Data Science newsletter – October 26, 2021

Noninvasive glucose monitoring devices are not currently commercially available in the United States, so people with diabetes must collect blood samples or use sensors embedded under the skin to measure their blood sugar levels. Now, with a new wearable device created by Penn State researchers, less intrusive glucose monitoring could become the norm.

Led by Huanyu “Larry” Cheng, Dorothy Quiggle Career Development Professor in Penn State’s Department of Engineering Science and Mechanics, the researchers published the details of the noninvasive, low-cost sensor that can detect glucose in sweat in Biosensors and Bioelectronics. The paper, available online, will publish in the journal’s December print issue.

The researchers constructed the device first with laser-induced graphene (LIG), a material consisting of atom-thick carbon layers in various shapes. With high electrical conductivity and a convenient fabrication time of just seconds, LIG appeared to be an ideal framework for the sensing device — but there was a significant caveat.

“The challenge here is that LIG is not sensitive to glucose at all,” Cheng said. “So, we needed to deposit a glucose-sensitive material onto the LIG.”

One of the highest potential sectors of e-textiles revolves around physiological monitoring. This sector is gaining popularity at an impressive rate. Integrating sensors into apparel helps to monitor a wide range of metrics that give insight into the body’s performance. There has been great interest in this sector from sports apparel and medical companies who have driven lots of R&D.

The combination of large amounts of money and interest in having access to a greater amount of relevant and specialised data means that the fields of both Sport and Medicine are likely to continue to invest in discovering the benefits of Smart Textiles. But due to problems associated with scalability and supply chains many companies still do not see producing en masse as viable at the moment. Many companies have therefore shifted their focus to medical products, as they see it as a more secure investment.

Because there are so many different types of workloads for the IoT (from the massive ML processing on autonomous vehicles to the needs of a battery-sipping remote sensor) there are myriad operating systems and chip configurations. A developer might be writing code for an embedded RTOS used on a microcontroller or a full application processor running Linux. If they are working with an MCU there are many different physical configurations and options that can also complicate things.

This time around ARM is attacking several problems and has also used some of its experience in building ecosystems for cloud computing (which is also more heterogeneous from a software perspective than Arm’s traditional world of smartphones) to bring development practices such as DevOps and Continuous Integration/Continuous Deployment from the world of cloud computing to the development of IoT products and software.

But first, Arm has to tackle the timeline problem. When Arm announces a new processor design, other chip companies take the IP and use it to build the silicon. That process can take a year or two before the chips are actually in the hands of people who want to build new products.

Facebook, Instagram, TikTok and other social media companies would face fines of up to $10 million if they failed to obtain parental consent for users under the age of 16 under proposed laws although it remains unclear how permission would be verified.

Most social networks require users to be 13 to open an account but changes in the draft online privacy bill would give the information commissioner powers to set up a mandatory code for social media companies, data brokers and big online platforms such as Apple, Google and Spotify in consultation with the tech industry.

Last month, Intel CEO Pat Gelsinger stepped to a podium on a hazy, wind-whipped day just outside Phoenix. “Isn’t this awesome!” Gelsinger exclaimed, gesturing over his shoulder. Behind him, two large pieces of construction equipment posed theatrically atop the ocher Arizona soil, framing an organized tangle of pipes, steel, and fencing at the company’s Ocotillo campus. “If this doesn’t get you excited, check your pulse,” he said with a chuckle. A handful of executives and government officials applauded at the appropriate points.

Despite the gathering dust storm, Gelsinger genuinely seemed to enjoy himself. He was in Arizona to announce not one but two new fabs that, when finished, will form a $20 billion bet that Intel can return to the leading edge of semiconductor manufacturing, one of the world’s most profitable, challenging, and cutthroat businesses.

The quest for national AI success has electrified the world—at last count, 44 countries have entered the race by creating their own national AI strategic plan. While the inclusion of countries like China, India, and the U.S. are expected, unexpected countries, including Uganda, Armenia, and Latvia, have also drafted national plans in hopes of realizing the promise. Our earlier posts, entitled “How different countries view artificial intelligence” and “Analyzing artificial intelligence plans in 34 countries” detailed how countries are approaching national AI plans, as well as how to interpret those plans. In this piece, we go a step further by examining indicators of future AI needs.

Evaluating fulfillment through factor analysis

Clearly, having a national AI plan is a necessary but not sufficient condition to achieve the goals of the various AI plans circulating around the world; 44 countries currently have such plans. In previous posts, we noted how AI plans were largely aspirational, and that moving from this aspiration to successful implementation required substantial public-private investments and efforts.

After World War II, biomedicine lived up to its promise, and American ideology turned strongly toward individualism. Anti-communist sentiment made advocating for social reforms hard—even dangerous—while consumerism fostered the belief that everyone had access to the good life. Seeing poor health as a matter of personal irresponsibility rather than of societal rot became natural.

Even public health began to treat people as if they lived in a social vacuum. Epidemiologists now searched for “risk factors,” such as inactivity and alcohol consumption, that made individuals more vulnerable to disease and designed health-promotion campaigns that exhorted people to change their behaviors, tying health to willpower in a way that persists today.

This approach appealed, too, to powerful industries with an interest in highlighting individual failings rather than the dangers of their products. Tobacco companies donated to public-health schools at Duke University and other institutions. The lead industry funded lead research at Johns Hopkins and Harvard universities. In this era, Rosner said, “epidemiology isn’t a field of activists saying, ‘God, asbestos is terrible,’ but of scientists calculating the statistical probability of someone’s death being due to this exposure or that one.”

As the use of machine learning has increased in areas such as criminal justice, hiring, health care delivery and social service interventions, concerns have grown over whether such applications introduce new or amplify existing inequities, especially among racial minorities and people with economic disadvantages. To guard against this bias, adjustments are made to the data, labels, model training, scoring systems and other aspects of the machine learning system. The underlying theoretical assumption is that these adjustments make the system less accurate.

A CMU team aims to dispel that assumption in a new study, recently published in Nature Machine Intelligence. Rayid Ghani, a professor in the School of Computer Science’s Machine Learning Department and the Heinz College of Information Systems and Public Policy; Kit Rodolfa, a research scientist in ML; and Hemank Lamba, a post-doctoral researcher in SCS, tested that assumption in real-world applications and found the trade-off was negligible in practice across a range of policy domains.

“You actually can get both. You don’t have to sacrifice accuracy to build systems that are fair and equitable,” Ghani said. “But it does require you to deliberately design systems to be fair and equitable. Off-the-shelf systems won’t work.”

Tom Sanchez, professor of urban affairs and planning, and Chris North, professor of computer science and associate director of the Sanghani Center for Artificial Intelligence and Data Analytics, have been awarded a planning grant from the National Science Foundation’s Smart and Connected Communities program.

The program is committed to accelerating creation of scientific and engineering foundations that will enable smart and connected communities to bring about new levels of economic opportunity and growth, safety and security, health and wellness, accessibility and inclusivity, and overall quality of life.

“Urban planning anticipates and guides the future physical and social conditions of communities to improve quality of life — all with a heavy reliance on increasingly large and varied datasets,” said Sanchez, who serves as principal investigator for the project. “In fact, cities have become primary sites of data collection and algorithm deployment, but the professional field of urban planning lacks a comprehensive evaluation of how artificial intelligence can and should be used to improve analytical processes. Our project will address that question.” [$150,000]

The world is awash with data, and the University of Virginia’s School of Data Science will help sort it.

UVA dignitaries, including President Jim Ryan, Vice Rector Robert Hardie and Data Science Dean Philip Bourne, donned hardhats and pitched ceremonial shovelfuls of dirt Thursday as they broke ground on the school’s new building in the Ivy Road entrance corridor, at the intersection of Ivy Road and Emmet Street. The 14-acre parcel eventually will hold the Data Science School, a hotel and conference center, a performing arts center and other proposed academic buildings.

“This groundbreaking is for the first building project of the Emmet/Ivy corridor, which we believe will be a hub for the University community, with classrooms; offices; gathering, exhibition and performance space; and a hotel and conference center,” Hardie said, speaking from the podium under a white tent during Thursday’s ceremony. “And it is fitting that our newest school should be located here. Data Science is the here and now, and the future, and we want to provide our students with cutting-edge facilities and resources so that they can go out in the world and make important contributions to industry, government and academia.”

Scientists are exploring why some creatures throw away bits of their DNA during development

The National Science Foundation has awarded researchers from industrial engineering, electrical engineering and computer science at the U of A $1.45 million to investigate of the potential of artificial intelligence as a driving force for changes to critical infrastructures and industries.

Their ultimate goal is to establish a collaborative research and workforce development/education program. This four-year multi-institution, multidisciplinary project, worth $6 million in total, will be led by North Dakota State University and the U of A, with other collaborators from University of Nevada-Las Vegas and Nueta Hidatsa Sahnish College in New Town, North Dakota.

Short thousands of student workers in campus dining halls, Michigan State University is asking faculty and staff for help.

In an email sent to MSU deans, directors and chairs on Monday, Vennie Gore, senior vice president for residential and hospitality services and auxiliary enterprises, invited faculty and staff to volunteer without extra pay in the dining halls.

“As you know, like other schools and universities across the country, Culinary Services is experiencing severe staffing shortages,” Gore wrote. “Many businesses in the local area and around the country are hiring, and we are all competing for the same available talent.”

An endless variety of virtual creatures scamper and scuttle across the screen, struggling over obstacles or dragging balls toward a target. They look like half-formed crabs made of sausages—or perhaps Thing, the disembodied hand from The Addams Family. But these “unimals” (short for “universal animals”) could in fact help researchers develop more general-purpose intelligence in machines.

Agrim Gupta of Stanford University and his colleagues (including Fei-Fei Li, who co-directs the Stanford Institute for Human-Centered AI and led the creation of ImageNet) used these unimals to explore two questions that often get overlooked in AI research: how intelligence is tied to the way bodies are laid out, and how abilities can be developed through evolution as well as learned.

Virtual game worlds provide a non-stop stream of open-ended challenges that nudge AI towards general intelligence.

“This work is an important step in a decades-long attempt to better understand the body-brain relationship in robots,” says Josh Bongard, who studies evolutionary robotics at the University of Vermont and was not involved in the work.

We are excited to announce that DeepMind has acquired MuJoCo and is making it freely available to everyone via our Download page. … MuJoCo is a physics engine that aims to facilitate research and development in robotics, biomechanics, graphics and animation, and other areas where fast and accurate simulation is needed.