Digital texts have become an increasingly important source of data for social studies. However, textual data from open platforms are vulnerable to manipulation (e.g., censorship and information inflation), often leading to bias in subsequent empirical analysis. This paper investigates the problem of data distortion in text classification when controlling type I error (a relevant textual message is classified as irrelevant) is the priority. The default classical classification paradigm that minimizes the overall classification error can yield an undesirably large type I error, and data distortion exacerbates this situation. As a solution, we propose the Neyman-Pearson (NP) classification paradigm which minimizes type II error under a user-specified type I error constraint. Theoretically, we show that while the classical oracle (i.e., optimal classifier) cannot be recovered under unknown data distortion even if one has the entire post-distortion population, the NP oracle is unaffected by data distortion and can be recovered under the same condition. Empirically, we illustrate the advantage of NP classification methods in a case study that classifies posts about strikes and corruption published on a leading Chinese blogging platform.

As artificial intelligence (AI) technologies and platforms become integral to advanced operations in nearly every industry, blockchain is inserting itself as a means to enhance AI applications in both form and function. Blockchain has the potential to allow AI technologies to become more collaborative in nature and therefore increase their operating efficiency. Additionally, the potential for bolstered revenue streams is also apparent, as blockchain is projected to grow to $20 billion by 2024 according to Transparency Market Research and the Grand View Research projects the AI market will be worth more than $35 billion by 2025. As previously noted, leaders in the AI landscape are turning to blockchain to finetune various applications. Active tech companies in the markets this week include Gopher Protocol Inc. (OTC:GOPH), Overstock.com

Budding authors face a minefield when it comes to publishing their work. For a large fee, as much as $3,000, they can make their work available to anyone who wants to read it. Or they can avoid the fee and have readers pay the publisher instead. Often it is libraries that foot this bill through expensive annual subscriptions. This is not the lot of wannabe fiction writers, it's the business of academic publishing.

JX Press has designed a newsroom staffed by engineers and artificial intelligence instead of employing journalists. According to Bloomberg, once the Japenese media firm finds news, it uses algorithms to create stories and filters out 99% of fake news stories. The firm has developed a tool called NewsDigest – a free mobile news app which generates the firm's advertising revenue, as well as filters fake news and finds breaking stories on social media programmes. Fast Alert is another one of its products which analyses social media posts and photos to find Japan's breaking news and reports major international developments that it deems as trustworthy. The team consists of 24 people, of which 17 are engineers and the others are in relevant business functions, not journalists or reporters.

Google will not renew a contract to do artificial intelligence work for the US Pentagon, company sources say. The decision follows strong opposition in the technology giant's workforce. A number of Google employees resigned and thousands of others signed a protest petition against taking part in the Pentagon project, known as Maven. They feared it was the first step towards using artificial intelligence for lethal purposes. There has been no official statement from Google.

God be with the days when AI meant a fella coming to get the cows pregnant. It can make a medical diagnosis or compose music or play chess. If artificial intelligence was a TV character it would be a mix of Sheldon from The Big Bang Theory and a handsome doctor from Grey's Anatomy. New York City Council is currently going through a process to establish exactly how many algorithms are involved in the governance of the city: because they don't know how many there are. They do know that it's involved in the allocation of police officers, food stamps and public housing.

I recently read the "Learning to generate reviews and discovering sentiment" paper by OPENAI and found it to be super cool. But I could not understand how they are using the language model as feature extractor. Suppose we have 150 characters in a review, how do we extract features from these 150 characters when our input is 64 characters at a time.

Im finding it hard to find the information on this, but im sure its fine. The gtx 1050 ti is listed as having a compute capability of 6.1. Or if someone can point me to a list of supported GPUs for matconvnet that would be stellar!

The world of wine is vast, and it seems like there are more wineries out there than grapes in a vineyard. As such, finding that perfect bottle of vino can feel like a nightmare. And, while asking your friends for recommendations can help refine the process, you can't always rely on their tastes to match your own. So, why not let data do the matching for you? Services like Spotify and Netflix already use artificial intelligence to match their users with music and TV - content that can be just as nuanced as a bottle of Merlot - and they do so with great effect.

Shawn Moynihan is Editor-in-Chief of National Underwriter Property & Casualty. A St. John's University alum, Moynihan has earned 11 Jesse H. Neal Awards, the Pulitzers of the business press; seven Azbee Awards, from the American Society of Business Press Editors; two Folio Awards; and a SABEW award, from the Society of American Business Editors & Writers. Prior to joining ALM, he served as Managing Editor/Online Editor of journalism institution Editor & Publisher, the trade bible of the newspaper industry. Moynihan also has held editorial positions with AOL, Metro New York, and Newhouse Newspapers. He can be reached at smoynihan@alm.com.