May 8, 2020
In COVID-related AI news, hospitals across the US are using an AI system called Deterioration Index to provide a snapshot of patients’s risks, even though the software has not yet been validated to be effective for those with COVID-19. Meanwhile, Qure.ai has retooled its qXR system, designed for chest x-rays, to detect COVID-induced pneumonia, and a preliminary validation study with 11,000 images found a 95% accuracy in distinguishing patients with and without COVID-19. The Digital Ethics Lab at University of Oxford has provided a set of ethical guidelines (16 yes/no questions) for those making COVID-19 Digital Tracking and Tracing (DTT) systems. And Carnegie Mellon provides five interactive maps for COVID-related issues in the US. The Joint AI Center unveils Salus, a prototype AI tool for examining where COVID-19 might impact logistics and supply chains. And Reuters spends time to debunk a false claim on the relation of AI to COVID-19. In regular AI news, Washington State passes major facial recognition legislation, defining how state and local government may use facial recognition. DARPA selects Georgia Tech and Intel to lead its Guaranteeing AI Robustness against Deception (GARD) program. And the Association for the Understanding of AI launches AIhub.org, to connect the public and AI community. In research, two German Institutes investigate the roles of different neurons in neural networks, and found populations that serve different functions; in addition, these populations could be extracted to a new network without having to train the new network on the same knowledge. Research from Bar-Ilan University demonstrate human brain learning mechanisms that outperform common AI learning algorithms, to include observing the same image 10 times in a second being more effective than observing the same imagine 1000 times in one month. The book of the week comes from Matthieu Thiboust, with Insights from the Brain, which aims to provide “neuroscience chunks of information related to AI.” And CBS News 60 Minutes has a report on BlueDot, the company that warned its clients about the COVID-19 outbreak a week before the CDC.