TensorFlow gets its quantum of solace, lid lifted on 'all-seeing crime-detecting' AI upstart, and more
Plus: Machine-learning software scans ancient texts
Roundup Here's a handy little roundup of all the bits of AI news that you may have missed.
Uh oh, another surveillance company has secretly been purloining data from social media: Banjo, the AI startup that believes its software can detect and surface crimes and other activities in real-time from all kinds of data feeds, also scraped information from people’s public social media profiles.
However, it wasn’t as brazen as Clearview, the controversial upstart known for downloading over three billion photos from Facebook, Instagram, YouTube, Twitter, and more to put together a massive dataset for facial recognition. Banjo apparently created a shadow company called Pink Unicorn Labs, according to Vice.
Pink Unicorn Labs went on to develop three apps directed at fans of things like the British boyband One Direction, EDM music, and Formula One racing. These apps asked users to connect and sign-in using their accounts on social media platforms like Facebook, Twitter, Instagram, Google Plus, FourSquare, as well as VK and Sina Weibo, commonly used in Russia and China. Linking the Pink Unicorn Labs Apps to people’s accounts makes it possible to scrape those netizens' data, such as images or location history.
Code across all the three apps contained links to Banjo’s website. Both companies were registered at the same address in Redwood City, California and headed by Banjo’s CEO Damien Patton.
Pink Unicorn Labs’ apps were removed from the Google Play Store in 2016. Even though data might be publicly posted on people’s accounts, scraping them to use for commercial purposes is against the terms of service of these platforms.
AI helps historians read messages carved on ancient bones: Researchers from Southwest University in China used a convolutional neural network to classify and read ancient scripts carved on bones dating back to more than 3,000 years between 1600 to 1046 BC.
The Chinese characters written in Yi script, the oldest examples show it was used in the Middle Kingdom from the 15th century. Studying these ancient texts is difficult; not only does it require extensive knowledge of the language and its history, but the messages imprinted on these bones are cracked and worn out over time.
Here’s where the machine learning bit comes in. A convolutional neural network was trained on images of these texts where each character was labelled so it could recognize scripts carved on other types of bones, according to a paper published in IEEE Computer Graphics and Applications.
“The researchers used a dataset consisting of 1,476 tortoise shell rubbings and 300 ox bone rubbings, from which they chose one-third as the test set and two-thirds as the training set. Experiment results show the proposed method reaches a level close to that of oracle experts,” Synced explained this week.
“As I said, classification is the first step,”Shanxiong Chen, first author of the paper and an associate professor of computer and information science, told Synched.
“This study specifically focused on telling between animal bones and tortoise shells, and we’re continuously working with Capital Normal University’s Center for Oracle Bone Studies on further classifying different types of animal bones.”
ICLR 2020 goes virtual: Tech conferences are dropping like flies amidst the current outbreak of the coronavirus. Now, the International Conference on Learning Representations (ICLR), a top academic machine learning conference, has decided to cancel its physical event due to take place in Addis Ababa, Ethiopia, next month.
“Due to growing concerns about COVID-19, ICLR2020 will cancel its physical conference this year, instead shifting to a fully virtual conference,” it announced this week. “We were very excited to hold ICLR in Addis Ababa, and it is disappointing that we will not all be able to come together in person in April.”
Organisers have called all academics with accepted papers to create a five minute video as presenting their work part of its virtual poster session. For those that were invited to give a talk, that video will be extended to 15 minutes and information should be conveyed in a series of slides. Workshops are a little trickier to put together; ICLR is currently contacting speakers to coordinate.
All registration fees and travel purchased for the conference will be reimbursed. Now, the price to attend the digital conference has dropped down to $50 for students and $100 for non-students.
New TensorFlow library! If you’re bored at home and ‘social distancing’ from all your friends, family, and colleagues then try this: TensorFlow’s latest library that allows you to build quantum AI models.
Your brain will probably turn to mush trying to understand and combine both quantum computing and machine learning. The library known as TensorFlow Quantum (TFQ) was built by folks over at Google, the University of Waterloo, X, and Volkswagen, to give developers tools to process data that could, theoretically, run on quantum computers.
“We announce the release of TensorFlow Quantum (TFQ), an open-source library for the rapid prototyping of quantum ML models,” the Chocolate Factory said this week. “TFQ provides the tools necessary for bringing the quantum computing and machine learning research communities together to control and model natural or artificial quantum systems.” ®