Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Joshua New
by
Waymo

This week’s list of data news highlights covers December 1-7, 2018, and includes articles about robotic janitors at Walmart and an AI system that can identify high-performing engineers.

1. Predicting the Structure of Proteins

Researchers at DeepMind have developed an AI system called AlphaFold that creates accurate 3D models of how proteins fold. AlphaFold predicts the distances between pairs of amino acids, which make up proteins, and the angles between chemical bonds that connect amino acids. Figuring out the structure of a protein is a challenging task for scientists, but the shape of a protein affects its function and misfolded proteins are thought to cause several diseases, such as Alzheimer’s.

2. Using AI to Detect Depression

Researchers from Stanford University have developed an AI system that can accurately predict if someone is depressed by analyzing their facial expressions, voice tone, and word choice. To create the system, the researchers trained a machine learning algorithm on footage of depressed and non-depressed people speaking to a physician through an avatar. The AI system can predict if someone is depressed with 80 percent accuracy and the researchers believe smartphones could use this technology to supplement traditional mental health care.  

3. Waymo is Offering an Autonomous Vehicle Ride-Hail Service

Waymo, the autonomous vehicle subsidiary of Google parent Alphabet, has begun offering Waymo One, an autonomous vehicle commercial ride-hail service, to hundreds of pre-approved riders in the greater Phoenix, Arizona area. Riders use an app to arrange and pay for a ride in the autonomous vehicle, but Waymo-trained drivers will supervise the vehicles at first. In addition, Waymo is still offering fully driverless rides to individuals who are apart of its early rider volunteer program.

4. Building Virtual Worlds From a Sketch

Researchers at Nvidia have developed AI software that can generate realistic 3D imagery for a virtual environment after analyzing just an outline of a scene. By feeding the software a basic outline of an environment indicating where different objects should be, the software will generate images to fill in the scene, such as cars, trees, and buildings. This approach could help dramatically reduce the amount of work required to develop realistic virtual environments for video games or virtual reality applications.

5. An AI App Can Translate Children’s Books Into Sign Language

Chinese technology company Huawei has released StorySign, an Android app that uses AI to translate children’s books into sign language. Huawei developed the app, which supports several sign languages including British, French, and Spanish, in collaboration with the nonprofit European Union for the Deaf, publishing company Penguin, and British animation studio Aardman. The app uses image recognition technology to detect words on a page, which an avatar translates into sign language.

6. An AI App Can Detect Anemia

Researchers from Emory University have developed an app that can predict if a person has anemia, a common blood disorder, by analyzing pictures of his or her nails. Individuals with anemia lack enough healthy red blood cells, which carry the distinctive red protein hemoglobin, and fingernail beds are an indicator of hemoglobin levels because there are no skin cells to mask the color. A study of more than 300 people found that the app predicts hemoglobin levels better than a physician’s physical assessment, however a blood test is still more accurate.

7. Robot Janitors Will Clean Walmart

Walmart will be using 360 autonomous robots to clean the floors of some of its stores by the end of January 2019. The robots, made by San Diego startup Brain Corporation, use an array of sensors to detect objects and look similar to a small Zamboni. Before they can clean autonomously, however, humans must drive the robots to teach them the store layouts. Once the robots learn a store’s layout, they can clean autonomously and even scrub the store floors when customers are shopping.

8. An Algorithm Predicts Cryptocurrency Pump-and-Dump Schemes

Researchers from the Imperial College of London have developed an algorithm that predicts pump-and-dump schemes, in which investors illegally attempt to inflate the price of an asset prior to selling, involving cryptocurrencies. The researchers trained a machine-learning algorithm on the data from 237 pump-and-dump events, teaching the algorithm the signs of a potential pump-and-dump scheme, such as an individual purchasing coins of a largely inactive cryptocurrency. The researchers then exposed the algorithm to live cryptocurrency data for a week, and the algorithm identified six possible pump-and-dump schemes, five of which were actual pump-and-dump schemes.

9. AI Can Spot Unlikely Engineers

A company called Catalyte has developed an AI system that can evaluate job applicants based on their potential to become a good engineer, rather than their existing ability. The software runs on an applicant’s computer while they take an engineering test, analyzing how they behave as they answer questions. For example, the software observes how an applicant reacts to questions designed to induce stress, such as by tracking their keystrokes and web browser use, as well as whether applicants go back to answer earlier questions once given new information, which can indicate cognitive ability. Catalyte designed the software to identify potentially high-performing workers that go overlooked due to cultural reasons or dated hiring practices.

10. Finding the Missing Malaysian Airlines MH370 Flight

Martin Kristensen, an engineer at Aarhus University in Denmark, has published an analysis of data related to Malaysian Airlines MH370, a plane that mysteriously disappeared somewhere over the Indian Ocean in 2014, suggesting the plane could be at a previously unconsidered location near Christmas Island. Kristensen developed a model that could factor in complex satellite communication data as well as potential distortions, such as the plane’s movement, which identified a possible water crash site near Christmas Island, which has not yet been searched. Kristensen claims that there is above a 90 percent chance the plane crashed in this area, which is 87 miles long by 18.6 miles wide.

Image: Grendelkhan

You may also like

Show Buttons
Hide Buttons