representative sample of volunteer participants to collect data about wearer’s physical activity, heart rate, sleep patterns, and other biometric activity for one year.

7. Making Quantum Computers More Trustworthy

Researchers at IBM, the University of Maryland, and Georgia Tech have developed a program for quantum computers that can detect when data has become corrupted, which can cause errors and make quantum computing analysis unreliable. Normal computers use duplicated data to detect and fix errors by reconstructing corrupted data based on uncorrupted data elsewhere on the machine. With quantum computers, duplicating a quantum state requires measurement, but measuring them loses information, making traditional error correction approaches ineffective. The researchers’ program uses qubits to recognize erroneous data and automatically change the state of other qubits to correct this information while the computer operates. The program reduces the error rate of quantum programs from between 10 and 15 percent down to 0.1 percent.

8. Opening Uber’s AI to the Public

Uber’s artificial intelligence lab has made Pyro, a programming language it built to develop AI applications, publicly available as open source. Uber designed Pyro to perform an analytical technique called deep probabilistic modeling that combines deep learning and another statistical technique called Bayesian modeling, making it well suited for dealing with models with high levels of uncertainty.

9. Training Robotic Workers with Virtual Reality

A startup called Embodied Intelligence is developing a system to train smart manufacturing robots by having them observe and learn from humans controlling them in virtual reality. The system will use virtual reality headsets to allow humans to operate a robot while it collects data about their actions, such as grasping and assembling complex objects, to serve as training data.

10. Machine Learning Can Make Your Bad Photos Better

Ukrainian startup Let’s Enhance has created photo enhancing software that uses machine learning to upscale low-resolution images to make them appear to be high-resolution. The software, which people can access for free online, enhances images in three steps. First, it identifies and removes artifacts caused by the low resolution, such as jagged lines. Then, it sharpens the image while preserving details and maintaining the edges of the image content. Finally, it tries to identify the contents of the image and find similar examples in a database so it can add in details that might not have been captured in the original image.

Image: Martineric

You may also like

Show Buttons
Hide Buttons