Just a quick reminder that the full version of Inside AI is now free on Fridays only. To receive full issues six days a week, click here to upgrade to premium! We offer a $10/month plan, or the option to be billed $100 annually.

PS: If you have tips or suggestions on content that is valuable to you, hit reply. I’d love to hear from you. 

Thanks again,

Beth / @Beth_Duckett

The supercomputers have a combined 400 petaflops of computing power, which will speed up calculations and molecular modeling, potentially unearthing new molecules that could result in a vaccine. NVIDIA is the latest to join the consortium, providing the scientists with expertise in AI and access to the company's SaturnV supercomputer. Other members include IBM and the U.S. Department of Energy, which are co-leading the effort, along with Google, Microsoft, Amazon, Hewlett Packard Enterprise, NASA, the National Science Foundation, Pittsburgh Supercomputing Center and national labs including Lawrence Berkeley, Argonne, Lawrence Livermore, Los Alamos, Oak Ridge, and Sandia, among others. MIT, the University of Texas, Austin, Rensselaer Polytechnic Institute, and the University of California, San Diego, are also involved.

Read more: Dario Gil, the director of IBM Research, explains how the COVID-19 High Performance Computing Consortium came together in less than a week.

The resources could help researchers, scientists, and doctors form new insights about the coronavirus, aggregate data, and identify new candidates for drug molecules and treatments, the company said. The tools include:

• AI deep search tool, which encompasses 13,335 documents provided by research groups, the White House, and databases from the DrugBank, Clinicaltrials.gov and GenBank. The system, which was trained on papers in the COVID-19 Open Research Dataset (CORD-19), taps machine learning to extract relevant information about the virus in queries.
• Functional Genomics Platform, which is free during the pandemic. The cloud-based repository contains information about proteins, genes, and other molecules, which could help identify molecular targets for drugs and treatments. Also, IBM has currently identified 3,000 small molecules as possible therapeutic candidates, which it has shared with researchers via an interactive molecular explorer tool.
• IBM Micromedex and DynaMed, which are free for clinicians and healthcare professionals to access the latest COVID-19 research and evidence-based drug and disease information.

Cambridge University researchers want audio of your coughing, talking, and breathing. This, they say, will help them develop new machine learning algorithms for their COVID-19 Sounds App, which is designed to automatically detect if a person has COVID-19 based on the sound of their voice. The app is available on Chrome and Firefox (with Android and iOS versions coming soon).

To lend a voice, volunteers must fill out a survey answering questions about their medical history and demographics. After that, they are asked to provide some voice samples (based on text read from the screen) and several seconds of coughing via their phone’s microphone. Users are also asked to share their locations, which the university will store in servers for research purposes only. While it's not a replacement for actual COVID-19 testing, it could be useful for people who have no access to diagnostic tools otherwise.

The spike proteins, which are made up of amino acids, poke out of the virus itself and bind to human cells, spreading the infection. To create a song based on their structure, MIT researchers gave each amino acid a different note and used an algorithm to convert the data into a musical score. “We would need a high-powered microscope to see the equivalent detail in an image, and we could never see it all at once," said MIT Professor Markus Buehler, who led the project. "Sound is such an elegant way to access the information stored in a protein.” On a pratical note, Buehler said scientists could use the song to search for a new protein matching the rhythm and melody of an antibody, which would bind to the spike and stop it from infecting.

Creative Technologist Matt Reed created his digital doppelgänger using several screenshots of himself sitting at his desk while peering into his webcam. Reed applied the images to a custom webapp he created, which is based on the open-source tool Artyom.js for voice command recognition and speech synthesis. Using ManyCam, he created a virtual webcam feed. His "Zoombot" appears during conference calls and responds to questions posed directly to Reed, with responses such as “I’m having trouble hearing you." Reed also posted the code on GitHub, with the title "A highly advanced AI to handle Zoom calls."

Related reading: A technologist at the creative agency redpepper, Reed first made headlines back in 2018 as one of the creators of a "Where's Waldo?" robot, which uses machine learning and a camera to spot the children's character.

According to Google researcher Jeffrey Dean, the number of machine-learning papers published on Arxiv has been growing exponentially over the last several years, surpassing even the growth rate of Moore's Law (doubling of transistors on a chip every two years). Susmelj shares tips for finding this research.

Caballero, who departed from Apple early last year, is now a corporate vice president at Washington-based Microsoft.

The app, called Impressions, is currently available for iOS, with an Android release planned for the near future.

The vehicle transports samples between the clinic’s drive-thru testing site and lab.