Inside | Real news, curated by real humans
Inside AI

Inside AI (Oct 31st, 2018)

1. Oxford University researchers developed an AI system that mimics human society to better understand what causes sectarian violence. The model has thousands of agents representing various ethnicities, races, and religions. The program is being trialed by Norway and Slovakia to study the tensions that exist when Muslim immigrants settle in predominantly Christian areas. The research has been published in the Journal for Artificial Societies and Social Simulation (JASS). — BBC

  • Email gray
  • Permalink gray

2. Public Knowledge, a public interest group, is calling for a new federal authority to oversee how regulators and agencies approach the use of AI technology. The group released a white paper today outlining five challenges that make a federal authority necessary. According to Public Knowledge, the new government body is needed to advise regulators in various sectors about common issues (transparency, bias, privacy), establish best practices for implementing the technology, attract AI experts as consultants, identify policy issues, and coordinate international AI strategies. Creating such an authority would require Congressional legislation. — AXIOS

  • Email gray
  • Permalink gray

It’s Hacktoberfest! Join DigitalOcean, GitHub and Twilio in celebrating open source and earn limited-edition swag. Learn more.

3. Microsoft announced two new AI training programs at its Future Decoded event today. The first program, Microsoft AI Academy, offers training sessions to help professionals develop practical AI skills. Microsoft will also be using the academy to train its own staff. The second program is the Microsoft Research-Cambridge University Machine Learning Initiative, investing in the next generation of data scientists and machine learning engineers. The initiative is meant to address the academic needs caused by the "brain drain" as researchers leave universities for private sector jobs in the AI tech industry. — ZDNET

  • Email gray
  • Permalink gray

4. A new AI lie detector is being piloted at EU border crossing points in Hungary, Latvia, and Greece. The lie detector is part of the iBorderCtrl technology and includes a pre-screening step before the traveler arrives at the border. The pre-screening uses an online application that uploads passport photos, visas, and proof of funds, and includes a computer animated border guard to ask the traveler questions via a webcam and analyze the micro-expressions to determine if the traveler is lying. The pre-screening will flag high-risk passengers before they cross the border and meet actual border guards. The six-month pilot test begins this month and is being coordinated by Hungarian National Police. — EUROPEAN COMMISSION

  • Email gray
  • Permalink gray

4 ways to improve collaboration on your team today

Our collaboration eBook shows you the four things your team can do right now to improve the way they work together and gives examples of what strong collaborative cultures across industries have in common. You’ll also learn how Dropbox Business can power your team’s best work.

Download now

5. Researchers from Guelph University and the University of Toronto Mississauga collaborated on a machine learning system that proved it is possible for fruit flies (Drosophila melanogaster) to visually recognize and identify another fly, despite their limited optics and their seemingly identical appearance. The team modeled a Deep Convolutional Network (DCN) based on the D. melanogaster biology, with the same theoretical visual input and processing ability, and the system was able to reliably identify a specific fly. The project was funded by a CIFAR Catalyst grant and the research is published in the PLOS One journal. — CIFAR NEWS

6. According to a survey of 2,000 Americans conducted by Interactions and The Harris Poll, most people overlook the benefits of AI because of the "creepiness" factor. The four main things about the technology that make respondents uncomfortable: human-sounding AI that doesn't identify itself as a machine (70 percent of respondents), AI knowing other household members' past interactions with a company (52 percent), AI using social media data to make suggestions (50 percent), AI knowing purchase history from a different company (42 percent). — TECH REPUBLIC

7. Researchers from the University of Western Australia repurposed an AI program used for facial recognition on social media to spot galaxies in deep space. The new model, called ClaRAN, scans radio telescope images to spot the radio jets emitted by supermassive black holes (SMBH) at the center of most large galaxies. Like the open-sourced system on which it was based, ClaRAN is also publicly available on Github. — GEEK.COM

8. The American College of Radiology Data Science Institute (ACR DSI) released a series of standardized use cases for AI in medical imaging. The structured use cases include flowcharts that describe how the algorithms work, parameters for training, testing, and validation for regulatory approval and clinical use, and instructions about how the technology is deployed and monitored for effectiveness. The tools are meant to advance the adoption of AI in clinical practice. — HEALTH IMAGING

9. MIT researchers developed a Halloween-themed AI system for generating scary music called the Uncanny Musicbox. The same team of researchers built an AI to create scary images in 2016, and one to produce scary stories in 2017. Uncanny Musicbox was trained on a large dataset of midi sound files and several horror movie soundtracks. — MOTHERBOARD

10. Computer vision researcher Janelle Shane's latest project is a machine learning algorithm to generate bizarre Halloween costume ideas. The algorithm was trained on a dataset of 7,182 costumes and the results were paired with illustrations by artist Jessia Ma. The project is featured in The New York Times. — FAST COMPANY

  • Email gray
  • Permalink gray

In yesterday's edition of Inside AI, we incorrectly identified the EPFL machine learning program as "SwiftML." The program is called ShiftML. Inside AI regrets the error. 

  • Email gray
  • Permalink gray

Written and curated by Deb Dion Kees, a writer, editor, and publisher based in Telluride, Colorado. Kees is a lover of science, technology, skiing, and adventure, and does her best work using a mobile hotspot to write from her Ford camper van office.

Editing team: Lon Harris (editor-in-chief at Inside.com, game-master at Screen Junkies), Krystle Vermes (Breaking news editor at Inside, B2B marketing news reporter, host of the "All Day Paranormal" podcast), and Susmita Baral (editor at Inside, recent bylines in NatGeo, Teen Vogue, and Quartz. Runs the biggest mac and cheese account on Instagram).

  • Email gray
  • Permalink gray

Subscribe to Inside AI

MORE NEWSLETTERS

David Strom's in-depth cybersecurity news and analysis

Inside Security

David Strom's in-depth cybersecurity news and analysis

DAILY
David Strom's in-depth cybersecurity news and analysis

Inside Security

DAILY

SUBSCRIBED!

Share via

The present and future of virtual/augmented reality news and technology

Inside VR & AR

The present and future of virtual/augmented reality news and technology

DAILY
The present and future of virtual/augmented reality news and technology

Inside VR & AR

DAILY

SUBSCRIBED!

Share via

Fascinating, curious and amazing journalism, all in one link.

ReadThisThing

Fascinating, curious and amazing journalism, all in one link.

DAILY
Fascinating, curious and amazing journalism, all in one link.

ReadThisThing

DAILY

SUBSCRIBED!

Share via

Financial, legislative, agricultural, and all the other most important news about the cannabis industry

Inside Cannabis

Financial, legislative, agricultural, and all the other most important news about the cannabis industry

DAILY
Financial, legislative, agricultural, and all the other most important news about the cannabis industry

Inside Cannabis

DAILY

SUBSCRIBED!

Share via