Nearly two dozen higher education groups warn the government to be cautious when advising US research universities to keep an eye on students and faculty with ties to certain Chinese institutions.
Purdue University, The Exponent student newspaper, Alexandra Weliever
from
A Boilermaker used skills he learned as a freshman studying data science to determine if Purdue pays male and female employees equally.
Earlier this summer sophomore Sanjeev Lingam-Nattamai analyzed salary data obtained from Purdue’s public records office detailing every employee paid by the University in 2018.
Focusing on over 12,000 West Lafayette employees, Lingam-Nattamai discovered what he believes is a wide gap between how much men and women are paid in faculty and athletic roles.
The U.S. Department of Energy, through its University Turbine Systems Research program, has awarded researchers at the University of Pittsburgh’s Swanson School of Engineering $802,400 to find an effective quality assurance method for additive manufacturing of new-generation gas turbine components. The three-year project has received additional support from the University of Pittsburgh ($200,600), resulting in a total grant of $1,003,000.
Xiayun (Sharon) Zhao, PhD, assistant professor of mechanical engineering and materials science at Pitt, will lead the research, working with Albert To, associate professor of mechanical engineering and materials science at Pitt, and Richard W. Neu, professor in the Georgia Institute of Technology’s School of Mechanical Engineering. The team will use machine learning to develop a cost-effective method for rapidly evaluating, either in-process or offline, the hot gas path turbine components (HGPTCs) that are created with laser powder bed fusion (LPBF) AM technology.
New York Gov. Andrew Cuomo signed legislation in late July to create a temporary state commission that will examine how artificial intelligence impacts his state.
In doing so, New York joined Vermont and Washington in establishing an A.I. task force that will examine the cutting-edge technology and then make recommendations about how it should be regulated. The groups vary in their mission, but the general message is the same: companies pushing A.I., the brains behind innovation like robotics and facial recognition software, can’t necessarily be trusted to do what’s in the best interest of state residents.
The University of Miami (UM) announced that their new Triton supercomputer is installed and helping their researchers explore new frontiers of science. The new supercomputer will be UM’s first GPU-accelerated HPC system, representing a completely new approach to computational and data science for the university’s campuses. Built using IBM Power Systems AC922 servers, the new HPC system was designed to maximize data movement between the IBM POWER9 CPU and attached accelerators like GPUs.
My family represents three generations of University of Texas (UT) students: my father, brother, and myself are all proud Austin graduates, and today my oldest daughter is enrolled at UT Dallas to study computer science and animation. With this background, and my role as a tech analyst focused on wireless telecommunications and enterprise networking, I was understandably excited to learn of the school’s strategic role in shaping the future of cellular communications. I recently had the opportunity to visit UT and speak to one of the professors, Dr. Sanjay Shakkottai, who, along with 22 other professors, all contribute to the research effort with the Wireless Networking & Communications Group (WNCG) in Austin.
All over the city, Bostonians are forgoing pints at the local tavern for a night of Netflix at home—and corner bars are slowly disappearing. But in an era of increasing isolation, the co-owner of the iconic pub the Plough & Stars wonders what it is we’re really losing.
Pressure is mounting against Amazon’s continued involvement with US Immigration and Customs Enforcement (ICE), with Whole Foods employees now demanding the company cease ties with the agency through its business dealings with the controversial government contractor Palantir. The demands, posted in a public letter earlier today by the Whole Foods unionization group Whole Worker, direct harsh words at the parent company of the grocery chain, which Amazon purchased for $13.7 billion in 2017.
In particular, the letter calls Amazon’s rule “dystopian,” and Whole Foods employees say they will continue to combat the company by leaking information and attempting to undermine policies and business dealings that lead to the deportation of undocumented people and other rights abuses.
Opioid addiction is an issue that hits close to home: In 2017, King County saw 260 opioid-involved overdose deaths, a number that’s on the rise. Shyam Gollakota, an associate professor at the University of Washington’s Paul G. Allen School of Computer Science and Engineering, is working to stop that trend.
With a team of UW researchers, Gollakota has developed the Second Chance app, a smartphone-based intervention intended for people who are using injectable opioids illegally.
Still in the testing phase, the app uses sonar to detect overdose-related breathing irregularities in drug users, and when an irregularity occurs, the app alerts lifesaving services.
The value of underwritten insurance premiums is estimated to exceed $20 million within five years, an increase from $1.3 billion this year, according to new research.
The growth is being driven by faster customer onboarding, streamlined underwriting processes and reductions in operations costs enabled by AI, according to the study by Juniper Research.
Global revenue from telematics, devices that capture and transmit data often from vehicles, is expected to grow from $1.2 billion this year to $5.4 billion by 2024.
The increased use of telematics and Internet of Things management tools in the motor, home, life and health insurance markets will create more efficiencies in underwriting, according to Juniper.
It seems unlikely anyone will ever come across a loquacious linden. But for the arbor-curious, a red oak at the Harvard Forest in Petersham has been tweeting as @awitnesstree since July 17. Outfitted with sensors and cameras, and programmed with code that allows it to string together posts with prewritten bits of text, the Harvard Forest Witness Tree has been sharing on-the-ground insights into its own environmental life and that of its forest.
Already renowned in certain circles as the subject of the popular climate-change book “Witness Tree” by Lynda Mapes, the century-old oak’s social-media debut was the brainchild of Harvard Forest postdoctoral fellow Tim Rademacher and is now a team effort with Clarisse Hart, who heads outreach and education for the forest. Its online presence is modeled after similar “twittering” trees that chronicle their life experiences as part of a tree-water and carbon-monitoring network based in Europe called TreeWatch.net.
WCAI, Living Lab Radio, Heather Goldstone & Elsa Partan
from
By tapping into what moves people, the emerging field of environmental humanities can help spur climate action. Scholars of history, philosophy, religious studies, literature and media are exploring many aspects of humans’ relationship with the Earth. An entire literary genre of climate fiction, or “Cli-Fi,” depicts often-apocalyptic visions of climate impacts on humanity. Social scientists have worked out how civilizations like the ancient Maya and medieval Icelanders dealt with climate shocks.
Together with scientists, environmental humanists are reforming scenarios used in climate modeling. Scenarios originated as an improvisational form of theater, and humanists are reclaiming them as a rehearsal space for the massive societal shifts required to avert dangerous climate change. [audio, 13:42]
… Cybersecurity consultant Brad Dixon from Carve Systems detailed how he managed to cheat in Zwift — a popular training app for cyclists that let users compete against one another, during a talk at this year’s DEF CON. The app requires players to remove the back wheel of their bike and set it in a stationary trainer device that uses sensors and resistance to translate power and speed to the online avatar. Dixon wanted to see if he could manipulate that data traveling from the bike to the Zwift and managed to successfully trick the system into thinking he was superhuman when in reality he wasn’t peddling at all.
A new analysis shows that the number of Chinese AI researchers has increased tenfold over the last decade, but the majority of them live outside the country.
Denver, CO October 20-24. “Teradata (NYSE: TDC), the industry’s only Pervasive Data Intelligence company, today shared new and expanded elements for this year’s Teradata Universe, the conference where customers, business leaders, experts and innovators come together to focus on the possibilities, opportunities and challenges of next-gen analytics. With a goal of using data and analytics to find answers to the toughest business challenges, this year’s event will include thought-leading keynotes from industry pundits, nearly 100 sessions about real-world implementations of Teradata solutions and significantly expanded education offerings throughout the week.”
Irvine, CA August 29, starting at 8:30 a.m. “If you are interested in data and California water issues please register for one or both datathon events. Note that in person attendance is required to participate in either of the datathon events. Also, note you do need to register for the Datathon in order to participate in that event, but registration at the CA Water Data Summit or Water Solutions 4 conferences is optional for Datathon participants.” [free, registration required]
Barcelona, Spain January 27-30. “The ACM FAT* conference has predominantly focused on fairness, accountability, and transparency. The success of the field of fairness, accountability, and transparency has also attracted much critique and renewed attention to the limitations of achieving these goals in systems that implement statistical, machine learning, optimization, or autonomous computing techniques. A number of prominent studies acknowledge that addressing societal problems embedded in such computing systems may require more holistic approaches.” Deadline for pre-proposal submissions is September 9.
“The Facebook Fellowship Program is designed to encourage and support promising doctoral students who are engaged in innovative and relevant research in areas related to computer science and engineering at an accredited university.” Deadline for applications is October 4.
With the NFL season just around the corner, hundreds of athletes are getting prepared. Players range from humongous linemen who sacrifice speed for strength; to defensive backs with incredible speed — but stereotypically bad hands — to wide receivers who are quick, agile, and can pluck a ball out of the air with the greatest of ease. One important measure of a player’s speed is the 40-yard dash, with a 4.4 second time considered excellent, and a few who can even run faster. In celebration of the upcoming season, software/hardware collective “Flopperam” decided to make their own DIY run timer.
Research teams from three universities recently released a dataset called ImageNet-A, containing natural adversarial images: real-world images that are misclassified by image-recognition AI. When used as a test-set on several state-of-the-art pre-trained models, the models achieve an accuracy rate of less than 3%.
In a paper published in July, researchers from UC Berkeley, the University of Washington, and the University of Chicago described their process for creating the dataset of 7,500 images, which were deliberately chosen to “fool” a pre-trained image recognition system. While there has been previous research on adversarial attacks on such systems, most of the work studies how to modify images in a way that causes the model to output the wrong answer. By contrast, the team used real-world, or “natural” images collected un-modified from the internet. The team used their images as a test-set on a pre-trained DenseNet-121 model, which has a top-1 error rate of 25% when tested on the popular ImageNet dataset. This same model, when tested with ImageNet-A, has a top-1 error rate of 98%. The team also used their dataset to measure the effectiveness of “defensive” training measures developed by the research community; they found that “these techniques hardly help.”
The problem is that zip codes are not a good representation of real human behavior, and when used in data analysis, often mask real, underlying insights, and may ultimately lead to bad outcomes. To understand why this is, we first need to understand a little more about the zip code itself.