NYU Data Science newsletter – September 13, 2016

GROUP CURATION: N/A

Sometimes AI techniques get it right, and sometimes they get it wrong. Only rarely will those errors be seen by the public: like the Vietnam war photograph, or when a AI ‘beauty contest’ held this month was called out for being racist for selecting white women as the winners. We can dismiss this latter case as a problem of training data?—?they simply need a more diverse selection of faces to train their algorithm with, and now that 600,000 people have sent in their selfies, they certainly have better means to do so. But while a beauty contest might seem like a bad joke, or just a really good trick to get people to give up their photos to build a large training data set, it points to a much bigger set of problems. AI and decision-support systems are reaching into everyday life: determining who will be on a predictive policing ‘heat list’, who will be hired or promoted, which students will be recruited to universities, or seeking to predict at birth who will become a criminal by the age of 18. So the stakes are high.

Scholars have debated whether social media platforms, by allowing users to select the information to which they are exposed, may lead people to isolate themselves from viewpoints with which they disagree, thereby serving as political “echo chambers.” We investigate hypotheses concerning the circumstances under which Twitter users who communicate about elections would engage with (a) supportive, (b) oppositional, and (c) mixed political networks. Based on online surveys of representative samples of Italian and German individuals who posted at least one Twitter message about elections in 2013, we find substantial differences in the extent to which social media facilitates exposure to similar versus dissimilar political views. Our results suggest that exposure to supportive, oppositional, or mixed political networks on social media can be explained by broader patterns of political conversation (i.e., structure of offline networks) and specific habits in the political use of social media (i.e., the intensity of political discussion). These findings suggest that disagreement persists on social media even when ideological homophily is the modal outcome, and that scholars should pay more attention to specific situational and dispositional factors when evaluating the implications of social media for political communication.

What’s next in tech? Which areas will produce the Googles and Facebooks of the next decade? … One alternative seems to be “frontier tech”: a seemingly heterogeneous group that includes artificial intelligence, the Internet of Things, augmented reality, virtual reality, drones, robotics, autonomous vehicles, space, genomics, neuroscience, and perhaps the blockchain, depending on who you ask.

A strategic partnership between the University of Michigan and software company Yottabyte promises to unleash a new wave of data-intensive research by providing a flexible computing cloud for complex computational analyses of sensitive and restricted data.

The Yottabyte Research Cloud will provide scientists high performance, secure and flexible computing environments that enable the analysis of sensitive data sets restricted by federal privacy laws, proprietary access agreements, or confidentiality requirements.

Sailors, pilots, military service men and women deployed around the world, and government officials who make national security decisions all rely on the National Geospatial-Intelligence Agency (NGA) to provide them with timely geospatial information that is critical for planning and decision-making. The University of Missouri College of Engineering has just been awarded a five-year, $12 million contract to deliver a comprehensive data science education program that will provide cutting-edge analytical training for the NGA workforce and potentially other members of the U.S. Intelligence Community (IC). This new program will address key education and training needs identified by NGA.

Johns Hopkins Engineering for Professionals has launched a new master’s degree program in data science that students can complete online.

More universities gearing up for data science:

Marquette picks architects to design athletic research center (August 30, BizTimes Media, Ben Stanley)

U-M, Yottabyte partner to accelerate data-intensive research (September 08, University of Michigan, Michigan Institute for Data Science)

$12 Million Federal Contract to MU Will Establish Education Program for National Intelligence Agency (August 31, University of Missouri, News Bureau)

A little less than two months ago, the Mozilla Science Lab closed its second-annual call for fellows. During that time we processed an impressive number of applications — from a record of 483 submissions –, conducted de-biased blind reviews, scheduled two- rounds of interviews and follow-ups, and evaluated a solid set of around 138 top candidates according to their engagement with open science, their enthusiasm for learning by teaching, and their commitment to a free and open web.

With generous support from the Leona M. and Harry B. Helmsley Charitable Trust, we fund the four chosen fellows to build open science, open access, and open source practice throughout their networks.

As digital applications encroach on various aspects of daily life, the impact on the economy will help us live smarter and better

Amid a proliferation of research into the factors and components of systemic risk come some novel approaches from the likes of Moody’s Analytics, Santa Fe Institute and Thomson Reuters.

School of Public Health Assistant Professor Julian Wolfson was named an associate editor for reproducibility for the Journal of the American Statistical Association (JASA). The appointment is in support of the journal’s new requirement for authors to submit scientific code and data for review along with their papers.

The journal said it’s adding the editors and requirement to ensure the reproducibility of scientific results reported in its studies. The move follows similar action recently taken by many medical journals.

Most real-world social networks are inherently dynamic, composed of communities that are constantly changing in membership. To track these evolving communities, we need dynamic community detection techniques. This article evaluates the performance of a set of game theoretic approaches for identifying communities in dynamic networks. Our method, D-GT (Dynamic Game Theoretic community detection), models each network node as a rational agent who periodically plays a community membership game with its neighbors. During game play, nodes seek to maximize their local utility by joining or leaving the communities of network neighbors. The community structure emerges after the game reaches a Nash equilibrium. Compared to the benchmark community detection methods, D-GT more accurately predicts the number of communities and finds community assignments with a higher normalized mutual information, while retaining a good modularity.

Google calls it Project Sand Hill.

Since 2012, Suman Prasad and his team have worked with various Silicon Valley venture capital firms to identify “rocketship” startups before they really take off, and they help plug them into the Google machine. They help them build apps for Android phones, hook into Android Pay, and make use of countless other Google services, from Google Maps to Google ads. Prasad started the project in his Google “20 Percent Time,” but it has since grown into something much bigger. He’s now director of startups and VC partnerships, and at any given time, Project Sand Hill now serves a good 100 US startups, plus about 30 abroad, including places like Israel, India, and China.

Google also runs its own venture capital arm, GV, but it wants another way of tracking the ever-changing tech landscape—and keeping its increasingly enormous company at the forefront of innovation. “The speed at which startups go from being a small startup to becoming a material company was accelerating,” Prasad says. “We wanted to partner up with companies before they came up with the next big thing.”

I was really impressed with the demo of IBM WatsonPaths. IBM scientists have trained their system to interact with medical domain experts in a way that’s natural for them, enabling the user to more easily understand the structured and unstructured data sources the system consulted and the path it took in offering an option.

Since these kind of tools fascinate me a lot, I have attempted to build a fully web-based and open-source based version of a “almost similar” Cognitive Computing system. I used technologies such as scikit-learn, gensim, MLib, NLTK, numpy, D3js, bootstrap, flask and nginx. It took me around 24 hours to develop the system from scratch. At the moment, this is just an experiment.

campaign finance filings,
Google search trends,
vote activity from sitting members of congress,
new polls,
forecasts from 538,
Cook Political Report race ratings

Key features:
Bookmarkable state;
Notifications on client browser;
Progress indicators;
Modal windows;
Documentation for connecting to external dB.

Gathering petabytes of data about your customers is cool, but how can you take advantage of this data? BlazingDB lets you run high-performance SQL on a database using a ton of GPUs. The company is introducing a free community edition of its solution on stage at TechCrunch Disrupt SF in our Battlefield competition.