AI Health

Friday Roundup

The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.

February 11, 2022

In today’s Roundup: assessing algorithmic impact for healthcare; nerve stimulation to treat paralysis; teaching robots to generalize; extending sleep linked to reduced caloric intake; countering quantum hackers; making space for compassionate care; reconsidering the toll from the Black Death in medieval Europe; more weirdness from the Burgess shale; when caregivers are machines; spotlight on ad targeting and data sharing practices; much more:

Deep Breaths

Fossil specimen of Opabinia regalis from the Burgess shale on display at the Smithsonian in Washington, DC. Public domain image via Wikipedia.
Public domain image via Wikipedia
  • “We are accustomed to seeing galaxies, from afar, as soft, glowing eggs of light or as majestic, bejeweled whirlpools. Rarely do we glimpse the roiling beneath the clouds — all the forms of frenzy that a hundred million or so stars can get up to.” All the light you cannot see: This New York Times Science article by Dennis Overbye narrates a radio spectrum view of our home galaxy, courtesy of the MeerKAT radio telescope, in all its complex glory.
  • The Burgess Shale is the gift that keeps on giving, at least as far as delightfully odd Precambrian life is concerned: “The fossil had been unearthed in western Utah, and it had zigzagging body flaps and a tail brimming with enough spikes to make a Stegosaurus jealous. The traits were reminiscent of Opabinia, but the creature’s poorly preserved head was little more than a crimson smear, obscuring the proboscis and generous allotment of eyes.” The New York Times’ Jack Tamisiea has the story.


Photograph or 3-D render of an industrial robotic arm with a grasping claw at one end. Image credit: Sufyan/Unsplash
Image credit: Sufyan/Unsplash
  • “Existing robotics research has made strides towards allowing robots to generalize to new objects, task descriptions, and goals. However, enabling robots to complete instructions that describe entirely new tasks has largely remained out-of-reach. This problem is remarkably difficult since it requires robots to both decipher the novel instructions and identify how to complete the task without any training data for that task.” A post at the Google AI blog by Chelsea Finn and Eric Jang describes recent work aimed at developing robots capable of performing specific new tasks without ever having been trained to do them.
  • “…the focus of this research is a context where the public and private sector use of AIAs [algorithmic impact assessments] intersect – a public health body that has created a database of medical imaging records and, as part of the process for granting access, has requested private sector and academic researchers and developers complete an AIA.” A new report from the Ada Lovelace Institute describes a pilot effort at creating an algorithmic impact assessment in for an AI application in the healthcare setting.
  • “…asking for various kinds of technical fairness requires compromising on overall error, and adding more protected groups increases error rates across all groups. Our goal is to break though such accuracy-fairness tradeoffs. We develop a simple algorithmic framework that allows us to deploy models and then revise them dynamically when groups are discovered on which the error rate is suboptimal.” A preprint by a group of researchers from Amazon Web Services’ AI describes an approach for achieving algorithmic fairness without accompanying loss of accuracy (H/T @arXiv_Daily).
  • “While Deepmind hasn’t necessarily cracked the code on democratizing the complex programming involved in crafting a neural network, the new system uses transformer-based language models to automatically write code, then spin it into programs that can win in a coding contest.” A brief article by Kate Kaye at Protocol describes Google Deepmind’s AlphaCode, the AI division’s foray into automating the writing of computer code.
  • “As in a cheesy time-travel trope, the machines that don’t yet exist endanger not only our future communications, but also our current and past ones. Data thieves who eavesdrop on Internet traffic could already be accumulating encrypted data, which they could unlock once quantum computers become available, potentially viewing everything from our medical histories to our old banking records.” In the “one more thing to worry about” department comes this article at Nature by Davide Castelvecchi, which delves into the threat that quantum computing poses to conventional cryptography – and what can be done to protect computer security in a quantum era.
  • “…the tracking can have particular implications for patient populations. In the process of reidentifying users across multiple sites, for example, a third-party tracking tool could gather together information about a user’s health status while also building a broader profile of their interests, profession, device fingerprints, and geographic region. And the interconnectedness of the ad ecosystem means that this composite picture can potentially pull in information from all sorts of web browsing, including activity on sites like Facebook.” An article by Lily Hay Newman published in Wired Magazine highlights recent work (featured in a recent Roundup) by Light Collective’s Andrea Downing and Duke Clinical Research Institute’s Eric Perakslis that probed the data sharing and privacy practices of Facebook and some of its third-party advertisers.


Black and white photo of a man taking a nap on a couch, arms folded across chest and a hat over his face. Image credit: Laurence BL/Unsplash
Image credit: Laurence BL/Unsplash
  • “…sleep extension reduced energy intake and resulted in a negative energy balance in real-life settings among adults with overweight who habitually curtailed their sleep duration. Improving and maintaining healthy sleep duration over longer periods could be part of obesity prevention and weight loss programs.” A randomized controlled trial reported by Tasali and colleagues in JAMA Internal Medicine finds that a group of overweight study participants who were randomized to receive a “sleep extension” of 2 hours (in addition to their typical <6.5 hours) for two weeks saw an average reduction of 270 kCal in their daily energy intake, without any significant change in energy expenditure.
  • “Previous studies have shown stimulating an injured spinal cord with epidural electrical stimulation (EES) can, with time, help restore mobility in some people with paralysis, but a paper published in Nature Medicine on Monday goes a step further. The findings suggest people with complete paralysis could regain a broader range of motion within days if dormant spinal nerves that mediate leg and upper-body movement are reengaged with a personalized device.” STAT News’ Isabella Cueto reports on a recent first-in-human study that reports success in restoring motor function to patients paralyzed below the waist by spinal cord injuries.
  • “By analyzing ancient deposits of pollen as markers of agricultural activity, researchers from Germany found that the Black Death caused a patchwork of destruction. Some regions of Europe did indeed suffer devastating losses, but other regions held stable, and some even boomed.” A New York Times’ article by Carl Zimmer highlights research just published in Nature Ecology & Evolution by Izdebski and colleagues that challenges previous conventional understandings of the extent and uniformity of mortality during the initial 14th Century outbreak of plague in Europe known as “the Black Death.”
  • “Hundreds of people die every year in the United States after eating food tainted with salmonella, listeria and other dangerous pathogens. As wrenching as those deaths are, though, they are only the tip of the toll that food poisoning takes on the United States, where millions more people are sickened each year….For many of those victims, the effects can be life-altering. There can be kidney or gastrointestinal troubles that persist for years.” An investigative report by Maryam Jameel and co-published by ProPublica and the Texas Tribune uncovers the largely hidden – and extensive – toll of food poisoning outbreaks.
  • “Omicron has dozens of mutations not seen in the original SARS-CoV-2 strain that researchers first detected in Wuhan, China. More than 30 of those mutations are in the spike protein on the coronavirus’s surface, which helps the virus to latch on to and infect host cells. No previous SARS-CoV-2 variant seems to have accumulated so many genetic changes.” A Nature feature article by Diana Kwon summarizes recent research pointing to the genetic clues that may explain why the COVID Omicron variant is simultaneously more transmissible than other variants but also seems to cause less severe disease on average.
  • “A significant number of these prescriptions come from a small minority of doctors who are willing to write them, often using telemedicine to do so, according to Kolina Koltai, a misinformation researcher at the University of Washington. The same doctors frequently promote anti-vaccine conspiracy theories.” An NPR report by Geoff Brumfiel investigates the intersection of conspiracy theories, telemedicine, and the provision of unsanctioned COVID treatments by a group of contrarian physicians, some of whom have had run-ins with the law and with professional accreditors. However, as an Ars Technica article by Beth Mole shows, lawmakers in some states are lining up to provide cover for physicians who prescribe unapproved treatments such as ivermectin and hydroxychloroquine for COVID patients: “[The Kansas] Senate bill is SB 381, which would specifically authorize doctors to prescribe off-label and unproven COVID-19 treatments—namely hydroxychloroquine sulfate and ivermectin. And it would force pharmacists to dispense the drugs, even if doing so is against their professional judgment.”


Red and yellow toy robot with arms stretched out as if offering a hug. Image credit: Tincho Franco/Unsplash
Image credit: Tincho Franco/Unsplash
  • “In this framework, care can be separated from the notion of protecting the sanctity of the individual as defined by humanism. Science fiction can accordingly help restructure our language so that it draws from our interconnectedness with machines and with other beings. The pre-emptive dismissal of robotically performed care derives from a fixation on individualized human autonomy, with the lines drawn between self and other traversable only by labor with a definite giver and recipient.” A thought-provoking article by Anabelle Johnston in Real Life probes the nature of care when the caregiver is a machine (H/T @MarkRDeLong).
  • “As the worst effects of the omicron variant start to wane, companies will again start to make noise about bringing people who’ve been working from home on their computers for the last two years back to the office. Thanks to an incredibly tight labor market, however, these employees have more leverage than they typically do to get what they want. How this plays out will shape how work is done for years to come.” An article by Rani Molla at Vox’s Recode tries to go beneath the surface of recent, pandemic-driven changes to the world of work – and the longer-term implications.
  • Content warning: the following item links to a story that discusses suicide. “…suicides linked to sales of the preservative through Amazon have continued. The New York Times identified 10 people who had killed themselves using the chemical compound after buying it through the site in the past two years…Enough people purchased the preservative to attempt suicide that the company’s algorithm began suggesting other products that customers frequently bought along with it to aid in such efforts.” A New York Times report by Megan Twohey and Gabriel J.X. Dance illuminates a deeply disturbing facet of unsupervised e-commerce algorithms that not only point users toward substances that can be used to commit suicide, but also “suggest” additional purchases related to those searches.
  • “Instead of waiting to put together an entire preprint, scientists can now highlight the first figure of their to-be paper on Twitter, garnering excitement and buzz from the relevant scientific community….This a huge win for scientists who want to get feedback and generate buzz for future publications. But it can also be a nightmare when comments or findings are taken out of context or misinterpreted and explode in the mainstream media.” A post by Matthew Bauer at Harvard’s Petrie-Flom Law School’s “Bill of Health” blog weighs the pros and cons of preprints – especially when those preprints attract the attention of news services and social media.
  • “Journals and editors are eager to add a diverse range of scientists to their talent pools, including early- career researchers from around the world…. Yet junior researchers can be invisible to editors because they haven’t published much or lack an online presence.” A Nature feature article by Amber Dance makes a case for early-career researchers taking the plunge as peer reviewers – despite the seeming “thanklessness” that the work sometimes seems to exhibit.


Closeup photo of two people tightly clasping hands. Image credit: National Cancer Institute
Image credit: National Cancer Institute
  • “I find hope in knowing that some critical system solutions are only an arm’s length away, though the answer isn’t more blankets. It’s recognizing and meeting the basic needs of the humans belonging to those arms: frontline workers who want to deliver compassionate care and patients who, even in a crisis, deserve to receive it.” An opinion article at STAT News by emergency physician Jay Baruch lays bare the moral injury suffered by healthcare professionals who find their attempts to provide compassionate care running into systemic barriers.
  • “The perpetual exclusion and misrepresentation of Asian American experiences in health research is exacerbated by three racialized stereotypes—the model minority, healthy immigrant effect, and perpetual foreigner—that fuel scientific and societal perceptions that Asian Americans do not experience health disparities. This codifies racist biases against the Asian American population in a mutually reinforcing cycle.” An article by Yi and colleagues in Health Affairs examines the intertwining effects of structural racism and data quality on the health of Asian Americans.
  • “Failure to enroll a diverse undergraduate population has already excluded outstanding people from science, and limiting affirmative action will only make matters worse. But much more insidious are the messages these fights continue to send.“ Science editor (and former UNC Chancellor) Holden Thorpe makes a case for affirmative action in science as a challenge to the academic policy is being heard by the US Supreme Court.