NYU Data Science newsletter – March 9, 2016

GROUP CURATION: N/A

Horizon Robotics, led by Yu Kai, Baidu’s former deep learning head, is developing AI chips and software to mimic how the human brain solves abstract tasks, such as voice and image recognition. The company believes that this will provide more consistent and reliable services than cloud based systems.

… Onalytica has brought insights into this field by reviewing experts and influencers in Artificial Intelligence and Machine Learning. They have used their models and methods to map the community of thought leaders in social media, extracting those showing the greatest amount of influence on the topics.

Onalytica analyzed 1.1M+ tweets from November 30, 2015 through February 24, 2016 mentioning keywords associated with “Artificial Intelligence” and identified the top 100 most influential brands and individuals leading the discussion on Twitter.

Look at these blocks. Do you think they’re going to fall?

That simple question drives a new experiment from Facebook’s artificial intelligence research lab – an attempt to create software that can observe a simple version of the world and predict what will happen.

I recently spoke to Finn Brunton of NYU about his book Spam: A Shadow History of the Internet for the New Books in Media & Communications podcast. … Finn Brunton’s Spam: A Shadow History of the Internet (MIT Press, 2013) is a cultural history of those communications that seek to capture our attention for the purposes of exploiting it. From pranks on early computer networks in the 1970s to commercial nuisances in the 1990s to the global criminal infrastructure of today driven by botnets and algorithms, spam’s history surfaces and shifts with the Internet itself.

In Massachusetts, big data is big business.

A new report finds that 53 new big data companies have come to Massachusetts since 2014, bringing the number to 537 — a jump of about 10 percent.

The report by Mass Big Data, a project of the Massachusetts Technology Collaborative, also found that the number of schools offering data science certificate or degree programs rose from zero to five within the past three years: Harvard, Northeastern, UMass Dartmouth, Worcester Polytechnic Institute, and Becker College. UMass Amherst recently opened new center for data science, although it does not have a degree program in data science.

With the rapid pace of scientific advancement, it can be challenging for members of the public to stay informed. News and social media play an important role in translating new science for a general audience. But over-reliance upon second-, or even third- or fourth-hand reports on science can lead to misinformation being spread. As my elementary school librarian taught me, nothing beats reading a primary source.

Despite the fact that scientific research is increasingly being published in open-access journals, and even though science communications aimed at the public often contain links to original sources, I suspect that few members of the public actually read original research articles. And there is good reason for that. Articles published in peer-reviewed journals are targeted at specialists in the field.

… Over the past eighteen months, a team of researchers at a Google AI lab in London have worked to build an artificially intelligent system that can make this kind of leap, and AlphaGo has already shown its worth. In October, during a closed-door match, it beat the three-time European Go champion, Fan Hui. But now comes the bigger test. Today at the Four Seasons, AlphaGo begins a five-game, seven-day, one-million-dollar match against the Korea-born Lee Sedol, who has won more international Go titles than all but one other player. Google bills it as a battle with “the Roger Federer of the Go world.”

… Campaigns and officials worldwide now use bots for a multitude of tasks beyond simple social media account management. These uses range from the seemingly mundane—sending out messages on particular pieces of legislation—to the downright nefarious—spreading covert political propaganda.

Of all the theories to explain the unexpected success of Donald Trump’s presidential campaign, this, surely, is the most novel. Forget about a disaffected working class buffeted by globalisation and automation, pent-up racial resentments finding an outlet or the advent of the 24-hour news cycle. No: in the assessment of Daniel Drezner, a professor of international politics at Tufts University, it’s the political scientists who are to blame.

Also:

How politicians should use Twitter bots. (Slate, Tim Hwang and Samuel Woolley, March 8)

A man had just gone on a shooting rampage in Kalamazoo, Mich., allegedly killing six people while driving for Uber. Sherry Towers, an Arizona State University physicist who studies how viruses spread, worried while watching the news coverage.

Last year, Towers published a study using mathematical models to examine whether mass shootings, like viruses, are contagious. She identified a 13-day period after high-profile mass shootings when the chance of another spikes. Her findings are confirmed more frequently than she would like.

… You wouldn’t see it in most classrooms, you wouldn’t know it by looking at slumping national test-score averages, but a cadre of American teenagers are reaching world-class heights in math—more of them, more regularly, than ever before. The phenomenon extends well beyond the handful of hopefuls for the Math Olympiad. The students are being produced by a new pedagogical ecosystem—almost entirely extracurricular—that has developed online and in the country’s rich coastal cities and tech meccas. In these places, accelerated students are learning more and learning faster than they were 10 years ago—tackling more-complex material than many people in the advanced-math community had thought possible. “The bench of American teens who can do world-class math,” says Po-Shen Loh, the head coach of the U.S. team, “is significantly wider and stronger than it used to be.”

If you’re a data management enthusiast like me (yes, we exist, and there’s actually a bunch of us), you’ve probably head about Kristin Briney’s Book, Data Management for Researchers. I received a copy for review a few months ago, and have been taking my time to savor it. But if you’ve heard of this book, chances are that although you’ll certainly find aspects of it useful, you’re probably the metaphorical choir that we, the data managers, are preaching to. You might even argue that there are lots of data management resources out there- why a book? But Briney does something unique here, and I have been enthusiastic to recommend it to everyone around me.

The Guardian offers an API as deep and robust as the New York Times Article API when it comes to content analysis.

The Guardian’s API offers more than “1.7 million pieces of content”, with published items as far back as 1999. You can register as a developer here, which gets you 5,000 API hits a day and an API key.

Leaf is a Machine Intelligence Framework engineered by software developers, not scientists. It was inspired by the brilliant people behind TensorFlow, Torch, Caffe, Rust and numerous research papers and brings modularity, performance and portability to deep learning. Leaf is lean and tries to introduce minimal technical debt to your stack.

Leaf is a few months old, but thanks to its architecture and Rust already one of the fastest Machine Intelligence Frameworks in the world.

The business world is full of streams of items that need to be filtered or evaluated: parts on an assembly line, resumés in an application pile, emails in a delivery queue, transactions awaiting processing. Machine learning techniques are increasingly being used to make such processes more efficient: image processing to flag bad parts, text analysis to surface good candidates, spam filtering to sort email, fraud detection to lower transaction costs etc.

In this article, I show how you can take business factors into account when using machine learning to solve these kinds of problems with binary classifiers. Specifically, I show how the concept of expected utility from the field of economics maps onto the Receiver Operating Characteristic (ROC) space often used by machine learning practitioners to compare and evaluate models for binary classification. I begin with a parable illustrating the dangers of not taking such factors into account.