Weekly News Shogi

Published on December 8th, 2017 | by

0

10 Bits: the Data News Hotlist

This week’s list of data news highlights covers December 2-8, 2017, and includes articles about a new AI-powered genetic analysis tool and Japan’s first self-driving taxis.

1. Tracking Thailand’s Economy with Big Data

Thailand’s central bank is using data from social media, ecommerce platforms, and other sources to supplement official statistics about the national economy. Thailand has the lowest unemployment rate in Asia at 1.3 percent, but this is likely due to the fact that official statistics do not accurately measure the large informal sector of Thailand’s economy. The Bank of Thailand is developing an employment index built on data from online job-search platforms, and it has conducted analysis about household debt and the export industry revealing a much less positive outlook on Thailand’s economy than official data sources would suggest.

2. Understanding Your Genome with AI

Google has made its AI genetic analysis tool DeepVariant freely available as open source, allowing researchers to more accurately analyze sequencing data. The widespread accessibility of genetic sequencing in recent years was made possible due to a technique called high-throughput genetic sequencing, which is fast and inexpensive but error-prone, which limits the usefulness of data collected from this technique. DeepVariant uses machine learning to differentiate between these sequencing errors and actual genetic mutations so it can filter out these errors and create a more accurate picture of a genome.

3. An AI Chatbot Can Help You Fight Sexual Harassment

Montreal startup Botler AI has developed a free-to-use AI chatbot that can talk to users who have been sexually harassed to give them guidance about their situation and provide them with relevant legal information. The chatbot asks users information about their situation, such as where they live and when the incident occured, and then generates an incident report users can give to authorities based on its predictions about which laws may have been broken.

4. Japan is Finally Getting Self-Driving Taxis

Nissan has announced that it will begin deploying its self-driving taxi service in Japan in March 2018, making it the first domestic car manufacturer to due so. Despite Japan’s high-tech reputation and sizeable car manufacturing industry, efforts to deploy self-driving cars have slowed due to regulations that only recently changed to allow the testing of autonomous vehicles on public roads, but that are still cumbersome and require testers to coordinate with local municipalities and law enforcement.

5. AI Masters Two New Board Games

DeepMind’s AI system AlphaGo Zero has outperformed the world’s leading AI systems for playing chess and the Japanese board game shogi after only several hours of training. AlphaGo Zero was able to beat or tie the leading chess-playing software Stockfish in every one of the 100 games they played after only four hours of training on the game. AlphaGo Zero also was able to beat the leading shogi-playing AI, named Elmo, 90 times in 100 games after training for just two hours.

6. Training Algorithms to Reduce Bias

Researchers at Google have modified a smile detection system by developing racial and gender classifier algorithms that can help the system more accurately detect smiling faces than “color blind” systems that ignore racial and gender categories in an effort to reduce bias. Many developers deliberately avoid using such classifiers under the assumption that it could introduce bias to their system and make it less effective at identifying faces of a certain race or gender. However by explicitly focusing on and training on racial and gender differences, the researchers’ approach increased the smile detection system’s accuracy by 1.5 percent.

7. Measuring the Performance of Australian Cities

Australia’s Digital Transformation Agency has launched a new tool called the National Cities Performance Framework that uses data from Australia’s 2016 Census to create a dashboard benchmarking economic and social performance indicators in different Australian cities. The dashboard allows members of the public to easily access and compare statistics such as cities’ unemployment rates, greenhouse gas emissions per person, patent applications per 100,000 residents, and jobs accessible within a 30-minute commute.

8. Inventing New Alloys with AI

HRL Laboratories in California is using machine learning to develop recipes for creating new alloys that can be 3D printed for use in airplane bodies. 3D printing can help create sophisticated and intricate metal parts for airplanes, however it requires powdered metal alloys which typically are not structurally sound at the atomic level when the printer welds them together. HRL generated 10 million different variations of a powdered aluminum alloy that could make it strong enough for airplane parts and uses machine learning algorithms to identify formulations that could meet the necessary physical and chemical requirements.  

9. Turning Space Debris into an Internet of Things Platform

Researchers at Fudan University in Shanghai have launched a microsatellite with a suite of microchips as an attachment to a rocket that was used to propel a satellite into orbit in November, with the goal of establishing a space-based wide area Internet of Things network. During a space launch, rockets jettison portions of themselves into space as they run out of fuel to reduce their weight. The researchers installed a microsatellite weighing just 30 grams on one of these rocket portions and have successfully communicated with it in orbit. With this initial success, the researchers plan on launching more of these microsatellites with future rockets to create a network of sensors in orbit that can communicate with each other and with systems on the ground.

10. Predicting How a Neighborhood Votes Based on Google Street View

AI researchers at Stanford University have developed a system that can predict the voting patterns of a neighborhood based on the cars that appear in Google Street View. The system uses one algorithm to classify cars that appear in Street View imagery by make and model, and another algorithm to analyze data from the census and the 2008 elections. By combining these factors, the researchers were able to identify relationships between car ownership and voting preferences. For example, the system was able to correctly identify how 58 out of 60 precincts in Gilbert, Arizona voted based on the proportion of pickup trucks to sedans in each precinct. The researchers also found strong correlations between the demographic makeup of a neighborhood and ownership of particular car brands.

Image: Immanuel Giel

Tags: , , , , ,



Comments are closed.

Back to Top ↑

Show Buttons
Hide Buttons