Amazon's facial recognition software incorrectly matched the faces of 28 U.S. legislators to images in a mugshot database. A test of Amazon's facial recognition software incorrectly matched the faces of 28 U.S. legislators to images in a mugshot database, with people of color misidentified disproportionately, according to the American Civil Liberties Union (ACLU). The organization assembled a face database and search tool from 25,000 public arrest photos, then cross-referenced that data with public photos of every member of Congress. Eleven of the misidentified lawmakers were people of color, representing nearly 40% of those wrongly matched, even though minorities comprise only 20% of those in Congress. Says the ACLU Foundation of Northern California's Jacob Snow, "Our test reinforces that face surveillance is not safe for government use."

With all of the buzz around regtech, it's easy to forget that banks have leveraged technology for compliance and reporting for decades. But thanks to recent developments in data architecture, artificial intelligence and more, regtech is on the rise, and it's evolving into something a lot more sophisticated. The definition of regtech is simple. According to New-York-based analytics firm CB Insights, regtech is "technology that addresses regulatory challenges and facilitates the delivery of compliance requirements." Regtech can be as simple as using an Excel spreadsheet for financial reporting or as complex as using adaptive algorithms to monitor markets.

Google is reportedly going to launch a censored version of its search engine in China. The tech giant has been secretly planning to launch the product since last year, as part of a project referred to inside the company as'Dragonfly,' according to The Intercept, which was given internal documents from a whistleblower. It comes as Google has tried and failed to make inroads in the Chinese market over the past several years. Google has been planning to launch the product since last year, as part of a project referred to inside the company as'Dragonfly.' While China is home to the world's largest number of internet users, a 2015 report by US think tank Freedom House found that the country had the most restrictive online use policies of 65 nations it studied, ranking below Iran and Syria.

In 2013, an algorithm determined Eric Loomis' six-year prison sentence in Wisconsin for attempting to flee a traffic officer and operating a motor vehicle without the owner's consent. No one knew how the software, Correctional Offender Management Profiling for Alternative Sanctions, or COMPAS, worked--not even the judge who delivered the sentence. Analyses conducted by ProPublica later found the predictive artificial intelligence used in this case, which attempts to gauge the likelihood of an offender committing another crime, to be racially biased: A two-year study involving 10,000 defendants found that the A.I. routinely overestimated the likelihood of recidivism among black defendants and underestimated it among whites. The U.S. Supreme Court declined to review Eric Loomis' case, so the sentence stands. Increasingly, A.I. has the power to alter the course of people's lives.

With increasing institutional use of Al and machine learning methods, more and more industry regulators are making use of such software to provide financial stability through services and systemic risk surveillance. Central to the UK Financial Conduct Authority's (FCA) agenda is their intent to develop "both their technology side – cloud analytics – as well as building out our human side – our data science capability" to allow an "osmotic" expansion across the wider organisation, "effectively enabling us to leverage machine learning," says head of regtech and advanced analytics at FCA, Nick Cook. Investigations of suspicious transactions are highly time-consuming, and often, due to overly-defensive mechanisms, fall victim of unsuccessful outcomes. Through the identification of intricate patterns, machine learning would flag up serious, rather than harmless transactions, and, consequently require the attention of supervisors. For instance, with the aid of Al and machine learning, the Monetary Authority of Singapore (MAS) examines suspicious transactions that require further analysis and attention, which, in turn, enables supervisors to spend their time and resources on higher risk transactions.

SAN FRANCISCO -- Google may launch a censored version of its search engine in China in a move that would largely reverse its 2010 decision to withdraw from the country, The Intercept reported. The news outlet says Google has been working on a project called Dragonfly since last year and has demonstrated a version of the censored search engine to Chinese officials. The Android app could launch in six to nine months as part of a joint venture with an unnamed partner company, likely in China. It would automatically block websites blacklisted by Beijing as well as search terms on human rights, democracy, religion, and peaceful protest, The Intercept said. "We provide a number of mobile apps in China, such as Google Translate and Files Go, help Chinese developers, and have made significant investments in Chinese companies like JD.com. But we don't comment on speculation about future plans," Google said in a statement.

Why are so many AI researchers so worried about lethal autonomous weapons? What makes autonomous weapons so much worse than any other weapons we have today? And why is it so hard for countries to come to a consensus about autonomous weapons? Not surprisingly, the short answer is: it's complicated. In this month's podcast, Ariel spoke with experts from a variety of perspectives on the current status of LAWS, where we are headed, and the feasibility of banning these weapons. Guests include ex-Pentagon advisor Paul Scharre (3:40), artificial intelligence professor Toby Walsh (40:51), Article 36 founder Richard Moyes (53:30), Campaign to Stop Killer Robots founders Mary Wareham and Bonnie Docherty (1:03:38), and ethicist and co-founder of the International Committee for Robot Arms Control, Peter Asaro (1:32:39). You can listen to the podcast above, and read the full transcript below. You can check out previous podcasts on SoundCloud, iTunes, GooglePlay, and Stitcher. If you work with ...

The latest: Several nonprofits are among the latest members in the Partnership on AI, a group established to address the ethical and other challenges presented by artificial intelligence. The effort began with big companies like Apple, Google, Facebook and Microsoft, but now includes a growing roster of academic institutions and nonprofit groups alongside some of the biggest names in tech. LGBT rights organization GLAAD is among the new members, as is the Joint Center for Political and Economic Studies, which was set up in 1970 as a think tank serving black elected officials. Joint Center president Spencer Overton said his organization is especially concerned with how AI could disproportionately impact employment in communities of color. Overton said that 27% of black workers are concentrated in just 30 jobs at high risk to automation.

China's relationship with Google is fractious at best, but it's no secret that the search giant wants to make inroads in what is a largely untapped market. However, its latest alleged plan could send tech's political sphere into a tailspin. According to The Intercept, Google is working on a censored version of its search engine for the country -- one which will blacklist websites and search terms about human rights, democracy, religion and protest. The project, codenamed Dragonfly, has been in the works since spring 2017 and could be ready to launch within the next six to nine months, according to unnamed sources familiar with the plan. Apparently the project was given the go-ahead during a meeting between Google's CEO Sundar Pichai and a top Chinese government official, although it still needs final approval from China.

Note: This article will likely be revised and expanded before being submitted for review and publication. At the moment it is missing critical sections, that will be added later. If we have suggestions for improvement, please send them to me directly. In this article I will discuss the well-known technique of feature hashing, but with the modification of performing the hashing step on the client-side before sending data to a server or daemon performing model training and prediction. By using this approach, we can ensure that the system performing the training cannot have any knowledge of the underlying data being received, since the learning takes place only using the hashed representation of the data.