US Homeland Security installs AI cameras at the White House, Google tries to make translation less sexist

Plus: European AI researchers to create a new lab

face_id_malware

Roundup Hello, welcome to this week's AI roundup.

This week we have Google trying to fix its gender bias problem by offering both male and female pronouns for its translation service, but it dismisses gender neutral ones. The US Secret Service is testing facial recognition software to track "subjects of interest," but the ACLU has concerns.

Gender bias in Google Translation: Google is attempting to make its translation service less sexist by including both masculine and feminine translations for the same sentence.

Previous research showed that Google was more likely to assign male pronouns to translations from sentences in languages with gender neutral pronouns - especially if the subject is related to fields that are traditionally more male dominated, like science and engineering.

A sentence such as “she is a surgeon” was translated to a language with no gendered pronouns such as Hungarian, and translated back to one with gendered pronouns like English, only to get back “he is a surgeon”.

So, what is Google doing to fix that?

“Now you’ll get both a feminine and masculine translation for a single word—like “surgeon” — when translating from English into French, Italian, Portuguese or Spanish. You’ll also get both translations when translating phrases and sentences from Turkish to English.

For example, if you type “o bir doktor” in Turkish, you’ll now get “she is a doctor” and “he is a doctor” as the gender-specific translations,” it announced.

Still, not everyone is happy about the move, as it forces the use of gendered pronouns and ignores non-gendered ones that others might prefer such as it or they.

Sexism in Google Translation is from biased training data. As the system is trained on millions of text scraped from the web, these sentences carry the historical and social biases humans have over time.

ELLIS: A new European AI Lab: A team of European AI experts are working together to support a new European research hub nicknamed ELLIS.

ELLIS, which stands for European Laboratory for Learning and Intelligent Systems, was previously proposed back in April to create new jobs and encourage academic research. USA and China are frequently seen as world leaders in AI and European countries want to remain competitive.

Fast forward eight months, and it looks like the plan is slowly coming into fruition. The formation of the lab was announced at the NeurIPS (Neural Information Processing Systems) conference in Montreal.

A team of AI researchers, including: Nicolò Cesa-Bianchi, Zoubin Ghahramani, Sepp Hochreiter, Cordelia Schmid, Jürgen Schmidhuber, Bernhard Schölkopf, Max Welling and many others will help lead efforts.

ELLIS will focus on modern machine learning methods such as deep learning and neural networks. “The comprehensive plan for ELLIS includes the creation of a network to advance breakthroughs in AI, a pan-European PhD program to educate the next generation of AI researchers, and a focal point for industrial engagements to boost economic growth by leveraging AI technologies,” according to a statement.

Facial recognition is coming to the White House: The Department of Homeland Security is testing the use of facial recognition to track people walking in and around the White House.

A document published by the Department of Homeland Security (DHS) reveals that the system will be operated by the US Secret Service (USSS).

“Ultimately, the goal of the FRP [Facial Recognition Pilot] is to identify if facial recognition technologies can be of assistance to the USSS in identifying known subjects of interest prior to initial contact with law enforcement at the White House Complex," it says.

At the moment, the FRP is restricted to only identifying staff that have volunteered to test the technology. Cameras have been set up in two locations that will capture images of people walking along the public streets and parks next to the White House.

The images of the faces captured in the video feed will be matched against a database containing the faces of the volunteers. Only pictures that correspond to a positive match will be kept, negative matches will be deleted automatically. A match will send an alert to the USSS, where it will be confirmed by staff.

DHS said it would be transparent about its process and has provided a public notice and written notice to volunteers. After the tests have been carried out, the faces kept on the database will be deleted.

The American Civil Liberties Union (ACLU)is concerned over facial recognition technology, and called it “one of the most dangerous biometrics from a privacy standpoint because it can so easily be expanded and abused”.

“We don’t exactly know how the Secret Service determines if someone is a “subject of interest,” it said in a statement. The technology could be used to identify protestors or for surveillance purposes.

“Exactly how wide a radius does the Secret Service want to monitor? Is there any reason to think it wouldn’t want to follow its “subjects of interest” 24/7 and nationwide if technology makes that easy enough?,” the ACLU asked.

“Whatever the answer is today, there is good reason to be concerned about what that answer might be in the future — especially if unregulated face recognition and other technologies make it cheap and easy to extend the tentacles of its surveillance outwards.” ®

Sponsored: Technical Overview: Exasol Peek Under the Hood

SUBSCRIBE TO OUR WEEKLY TECH NEWSLETTER




Biting the hand that feeds IT © 1998–2019