Data Science newsletter – December 1, 2021

We will develop CODA, a method to image in three dimensions large volumes of tumors at single-cell resolution.

Twitter has credited Agrawal’s engineering work on its revenue and consumer platforms with helping turn around its audience growth in 2016-17. Under Dorsey, who had been CEO since 2015, the company focused on making Twitter easier to use and more engaging, and also improving the reliability and scalability of the overall service and its advertising stack.

Agrawal noted that there were fewer than 1,000 Twitter employees when he joined the company in 2011. As of the end of September 2021, it has more than 7,100 employees worldwide, up 33% year over year.

This is scary. But the upside is that it could be the final boost of energy needed to enable a Twitter replacement. So if you’d been thinking of creating a Twitter replacement, this could be the time.

The Vermont Fish and Wildlife Department won a $106,256 competitive grant from the U.S. Fish and Wildlife Service to take advantage of groundbreaking new data that will help conservation planners protect plants, animals and their habitats in the face of climate change.

Vermont Conservation Design is a science-based assessment of Vermont’s ecologically functional landscape that helps guide strategic fish and wildlife conservation. “With this grant, we are excited to fine-tune our assessment to better identify lands and waters that contribute to Vermont’s healthy environment with climate change in mind,” said Director of Wildlife Mark Scott.

First released in 2015, Vermont Conservation Design maps the habitat needed to ensure Vermont’s wildlife remains healthy and abundant. Six years later, new state-wide “Lidar” data from the Vermont Center for Geographic Information provide an opportunity to upgrade this conservation tool.

Switzerland’s two federal institutes of technology have teamed up to offer a new joint doctoral programme in the learning sciences. It aims to train experts to draw on their technological and scientific knowledge in the interests of forwarding the research, teaching and learning.

Anyone who’s sat in a classroom knows how important it is to have the right learning environment, teaching methods and study materials in order to learn a subject properly. Learning sciences are a field of study that combines fundamental research into human cognition with information and communication technology to improve teaching outcomes and enhance the learning experience.

Ask most any woman about her experience with the American healthcare system and you will likely hear stories of medical maltreatment in the form of dismissal, undertreatment or incorrect diagnosis. Add racial bias to the mix and a woman’s likelihood of being victimized in medicine is even worse.

In the largest study of its kind to date, a 2020 analysis of more than 3 million U.S. patients’ hospital admissions between 2012 and 2017 found that adults who are Black or from other underrepresented racial or ethnic groups received up to 10% fewer early treatments for heart problems than white patients. Medical bias according to race and gender is so powerful that even mega stars like Serena Williams have nearly died from it.

Institutions including medical schools and hospitals have responded to the problem of bias with implicit bias training — the use of cognitive techniques to make people aware of their internalized assumptions about race and gender. But the data show that it doesn’t work.

A committee within Harvard’s Office of Undergraduate Education is developing a proposal to introduce double concentrations at the College, which it hopes to submit to a faculty vote in spring 2022.

The office established a subcommittee in September to develop the proposal. Dean of Undergraduate Education Amanda J. Claybaugh heads the subcommittee, which is part of the Standing Committee on Undergraduate Educational Policy.

Currently, undergraduates may only pursue joint concentrations, in which students pursue work in two departments and write a thesis that integrates the methods of the two fields. In 2019, the educational policy committee formed a separate subcommittee to study joint concentrations in response to “concerns” from students and faculty about the current system.

An interdisciplinary research team from the University at Buffalo was awarded a $378,940 grant from the U.S. National Science Foundation to explore how to better utilize the social media platform Twitter for improving disaster response.

Yingjie Hu, PhD, assistant professor in the Department of Geography in the UB College of Arts and Sciences, is the project’s principal investigator, with Kenneth Joseph, PhD, assistant professor in the Department of Computer Science and Engineering in the School of Engineering and Applied Sciences, serving as co-principal investigator.

The team also includes an industry partner, Geocove — a geographic information system (GIS) and disaster management company started by UB alum Karyn Senneff Tareen.

We study information substitutability in the financial market through a quasi-natural experiment: the pandemic-triggered lockdown that has hampered people’s physical interactions hence the ability to collect, process, and transmit soft information. Exploiting the cross- sectional and time-series variations of lockdown, we investigate how the difficulty to use soft information has prompted a switch to hard information and its implication on fund investment, performance, and risk management. We show that lockdown reduces fund investment in proximate stocks and generates a portfolio rebalancing toward distant stocks. The re- balancing negatively impacts fund performance by reducing fund raw (excess) return of an additional 0.76% (0.29%) per month during lockdown, suggesting that soft and hard information is not easily substitutable. Lastly, we show that soft information originates mainly from physical human interactions, primarily in cafés, restaurants, bars, and fitness centers; and the virtual world based on Zoom/Skype/Team fails to substitute physical interactions fully, thus cannot provide sufficient soft information.

Reports suggest that Raspberry Pi is planning an IPO in the Spring that would value it at over £370m.

It’s hard to track all the ways this pandemic has upended “normal” life, but surely one of the most significant changes has been how and where, and even when, we work.

You might call the last year or so a remote work revolution, but that’s not quite right. For one thing, remote work wasn’t an option for most of the country. But even for the fortunate people who were able to work from home, what they were doing wasn’t really working. It’s more like a panicked compromise forged under the chaos of a national emergency.

But as we inch our way toward the other side of this pandemic — or at least the closest we’ll get to the other side of it — we have an opportunity to rethink our broken relationship to work. The pandemic was an inflection point, and what happens or doesn’t happen next is up to us.

This is the case that Charlie Warzel and Anne Helen Petersen make in their new book, called Out of Office.

Shortly after the shellshock announcement that Texas and Oklahoma would leave the Big 12 for the SEC, the scorned conference reloaded with the planned addition of four teams that span nearly the full breadth of the country. The farthest-flung, Brigham Young and Central Florida, are 1,900 miles apart — 2,300 miles by car, for one of the longest conference road trips imaginable. But the new-look Big 12 will fit right in within the college football landscape: Every Power Five conference except the SEC already has member schools in excess of 1,000 miles apart.1

Conference realignment has tested and made a mockery of the limits of geography. Lines that were once drawn around proximity to campus and in the supposed best interest of student athletes have rapidly warped and been redrawn around media markets and potential revenue. The primary power brokers of the sport — the ACC, Big 12, Big Ten, Pac-12 and SEC — each hold billion-dollar TV rights contracts, and the addition of new member schools reopen those already lucrative agreements. Superconference blueprints are freely available online.

Some species do it in pairs, some without knowing the other parties involved, and some even do it on their own: when it comes to replication, nature is nothing if not versatile.

Now researchers say they have found that clusters of frog cells can undergo a form of replication never before seen in plants or animals. The spherical clumps, known as xenobots, can give rise to “offspring” by sweeping up loose cells and swashing them into yet more clusters.

“These things move around in the dish and make copies of themselves,” said Prof Josh Bongard, of the University of Vermont, a co-author of the research.

Micro-sized cameras have great potential to spot problems in the human body and enable sensing for super-small robots, but past approaches captured fuzzy, distorted images with limited fields of view.

Now, researchers at Princeton University and the University of Washington have overcome these obstacles with an ultracompact camera the size of a coarse grain of salt. The new system can produce crisp, full-color images on par with a conventional compound camera lens 500,000 times larger in volume.

It’s no secret that the tech giants gather (and buy and sell) tremendous amounts of data about their customers, which is almost all of us. We may rightly worry about how much of our personal data is in the hands of private companies. But we might spend less time thinking about what exactly they do with that data — including using artificial intelligence (AI) to exploit human decision-making.

Humans are pretty good at manipulating each other; in fact, we’ve likely been engaging in “tactical deception” for thousands of years. But thanks to the assistance of AI, software systems that learn for themselves, humans may be more vulnerable to that coercion ever.

When deployed the right way, artificial intelligence can persuade you to buy something, share a post, vote for a candidate, or do any number of things. Recently, a team of researchers from the Commonwealth Scientific and Industrial Research Organisation, Australia’s federal scientific and research agency, conducted a series of experiments that explored how AI influences human decision-making. The results showed that AI could locate and exploit weaknesses in human decision-making to guide people toward certain decisions. “The implications of this research are potentially quite staggering,” Amir Dezfouli, an expert in machine learning at CSIRO and lead researcher on the study, said in a press release.

Also, it reminds me that the US IS ALREADY SUPPOSED TO BE DOING THIS

EVCache is a distributed in-memory caching solution based on memcached. It is a Tier-0 system at Netflix with its footprint across ~18,000 servers holding ~14 petabytes of data, and still rapidly growing.

We had previously given an overview of how we perform cache warming on our EVCache clusters and why it’s needed. The architecture mentioned in the original cache warming article worked great for a vast majority of our use cases. However, as Netflix’s subscriber base grew over the years, the footprint of the data stored in EVCache has increased multiple fold and will continue to increase to meet new and additional demand. As the scale and sensitivity of the clusters increased, the architecture needed to evolve to keep up.

Coming March 2022. “Best practices for addressing the bias and inequality that may result from the automated collection, analysis, and distribution of large datasets.”