NYU Data Science newsletter – December 14, 2015

GROUP CURATION: N/A

A decade ago the late Jack Marburger, a physicist and former college president who served as science advisor to President George W. Bush, challenged academics to come up with scientific evidence on the impact of federal research investment. Even more adamantly, Congress required the National Science Foundation to “better articulate the value of grants to the national interest.”

A study in the latest issue of Science (online Dec. 10, in print Dec. 11) led by Professors Julia Lane of New York University, Bruce Weinberg of Ohio State University and Jason Owen Smith of the University of Michigan demonstrates a significant path by which federally and non-federally funded investment in research make an impact on the economy. Inspired by the Robert Oppenheimer maxim that “The best way to send information is to wrap it up in a person,” the researchers combed an array of new data, combined with Census Bureau information, to trace where doctoral recipients get their jobs and what their earnings are after receiving research training.

Two women who switched continents get to know each other through the data they draw and send across the pond.

As part of an ongoing exhibition series with the Hyundai Card Design Library in Seoul, Korea, MoMA senior curator Paola Antonelli has organized three capsule exhibitions that highlight new frontiers in contemporary design and encourage international dialogue. The second exhibition, Data Visualization, opened in Seoul in July 2015 and has recently concluded. We live in an age where we are bombarded by data gathered by sensors, arrayed by software, and dispersed via ever-proliferating networks. To visualize this data is to understand it. As the projects in this exhibition demonstrated, designers and scientists create diagrams, three-dimensional maps, and other graphics to help us make sense of the copious amount of information with which we are confronted daily. The New York-Seoul exchange has deepened during in-person visits to Korea where Antonelli met with designers to discuss their current work. Here, one of these designers, Sey Min, reflects on the specific circumstances of data visualization design in the Korea.

And so, my fellow scientists: ask not what you can do for reproducibility; ask what reproducibility can do for you! Here, I present five reasons why working reproducibly pays off in the long run and is in the self-interest of every ambitious, career-oriented scientist.

Tesla founder Elon Musk, big-name venture capitalist Peter Thiel, LinkedIn co-founder Reid Hoffman, and several other notable tech names have launched a new artificial intelligence startup called OpenAI, assembling a particularly impressive array of AI talent that includes a top researcher from Google. But the idea, ostensibly, isn’t to make money.

Overseen by ex-Googler Ilya Sutskever and Greg Brockman, the former CTO of high-profile payments startup Stripe, OpenAI has the talent to compete with the industry’s top artificial intelligence outfits, including Google and Facebook—but the company has been setup as a non-profit. “Our goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return,” Brockman said in a blog post.

---

More OpenAI coverage:

AI’s Real Risk, Michael Schrage in Harvard Business Review

Why OpenAI Matters, Miles Brundage

Professor John Denero will give a talk and answer questions about the new Data Science Undergraduate Education Program. In spring 2016, UC Berkeley will officially launch the Foundations of Data Science course, a course jointly developed and offered by Computer Science, Statistics, and the School of Information. In conjunction with the Foundations of Data Science course, there will be a series of connector courses that will allow students to explore data science from fields as diverse as cognitive science, history, literature, engineering, ecology, and beyond.

The classical music world has a lot of conversations about what we do.

Is there enough music by living composers or female composers? Do we in the United States give American composers their due? Do orchestras play too many of the warhorses that have name recognition with their audience and overlook some great music either by lesser known composers or seldom-heard works that are deeper in a well known composer’s catalogue?

Often what’s missing from these discussions is data. Last year we tried to offer up some of those numbers to make those conversations more meaningful by collecting and categorizing all of the music that 22 of the largest American symphony orchestras were playing in the 2014-2015 classical season. This year, we hope to do the same with a bit of a twist.

There is “widespread discrimination” on Airbnb against renters with African-American-sounding names, according to a working paper from Harvard researchers.

Overall, users with names like Latoya and Darnell were 16% less likely to be accepted as guests than those with names like Allison or Greg, according to the study.

Gregory Dobler is an astrophysicist who honed his craft by recording spectral images of quasars and black holes. Now, from a high-rise rooftop in Brooklyn, he is training his lens on the expanding universe of New York City.

Every 10 seconds for two years, Dr. Dobler and his colleagues at New York University’s urban observatory have taken a panorama of Manhattan. Across hundreds of wavelengths of light, they are recording the rhythmic pulse of a living city, just as astronomers capture the activity of a variable star.

“Instead of taking pictures of the sky to see what is going on in the heavens, we are taking pictures of the city from a distance to see if we can figure out how the city is functioning,” says Dr. Dobler, a scientist at NYU’s Center for Urban Science and Progress.

To turn back the invading force that is influenza, drugs can directly block the virus or adopt an indirect approach, a kind of scorched earth strategy. With the indirect approach, drugs could be used to prevent the virus from taking over host capabilities. Doing so is difficult, however, because influenza-host interactions are poorly mapped.

Enter “big data.” It is being used to survey the biochemical landscape of influenza-host interactions. Already, it has identified 20 previously unrecognized host proteins that are required for influenza A virus (IAV) interactions. One host protein in particular has been shown to be a particularly promising drug target. This protein, called UBR4, is useful to the virus because of its role in supporting viral budding from the host cell membrane.

These results were generated by scientists based at the Sanford Burnham Prebys Medical Discovery Institute (SBP) and the Icahn School of Medicine at Mount Sinai.

Why the neural net project creating wild visions has meaning for art, science, philosophy—and our view of reality.

Today’s artificial intelligence may not be that clever, but it just got much quicker on the uptake. A learning program designed by a trio of researchers can now recognize and draw handwritten characters after seeing them only a few times, just as a human can. And can do it so well that people can’t tell the difference.

The findings, published in the journal Science, represent a major step forward in developing more powerful computer programs that learn in the ways that humans do.

“For the first time, we think we have a machine system that can learn a large class of visual concepts in ways that are hard to distinguish from human learners,” study coauthor Joshua Tenenbaum from the Massachusetts Institute of Technology said in a news briefing.

Computer researchers reported artificial-intelligence advances on Thursday that surpassed human capabilities for a narrow set of vision-related tasks.

The improvements are noteworthy because so-called machine-vision systems are becoming commonplace in many aspects of life, including car-safety systems that detect pedestrians and bicyclists, as well as in video game controls, Internet search and factory robots.

Researchers at the Massachusetts Institute of Technology, New York University and the University of Toronto reported a new type of “one shot” machine learning on Thursday in the journal Science, in which a computer vision program outperformed a group of humans in identifying handwritten characters based on a single example.

More coverage:

Humans take note: Artificial intelligence just got a lot smarter, Los Angeles Times

A Learning Advance in Artificial Intelligence Rivals Human Abilities , The New York Times

People learning new concepts can often generalize successfully from just a single example, yet machine learning algorithms typically require tens or hundreds of examples to perform with similar accuracy. People can also use learned concepts in richer ways than conventional algorithms—for action, imagination, and explanation. We present a computational model that captures these human learning abilities for a large class of simple visual concepts: handwritten characters from the world’s alphabets. The model represents concepts as simple programs that best explain observed examples under a Bayesian criterion. On a challenging one-shot classification task, the model achieves human-level performance while outperforming recent deep learning approaches. We also present several “visual Turing tests” probing the model’s creative generalization abilities, which in many cases are indistinguishable from human behavior.

---

More from the newspapers:

Humans take note: Artificial intelligence just got a lot smarter, Los Angeles Times

A Learning Advance in Artificial Intelligence Rivals Human Abilities, The New York Times