NYU Data Science newsletter – September 6, 2016

GROUP CURATION: N/A

Four of the biggest AI startup acquisitions of the last five years have come from the UK, starting with Google’s purchase of Deep Mind in 2014 for a reported £400 million. Since then Apple has purchased Cambridge-based natural language processing specialists VocalIQ, Microsoft bought the machine learning powered keyboard SwiftKey in February and Twitter acquired Entrepreneur First alumni Magic Pony in June.

Here are some of the most interesting startups working in the area of artificial intelligence in the UK today, whether that is machine learning, deep learning, neural nets or computer vision. Could the next Deep Mind be lurking in there?

To advance artificial intelligence research, Tambe and Associate Professor Eric Rice of the USC School of Social Work have joined forces to co-direct the Center on Artificial Intelligence for Social Solutions, or CAISS. At the newly announced center, one of the first such university-based institutes dedicated to studying AI as a force for good, researchers will leverage artificial intelligence to address myriad problems ranging from climate change to security to health to homelessness.

“AI has continued its significant advancements in the past several years, and now there is greater potential than ever to apply computational game theory, machine learning, automated planning and multi-agent reasoning techniques to problems that are socially relevant,” Tambe said. “There is a very important opportunity here to really harness AI for social good.”

Silicon Valley is turning to the dismal science in its never-ending quest to squeeze more money out of old markets and build new ones. In turn, the economists say they are eager to explore the digital world for fresh insights into timeless economic questions of pricing, incentives and behavior.

“It’s an absolute candy store for economists,” [Peter Coles, a former Harvard Business School professor] said.

Research Scientist at OpenAI. Previously ML/CV PhD student at Stanford.

More Andrej Karpathy:

A Survival Guide to a PhD (September 07, Andrej Karpathy blog)

This Supercomputer Will Try to Find Intelligence on Reddit (August 15, MIT Technology Review, Will Knight)

On August 25, Bloomberg, in partnership with Western Digital, convened a half-day event to explore the vast promises of this technology revolution and weigh them against concerns over privacy, safety and control. The event examined how massive troves of data and deep learning technologies are teaching computers to see, understand natural language and think as humans do – and what this means for the potential of artificial intelligence to solve problems and do social good. Speakers explored the social, economic and security implications of these types of emerging technologies, and discussed policy prescriptions needed to balance rewards against disruption and ensure the democratization of gains.

A highlight of the event was a keynote address from Transportation Secretary Anthony Foxx, who led a discussion on the future of transportation powered by artificial intelligence.

Geoff Boeing at UC-Berkeley used Craigslist data from 11 million rental listings all over the US and demonstrates that in hot markets like New York and San Francisco significant dispersion in rental prices with a long tail at the top end. In cities like Detroit, there’s a stronger peak around what appears to be a market clearing price.

Others using real estate data:

Boston officials use Big Data to find deceptive rental ads, unsafe units (September 02, The Boston Globe, Matt Rocheleau)

Building a Team from the Inside Out: Alok Gupta on the Evolution of Data Science at Airbnb (September 06, Kaggle, No Free Hunch blog, Alok Gupta)

Technology is finally changing the apartment rental experience (September 04, TechCrunch, Omri Barzilay)

If you live in Baltimore, you may have the feeling that you’re being watched. You are. Baltimore Police track your cellphone use without a warrant. They secretly film the entire city from the air. And as concerns about the uses and privacy implications of that next-generation surveillance tech have mounted, these domestic spying scandals also raise another question: Why Baltimore?

It turns out that Baltimore checks off all the requirements to build a modern American urban panopticon: High crime rates, racially biased policing, strained community-police relations, and lack of police oversight have turned Baltimore into a laboratory of emerging surveillance techniques.

On this episode, I am so happy to have Pedro Domingos who is a professor at the University of Washington. … In this conversation we explore the sources of knowledge, the five major schools of machine learning, why white collar jobs are easier to replace than blue collar jobs, machine wars, self-driving cars and so much more. [audio, 1:02:33]

A research project at the University of Washington to see whether analyzing food product reviews on Amazon.com might help to predict recalls is showing promise, and the team leader hopes it will have practical applications for illness outbreak investigations.

Project lead Elaine Nsoesie, Ph.D.., assistant professor of global health at the university’s Institute for Health Metrics and Evaluation, said that while the Unsafe Food Project is still in the “mission learning stage,” initial results are encouraging after a summer fellowship program recently ended.

“What we did this summer is get all the data we needed together from the FDA and try and match that data to the Amazon product reviews we were getting. It was more challenging than we expected,” she said.

Data collection and analysis are an integral part of our society, and are important activities we use to inform our decisions. Big data is no exception. Made up of extremely large sets of data that can be analyzed for trends and other information, big data is extremely useful and relevant when determining strategies and plans for communities and companies. In fact, it’s changing the face of many different industries—including manufacturing. Let’s take a look at how big data is revolutionizing the manufacturing industry.

Leveraging large data sets, deep Convolutional Neural Networks (CNNs) achieve state-of-the-art recognition accuracy. Due to the substantial compute and memory operations, however, they require significant execution time. The massive parallel computing capability of GPUs make them as one of the ideal platforms to accelerate CNNs and a number of GPU-based CNN libraries have been developed. While existing works mainly focus on the computational efficiency of CNNs, the memory efficiency of CNNs have been largely overlooked. Yet CNNs have intricate data structures and their memory behavior can have significant impact on the performance. In this work, we study the memory efficiency of various CNN layers and reveal the performance implication from both data layouts and memory access patterns. Experiments show the universal effect of our proposed optimizations on both single layers and various networks, with up to 27.9x for a single layer and up to 5.6x on the whole networks. [full text pdf download]

Microsoft hopes to show that artificial intelligence can extend beyond “the cold world of computers” by working alongside Tate Britain to sponsor the artistic IK Prize 2016 – the winner of which devised an AI algorithm that pairs the Tate’s historic art collection with modern photojournalism from Reuters’ archive.

In this PG Podcast, Adam introduces the futuristic-sounding topic of human-assisted A.I., which his company is now bringing to the world. We discuss how we can get machines to help humans do what humans are best at, and what that means for the future of work. We also talk about the interplay between academia and industry, and how to keep a foot in both worlds.

You have done a lot of work on kernel methods. A year ago there was indication that kernel methods are not dead and that could match or outperform deep nets. Is that the case? Is it time for them to retire?

LS) I think kernel methods and deep learning are cousins. The field needs to combine them rather than throwing either away.

In fact, they share lots of similarities, such as both tries to learn nonlinear functions. One can design kernel functions to capture problem structure as much as one can choose the architecture of deep learning according to the problem at hand.

Technology companies have become a powerful way to build the future. Our goal with this series is to share advice about how you can do it, too.

NASA’s computer model predicts that the exoplanet Proxima b will appear as a pale purple dot when it is imaged for the first time.

In our quarterly analysis of companies pursuing healthcare-focused applications of AI, we reported that deals leapt from less than 10 in 2011 to 60 in 2015. So far this year (as of 8/23/2016), companies in this space have raised over 55 equity funding rounds. Some of the recent deals include a $25M Series A round raised by London-based health services startup, Babylon Health, backed by investors including Kinnevik and Google-owned DeepMind Technologies (Babylon will reportedly roll out a Siri-like voice recognition interface this year), and a $154M Series A round raised by China-based iCarbonX.

The 2013-2016 epidemic of Ebola virus disease in West Africa was of unprecedented magnitude, duration and impact. Extensive collaborative sequencing projects have produced a large collection of over 1600 Ebola virus genomes, representing over 5% of known cases, unmatched for any single human epidemic. In this comprehensive analysis of this entire dataset, we reconstruct in detail the history of migration, proliferation and decline of Ebola virus throughout the region. We test the association of geography, climate, administrative boundaries, demography and culture with viral movement among 56 administrative regions. Our results show that during the outbreak viral lineages moved according to a classic ‘gravity’ model, with more intense migration between larger and more proximate population centers. Notably, we find that despite a strong attenuation of international dispersal after border closures, localized cross-border transmission beforehand had already set the seeds for an international epidemic, rendering these measures relatively ineffective in curbing the epidemic. We use this empirical evidence to address why the epidemic did not spread into neighboring countries, showing that although these regions were susceptible to developing significant outbreaks, they were also at lower risk of viral introductions. Finally, viral genome sequence data uniquely reveals this large epidemic to be a heterogeneous and spatially dissociated collection of transmission clusters of varying size, duration and connectivity. These insights will help inform approaches to intervention in such epidemics in the future.

“Rudolf Kalman, a Budapest-born engineer and mathematician died on July 2 in Gainesville, Florida, at age 86. His fundamental contribution, an algorithm called the Kalman filter, made possible many essential technological achievements of the last 50 years. … Someone once described the entire GPS system—an Earth-girdling constellation of satellites, ground stations, and computers as ‘one enormous Kalman filter.'”

We’re interested in hearing about the challenges faced by early-career scientists worldwide, especially if you’ve recently started your own lab, are struggling to maintain a lab, or have left research. We want to hear your stories.

Public comments will close on Thursday, 15 September 2016.

The Foundation seeks Letters of Interest for grants in Open Science and Science and Technology Research and Development.

Deadline for submissions is Friday, 16 September 2016.

  Summer’s notable departures and arrivals at NYU Center for Data Science

• Foster Provost stepped down from the Moore-Sloan steering committee and handed over the CDS Interim Directorship to Claudio Silva.
• Roy Lowrance, CDS Managing Director, moved into industry.
• Dan Cervone, Moore-Sloan Fellow, took a job with the LA Dodgers starting in October.
• Pablo Barberá, a Moore-Sloan Fellow, started his tenure-track position at USC in July.
• Andrea Rooy-Jones joined Moore-Sloan in a research and outreach capacity.
• Michael Gill joined us as a Moore-Sloan Fellow.
• Our top-notch administrator David Clark started an MBA program this fall; we welcome Kathryn Angeles in his position.

[Astro Hack Week] is one approach to improving data literacy in astronomy while at the same time providing a collaborative venue for researchers to share ideas … “I will give an overview of the ideas that underlie Astro Hack Week, highlight some results from this week’s workshop, and try to convince you that your field needs a hack week too.”

Also in hackathons:

Data Hackathons Workshop in Denver, part of International Data Week (September 15)

Seattle Sports Tech Hackathon (September 10-11)

The open source Git project has just released Git 2.10.0, with features and bugfixes from over 70 contributors. Here’s our look at some of the most interesting new features.

“We’re announcing the open beta release of the newest set of features and open source components in Google Cloud Endpoints, a distributed API management suite that lets you deploy, protect, monitor and manage APIs written in any language and running on Google Cloud Platform (GCP).”

“I’ve listened to a bunch of machine learning and data science podcasts in the last few months, so I thought I’d share my favorites: 1. The Data Skeptic.”

The open preprint repository network, a singular interface for searching the major preprint archives.