The online version of Georgia Institute of Technology’s top 10-ranked master’s program in analytics welcomes 284 learners in the first cohort starting August 21, 2017. Georgia Tech’s Online Master of Science in Analytics (OMS Analytics) begins with seven courses. This is the Institute’s second at-scale degree program, following the 2014 launch of the Online Master of Science in Computer Science (OMS CS) program, which demonstrated that world-class education can be delivered at a lower price. The OMS Analytics is available for less than $10,000, one quarter of the cost of the Institute’s on-campus program.
“In a funny way, I have a new song-writing partner who doesn’t get tired and has this endless knowledge of music making,” [Taryn] Southern told CNN Tech. “But I feel like I get to own my vision; I iterate and choose what I like and don’t like. There’s a lot more control.”
There seems to be a modern day gold rush as companies explore how to use machine learning in clinical decision support software. Unfortunately for libertarians, FDA will regulate some of that software because of its risk profile. While the 21st Century Cures Act that passed last December exempted certain CDS from regulation and indeed FDA intends to exempt even more, FDA will continue to regulate high risk CDS. The question is: how will FDA regulate high risk CDS when the software involves machine learning?
Some might assume that machine learning in healthcare is so new, we have no idea how FDA will react. But that’s simply not the case. FDA has decades of experience regulating machine learning and, fortunately, that gives us some useful clues as to how FDA will respond to the expanded uses of that technology.
Sara Stewart on Twitter: “I made a graph showing my past relationships. It has an ex axis and a why axis.”
Government Data Science News
Daniel Kammen, Director of the Renewable and Appropriate Energy Lab at UC-Berkeley and science envoy for the US Department of State has resigned from Trump’s administration. Kammen’s resignation letter cites the President’s “attacks on the core values of the United States” including “destructive” policies on energy and the environment and Trump’s “failure to condemn white supremacists and neo-Nazis.”
Kammen returns to UC-Berkeley where he will work on the newly announced, still developing, California Climate Science and Solutions Institute. The Institute will involve all ten UC campuses along with Stanford and CalTech.
Germanywants to change academic publishing by paying publishers an annual lump sum for all articles with a German first author and then making those articles free to anyone in the world to read. The German institution would have access to the rest of the content in those journals, though they wouldn’t be able to share it globally. People more senior than I am in countries other than Germany: What would it take to lead an effort in your country to ask for the same terms?
Berkeley Industrial Engineering & Operations Research
from
The Department of Indusrial Engineering & Operations Research is delighted to announce that Professors Pieter Abbeel and Michael Jordan, two of the best-known experts in machine learning, have been appointed as joint faculty in IEOR in addition to their primary appointments in EECS (and Statistics for Jordan).
“Profs. Abbeel and Jordan are terrific colleagues that bring extremely valuable perspectives to our interests in robotics, automation, machine learning, and data science,” states Ken Goldberg, Chair of IEOR.
Jennifer Hill, a professor at NYU’s Center for Data Science and NYU’s Department of Applied Statistics, Social Science, and the Humanities, has won Political Analysis’s Miller Prize for her co-authored paper, “Bias Amplification and Bias Unmasking.”
The Miller Prize is awarded annually for the best work published in Political Analysis, the official journal of the Society for Political Methodology and the Political Methodology Section of the American Political Science Association.
Four of the council members who resigned confirmed in emails that they had done so and that the text of their letter to Trump, obtained by CQ Roll Call, is accurate.
The council members providing that confirmation were Dorgelo; Christy Goldfuss, a vice president at the Center for American Progress who chaired Obama’s Council on Environmental Quality; DJ Patil, a political independent who was formerly U.S. chief data scientist under Obama; and Daniel Tangherlini, former administrator of the General Services Administration.
In the letter, the resigning council members said they normally work in a bipartisan and collaborative fashion.
“Unfortunately,” they added, “our experience to date has not demonstrated that the Administration is adequately attentive to the pressing national security matters within the NIAC’s purview, or responsive to sound advice received from experts and advisors on these matters.”
Let us say that James Vacca is not necessarily the first person you’d think would begin a deeply necessary revolution to peel away some of the secrecy around technology that shapes government decisions. In the 1980s, Mr. Vacca admitted, he told an aide that it would be a waste of money to replace office typewriters with word processors.
Yet on Thursday, Mr. Vacca, 62, a Democratic City Council member from the Bronx, introduced a bill that would require the city to make public the computer instructions that are used, invisibly, in all kinds of government decision-making. Experts say that few, if any, major cities in the United States require transparency for those computer instructions, or algorithms.
Led by University of Massachusetts professors Alexandra Meliou and Yuriy Brun, Themis is like a software data scientist that excels at running its own experiments. “It’s actually using a very fundamental part of the scientific method called ‘causal inference,’” says Brun. “It’s this notion run in statistics that if you just observe the system, you can’t say if someone’s race causes the difference of output, or behavior of software.” But if you can test and observe a system, you can.
That’s actually a lot less complicated than it sounds, says Brun, who walks me through the example of how causal inference works when applying for a loan. “If I run my loan application–I happen to be white–and find software recommends I get a loan, then I can take that same loan application and change just the race,” Brun explains. “Then I ask, what about this application? Can you give this person a loan?” If all other variables on a loan application are the same, and the person is rejected for a loan, then the conclusion is obvious: This loan application is racist.
Using satellite remote sensing data sets can be a daunting task. Giovanni, a Web-based tool, facilitates access, visualization, and exploration for many of NASA’s Earth science data sets.
Not only did a Cornell CIS research paper receive the best paper award at the Conference on Computer Vision and Pattern Recognition (CVPR 2017), it also got a shoutout on Facebook from the site’s founder, Mark Zuckerberg, on July 25.
Cornell researchers Gao Huang, a postdoctoral fellow in the Department of Computer Science; Zhuang Liu of Tsinghua University; Kilian Weinberger, associate professor of computer science; and Laurens van der Maaten of Facebook received congratulations from Zuckerberg for their work on “Densely Connected Convolutional Networks” on his Facebook account.
Company Data Science News
Apple’s CEO Tim Cook wrote a company-wide email excoriating Trump’s recent behavior. Cook wrote, “hate is a cancer, and left unchekced it destroys everything in its path. Its scars last generations.” Cook is donating $1m of the company’s cash to the Southern Poverty Law Center and another $1m to the Anti-Defamation League in addition to matching employee donations and enabling donations via iTunes. This is what leadership looks like.
Netflix hired mega-popular television producer Shonda Rhimes from Disney. Her most popular shows include Grey’s Anatomy and Scandal. James B. Stewart chronicles more of the Hollywood intrigue between Netflix and Disney for the New York Times.
Facebook has the longest average number of years employees stick around of the top ten tech companies. However, it’s only 2.02 years, on average. A recent Bureau of Labor Statistics study reports that the typical American worker held ~12 jobs between ages 18 and 50. This amount of churn is greater than in the past and indicates how important it is for workers to stay relevant. This is especially true for technology fields and may drive increasing demand for education throughout career trajectories. There is some controversy around who should pay for training required to remain current: employers or employees?
Microsoft continues to make incremental progress towards on a hard technical problem: speech recognition. The company trimmed its error rate from 5.9% to 5.1% over the past year. The big deal is that the technology is now better than human transcribers who can listen more than once, consult with others, and rely on context clues. (Transcription is harder than it seems like it should be. I say this as someone who has transcribed 80 of my own interviews.) The news coverage does not comment on how accurate the transcription is when it encounters accents other than the nightly news/Dan Rather standard. The training set is the Switchboard corpus of telephone calls from the 1980s that has US southern accents, but not a whole lot of diversity overall.
Uber tried to give a large donation to Girls Who Code, a non-profit that trains middle and high school girls how to program. Girls Who Code accepted the donation, but organizations like Black Girls Code rejected a smaller donation due to discomfort accepting money from Uber, a company they do not recommend as an employment option to the girls in their programs.
Microsoft is getting into the AI hardware space with Field Programmable Arrays (FPGAs) that will support a variety of deep learning frameworks including their own CNTK, Google’s TensorFlow, and Facebook’s Caffe2. Their relatively late entry to the AI hardware party necessitated a more open, flexible approach that took advantage of existing technology. FPGAs have been around for years. Nvidia is still a clear leader in AI hardware with Google’s new AI chips generating a lot of enthusiasm, as well.
Netflix founder Todd Yellin explains how their recommender system works and where they are moving next. Note to Netflix users: they know you only *say* you love documentaries.
Apple has this new technical journal where they write about AI research. This month, they talk about the evolution of Siri.
Google Research NYC has posted a page where they’re sharing a bunch of algorithms to optimize infrastructure, protect privacy, and improve friend suggestions. Their goal is to “share our work and broaden our dialogue with the research and engineering community. Dear Google Research NYC: I will promise not to write about my visit to Google in the newsletter and give you consulting advice about [possibly] unforeseen consequences of your AI applications for some free lunches.
HireVue is using AI in the human resources space – in this case for making hiring decisions – an area I am watching very carefully. We have seen ethical lapses at the interface between AI and complex human behavior over and over again. This company uses video from job interviews to assess candidate’s facial expressions, body language, tone of voice, and keywords to predict which applicants are going to be the best employees. Note that the humans are talking to a camera, not another human, while they are being taped. This already seems like a big abstraction from an actual job where we have to deal with actual people. Diversity – no matter how we define it – is likely to be squeezed out by strategies like this. They are designed to find similarity between the current employees and the batch being considered which will, at best, replicate current diversity and at worst, shave off little elements of diverse traits and produce an ever-more similar workforce.
University of Michigan's Center on Finance, Law, and Policy
from
Ann Arbor, MI September 14-15 at University of Michigan. Economics meets law, finance, public policy, and psychology in this two-day symposium. [free, registration required]
The Aspen Institute’s New Voices Fellowship is a year-long media skills, communication and leadership program designed for standout development professionals from the developing world. Deadline for nominations is October 15.
“This post is part three of a series of posts introducing the Redis-ML module. The first article in the series can be found here. The sample code for this post requires several Python libraries and a Redis instance running Redis-ML. Detailed setup instructions to run the code can be found in either part one or part two of the series.”