NYU Data Science newsletter – July 7, 2016

GROUP CURATION: N/A

Delivering on the promise of genomics is dependent upon three main factors, most of which are within the purview of digital health: (1) ensuring broad access to diverse data sets used to deliver insights (2) removing barriers to clinical workflow incorporation, and (3) advancing technology, both in the lab and in the cloud.

Machines refine and improve products. Artificially intelligent machines will soon have the same effect on the law. Future developments in artificial intelligence and machine learning will dramatically reduce the costs currently associated with rules and standards. Extending this insight, we predict a world of precisely tailored laws (“micro-directives”) that specify exactly what is permissible in every unique situation. These micro-directives will be largely automated. If the state of the world changes, or if the objective of the law is changed, the law will instantly update. The law will become “self-driving.”

The evolutionary path toward self-driving laws will be piecemeal and incremental.

A year or two ago, if you’d asked, I would have suggested that yes, eventually, software and robots will take all our jobs. The only jobs that were potentially safe were the purely creative ones: writers, painters, musicians, artists of one kind or another. But this year I started painting, and recently I began collaborating with deep neural networks as part of my artistic process. Now I can see that the future seldom runs in a straight line from the present; it usually ends up being far more nuanced, with twists and turns that are only apparent and obvious in hindsight. Integrating an AI into my artistic process has enhanced my work and made me a better painter too.

… By bringing together doctors and data like never before, precision medicine aims to deliver the right treatments in the right dosage at the right time — every time. It helps target the causes of a condition rather than just the symptoms. This is one of the greatest opportunities we’ve ever seen for new medical breakthroughs, but it only works if we collect enough information first.

More on government precision medicine grants:

Scripps Research gets record $120M to change medicine (July 06, The San Diego Union-Tribune)

NIH awards $55 million to build million-person precision medicine study (July 06, National Institutes of Health)

Every day in Seattle, Aristo takes a mostly multiple-choice fourth-grade science test. Aristo has been taking these tests since 2013, learning from mistakes and filling in knowledge gaps, but has yet to pass one. It’s the thinking part that’s hard.

That’s because Aristo is a computer program.

The Allen Institute for Artificial Intelligence on Lake Union’s north shore has been working for three years on the program. Each day, Aristo’s memory is wiped clean so it can’t cheat by remembering past right and wrong answers. Each week, the tests—but not their difficulty level—are changed.

Scientists at the Computational Story Laboratory have analyzed novels to identify the building blocks of all stories.

Mobile applications are a ubiquitous part of our current digital age. In fact, for many they are an indispensable part of everyday life. We use them to shop, to order takeaway food and to book taxis, among myriad other functions.

However, it isn’t just consumer habits that have been fundamentally transformed by these apps; they also have an impact on how businesses market their products and services.

This was a key topic addressed at the 38th ISMS Marketing Science Conference held recently in Shanghai.

A quick method for making accurate, virtual universes to help understand the effects of dark matter and dark energy has been developed by scientists at University College London and the Aragon Centre for Cosmological Studies in Spain. Making up 95% of our universe, these substances have profound effects on the birth and lives of galaxies and stars and yet almost nothing is known about their physical nature.

Also in astronomy:

Modeling confounding by half-sibling regression (July 05, Proceedings of the National Academy of Sciences; Bernhard Schölkopf, David W. Hogg, Dun Wang, Daniel Foreman-Mackey, Dominik Janzing, Carl-Johann Simon-Gabriel and Jonas Peters)

The National Institutes of Health today announced $55 million in awards in fiscal year 2016 to build the foundational partnerships and infrastructure needed to launch the Cohort Program of President Obama’s Precision Medicine Initiative (PMI) (link is external). The PMI Cohort Program is a landmark longitudinal research effort that aims to engage 1 million or more U.S. participants to improve our ability to prevent and treat disease based on individual differences in lifestyle, environment and genetics. The awards will support a Data and Research Support Center, Participant Technologies Center and a network of Healthcare Provider Organizations (HPO).

“You just witnessed some live history in the making,” he told reporters assembled for the occasion. “That was the first official call from one of our state-of-the-art LinkNYC kiosks.”

De Blasio’s eagerness to label just about any accomplishment of his administration as historic is well-known, but he may have had a point in this case: The narrow, gleaming tower looming over him was the forerunner of a full-scale invasion. By the end of July, there will be 500 of them throughout the city. Initially they will replace what remain of the city’s antique pay phones, but when all is said and done, the links, as they’re being called, will number at least 7,500, a standing army of supersized digital foot soldiers blanketing streets throughout the five boroughs.

The National Institutes of Health is giving a La Jolla scientist a record $120 million to help medicine make a historic shift to treating patients based on their specific genetic makeup, lifestyle and environment.

Dr. Eric Topol will co-lead the effort to enroll and engage 1 million Americans in a study that will deeply explore people’s health and regularly provide them with information that they can share with their doctors.

It helps that Wall Street calls New York home, but a concerted effort by the area’s financial firms to partner with technology startups is cementing the city’s reputation as a fintech hub.

We describe a method for removing the effect of confounders to reconstruct a latent quantity of interest. The method, referred to as “half-sibling regression,” is inspired by recent work in causal inference using additive noise models. We provide a theoretical justification, discussing both independent and identically distributed as well as time series data, respectively, and illustrate the potential of the method in a challenging astronomy application.

There are four basic ingredients for making AI: data, compute resources (i.e., hardware), algorithms (i.e., software), and the talent to put it all together. In this era of deep learning ascendancy, it has become conventional wisdom that data is the most differentiating and defensible of these resources; companies like Google and Facebook spend billions to develop and provide consumer services, largely in order to amass information about their users and the world they inhabit. While the original strategic motivation behind these services was to monetize that data via ad targeting, both of these companies—and others who are desperate to follow their lead—now view the creation of AI as an equally important justification for their massive collection efforts.

We want to enable our authors to publish the best research and maximize the benefit of research funding, which includes achieving good practice in the sharing and archiving of research data. We also aim to facilitate authors’ compliance with institution and research funder requirements to share data.

To help accomplish these goals we are introducing a set of standardized research data policies that can be easily adopted by journals and understood by authors.

Surveillance drones don’t have to be creepy. They can also help protect endangered animals and their habitats.

BMW has announced a partnership with Intel and Mobileye, dedicated to building autonomous cars that are capable of a fully driverless mode.

The alliance, announced on Friday at a joint news conference, shows the power dynamics shifting in the automotive industry. Instead of BMW requesting parts and software or sourcing it internally, as it would have in the past, it is now using a more open and collaborative approach.

The victory in March of the computer program AlphaGo over one of the world’s top handful of go players marks the highest accomplishment to date for the burgeoning field of machine learning and intelligence. The computer beat Lee Se-dol at go, a very old and traditional board game, at a highly publicized tournament in Seoul in a 4–1 rout. With this defeat, computers have bettered people in the last of the classical board games, this one known for its depth and simplicity. An era is over, and a new one has begun. The methods underlying AlphaGo, and its recent victory, have startling implications for the future of machine intelligence.

Jim Simons was a mathematician and cryptographer who realized: the complex math he used to break codes could help explain patterns in the world of finance. Billions later, he’s working to support the next generation of math teachers and scholars. TED’s Chris Anderson sits down with Simons to talk about his extraordinary life in numbers. [video, 23:07]

“Apply Machine Learning like the great engineer you are, not like the great Machine Learning expert you aren’t.”

This is the first sentence in a Google-internal document I read about how to apply ML. And rightly so. In my limited experience working as a server/analytics guy, data (and how to store/process it) has always been the source of most consideration and impact on the overall pipeline. Ask any Kaggle winner, and they will always say that the biggest gains usually come from being smart about representing data, rather than using some sort of complex algorithm. Even the CRISP data mining process has not one, but two stages dedicated solely to data understanding and preparation.

So what is Feature Engineering?

Implementations of saliency models described in “Visualizing and Understanding Neural Models in NLP” by Jiwei Li, Xinlei Chen, Eduard Hovy and Dan Jurafsky.

Yelp is expanding the group of third party companies that can directly access its complete reviews data and announcing a new program called “Yelp Knowledge.” … The concept is that through the Yelp data, multi-location brands and franchises won’t simply see reviews (e.g., 3.5 stars) but gain a more holistic sense of how their locations are performing and how customer sentiment is trending.

In this article, I am explaining my trip from Python to Erlang. If you are not a Python developer (+ probably with a deep understanding of Python based web services), or you don’t need/want to scale thing to a very huge scale, you won’t be able to find this article much useful. If you don’t want to develop an infrastructure for your business or If you develop simple blogs, small asset management systems or Hello World-ish websites, This article won’t help you at all and If you are about to choose a language to start, please do not decide based on my words. I am going to tell you what problems I encountered using Python and how Erlang is able to solve those specific problems for me.