Inside AI - January 10th, 2020

Inside AI (Jan 10th, 2020)

San Diego facial recognition program / How to spot deepfakes / Gauge CNNs detect patterns on curved surfaces

Subscribe | View in browser

1. Current and future AI restrictions imposed by the U.S. on things like exports could threaten China-based research lab outposts established by American companies such as Microsoft and IBM, Wired reports. These include the U.S. Commerce Department's new restrictions, which took effect Monday, on U.S. exports of AI software related to geospatial imagery. They were the first requirements to take effect under the 2018 Export Control Reform Act, or ECRA, which focuses on AI and other emerging technologies that the U.S. government believes are essential to national security. According to Wired, the new restrictions "are unlikely to be the last to land on American AI algorithms, datasets, and chips," as the Commerce Department continues to mull over additional controls of AI software. As a result, it "will become more difficult for American companies to maintain labs" in talent-rich China, Wired reports. - WIRED

2. The now-shuttered facial recognition program previously used by San Diego police has not been linked to a single arrest, according to Fast Company, which described the initiative as "a flop." San Diego police agencies launched the program seven years ago and began compiling a database that eventually collected more than 65,000 face scans. However, the program was discontinued after California's legislature approved and enacted a three-year ban on police use of mobile facial recognition technology. Because law enforcement agencies didn’t actually track any results, it's unclear if it had much of an effect, if at all. A San Diego Police Department spokesperson confirmed that officers could not identify a single case in which someone was arrested through the use of facial recognition technology, which continues to raise concerns about surveillance, privacy, and inherent biases that could impact how it's deployed in the real world. - BUSINESS INSIDER

3. With the number of deepfakes expected to grow ahead of the 2020 presidential election, USA Today's Jessica Guynn interviews some experts on how to spot the AI-edited fake videos, images, and audio. Siwei Lyu, a computer science professor at the State University of New York at Albany and a member of the Deepfake Detection Challenge’s advisory group, recommends that people stay on the alert, particularly when a video appears "bizarre or exceptional." Some clear giveaways include if the video is low-resolution or grainy, relatively short (30-60 seconds), and discolored or blurry. Subbarao Kambhampati, a computer science professor at Arizona State University, says the subject in a deepfake could have "different-sized eyes or ill-formed teeth, or more than two eyes," along with general "inconsistencies in the background of the video." Another indication could be if the audio is not synced with the video or images. - USA TODAY

4. Researchers at the University of Amsterdam and Qualcomm AI Research developed so-called “gauge-equivariant convolutional neural networks,” or gauge CNNs, that can detect patterns on any kind of geometric surface, including spheres and asymmetrically curved objects. The new theoretical framework for building neural networks was outlined in a research paper published on In essence, the idea is to allow computers to detect patterns and other features in curved and higher-dimensional space. The goal of "lifting CNNs out of flatland" emerged in 2016 through the process known as geometric deep learning. The newer research into gauge CNNs "is a fairly definitive answer to this problem of deep learning on curved surfaces,” said Max Welling, who developed the framework along with Taco Cohen, Maurice Weiler, and Berkay Kicanaoglu. - QUANTA MAGAZINE

5. France-based medical tech startup Cardiologs has raised $15 million for its AI system that helps detect heart conditions, mostly the abnormal heart rhythm known as atrial fibrillation. The company has worked since 2014 to develop a large database of electrocardiography (ECG) recordings that track heart activity. Its deep learning system can analyze new recordings to find patterns and potentially diagnose patients faster, VentureBeat reports. The round was led by the Paris-based VC firm Alven, with participation from Bpifrance, Idinvest Partners, Kurma Diagnostics, ISAI, and Paris Saclay Seed Fund. - VENTUREBEAT

6. A documentary titled "In the Age of AI" is coming to PBS Distribution later this month. The “Frontline” doc, due out on DVD and digitally on Jan. 14, explores current and future ways AI can impact the world. An online description says the film "traces the battle between the U.S. and China to harness its power, examining fears about what AI advances mean for the future of work and revealing how AI algorithms are ushering in an age of both great problem-solving potential and of new and troubling threats to privacy and democracy." - MEDIA PLAY NEWS

7. Can AI help society become fairer? On his "Connections" program, host Evan Dawson explores this question through interviews with several experts, who discuss the idea that AI could actually exacerbate inequality by mainly serving society's most powerful. His guests include Matt Kelly, an independent journalist; Ehsan Hoque, Asaro-Biggar Family Assistant Professor of Computer Science at the University of Rochester; Jonathan Herington, lecturer in the Department of Philosophy, and assistant director of graduate education in the College of Arts, Sciences, and Engineering at the University of Rochester; and Hadi Hosseini, assistant professor in the Department of Computer Science at RIT. - WXXI

Written and curated by Beth Duckett, a former reporter for The Arizona Republic who wrote a book about the solar industry and frequently covers hobby and commercial drones. You can follow her tweets here.

Edited by Sheena Vasani, staff writer at Inside Dev.

Copyright © 2020, All rights reserved.

Our mailing address is:
767 Bryant St. #203
San Francisco, CA 94107

Did someone forward this email to you? Head over to to get your very own free subscription!

You received this email because you subscribed to Inside AI. Click here to unsubscribe from Inside AI list or manage your subscriptions.

Subscribe to Inside AI