This week’s list of data news highlights covers December 12, 2020 – December 18, 2020 and includes articles about using AI to control military jets and calculating how well COVID-19 quarantine measures are working with machine learning.
Researchers at the Los Alamos National Laboratory, a U.S. Department of Energy national laboratory, have used a supercomputer to optimize the distribution of COVID-19 vaccines. Using simulations of cities populated by individuals with different demographic characteristics, such as ethnicity and employment type, the researchers modeled the spread of COVID-19 under different conditions. They concluded that prioritizing vaccines for healthcare workers, rather than distributing the vaccine randomly, is more effective in reducing the number of COVID-19 cases and deaths.
The U.S. Air Force is using AI to control sensor and navigation systems on the U-2 Dragon Lady spy plane, marking the first time the military has used AI onboard an aircraft. During a two-and-a-half hour long test flight, the AI system controlled the plane’s radar system and tactical navigation. Developers trained the AI system on open-source software algorithms and adapted to the plane’s computer systems.
Mathematicians and cryptologists at the University of Melbourne have used a supercomputer to solve a 50-year-old coded message written by an as yet unidentified serial killer, known only as the Zodiac Killer. The message was a 340-character cryptogram that the researchers cracked using a code-breaking computer program that tested more than 650,000 combinations. The message revealed that the Zodiac Killer admitted responsibility for the death of at least five people in Northern California. The team plans to use this method to crack two other messages from the Zodiac Killer.
Danske Bank, the largest bank in Denmark, is using AI to investigate cases of money laundering. In 2018, the bank was hit with a €200 billion money laundering scheme through its Estonian branch. To ensure this does not happen again, the bank has partnered with Quantexa, a data analytics company in London, to use its AI technology to detect suspicious financial activity in high-risk areas such as foreign currency exchange, securities, and equities.
Researchers at the Massachusetts Institute of Technology have created a machine learning model that shows how quarantine measures impact COVID-19 rates. Using neural networks, the model identifies patterns in the number of infections and recoveries within a geographic region. Most importantly, the model quantifies the number of infected individuals who are not spreading the virus of others. This reflects how successful a region is in quarantining an infected individual. Public health officials have applied the model to COVID-19 data from 70 countries and now are applying it to 13 U.S. states, enabling them to understand if their states need to amend quarantine measures, such as stay-at-home orders.
Clinicians at the Center for Cancer and Blood Disorders in Texas are using AI to better manage the mental health of cancer patients. Researchers have developed an AI tool that stratifies the health risk of patients into different categories, such as whether their physical condition would deteriorate in the next six months or whether they are more likely to develop depression. By stratifying patients into risk categories, clinicians are able to use the resources available to them to better manage patient mental healthcare which is important in improving outcomes.
Researchers from the Spin Convergence Research Center and Kyung Hee University in South Korea have developed an AI system to identify the magnetic properties of semiconductor materials. The researchers trained the system using images of magnetic regions in semiconductors to identify the properties of magnetic materials, such as thermal stability and dynamic behavior. When tested, the system achieved a 99 percent accuracy rate in estimating the properties of magnetic materials.
Doctors at the University of Vermont have partnered with Biocogniv, an AI-diagnostics tool company, to detect the likelihood a patient has COVID-19 from routine blood tests using machine learning. The algorithm analyzes changes in these routine tests and assigns a probability of the patient having COVID-19. The team trained the tool using blood work data from 2,813 patients confirmed to have COVID-19. When tested, the algorithm assigned probabilities correctly 91 percent of the time.
Researchers at the University of Rhode Island have used AI to identify, classify, and distinguish Anopheles, a mosquito species that carries malaria. The researchers trained neural networks to identify the sex and species of Anopheles mosquitoes based on 1,700 images of adult mosquitoes from five geographic regions. When tested, the networks predicted mosquito species with 99 percent accuracy and the sex of mosquitoes with 98 percent accuracy.
Google has open-sourced the first AI model that supports interoperation between 16 Indic languages, including Bengali, English, Punjabi, and Urdu. The model better understands the context of statements made in local languages, such as whether the sentence connotes a positive or negative sentiment, and also supports transliteration detection for users using the English keyboard to type in their native language.
Image credit: Oleksandr Pyrohov