An AI to predict sepsis is often flawed

PLUS: The latest AI and tech news.

View this email in your browser  |   Manage newsletter subscriptions By Jennifer Conrad | 06.28.21 As many of us begin the transition back to office life, it's time to get to know our colleagues' quirks again: Where's that singing coming from? Who keeps microwaving fish? And when some of your collaborators are AIs, it's worth thinking about how much you trust them. A new government initiative aims to measure just that. There's reason for caution: A new report suggests that an artificial intelligence system used by more than 100 health organizations may not always be reliable.     An Algorithm That Predicts Deadly Infections Is Often Flawed A complication of infection known as sepsis is the number one killer in US hospitals. So it's not surprising that more than 100 health systems use an early warning system offered by Epic Systems. But, as Tom Simonite reports, a new study using data from nearly 30,000 patients in University of Michigan hospitals suggests Epic's system performs poorly. The authors say it missed two-thirds of sepsis cases, rarely found cases medical staff did not notice, and frequently issued false alarms. (The company disputes the findings, which were published last week in JAMA Internal Medicine.) Automated sepsis warnings could be very valuable to hospitals because key symptoms of the condition can have other causes, making it difficult for staff to spot early. Starting sepsis treatment such as antibiotics just an hour sooner can make a big difference to patient survival. The Bigger Issue The findings illustrate a broader problem with the proprietary algorithms increasingly used in health care. "They're very widely used, and yet there's very little published on these models," says Karandeep Singh, an assistant professor at the University of Michigan who led the study. "To me that's shocking." Read more about the study's findings here. PLUS: Listen to Tom Simonite discuss the alarming blind spots in health care AI on the WIRED Gadget Lab podcast. The World Needs Deepfake Experts to Stem This Chaos French authorities indicted four spyware executives for aiding torture, accusing them of selling tools to Libya and Egypt that were used against activists and dissidents. Human rights advocates hope such indictments can help slow the global trade of surveillance technology. Andreessen Horowitz is looking to cash in on crypto with a new $2.2 billion fund to invest in cryptocurrency startups. In an op-ed, Nathan Schneider and Morshed Mannan suggest giving users partial ownership of the tech companies they helped build. These startups are betting on a remote-first world, and reimagining everything from housing to education. A constellation of nanosatellites monitoring the Earth's tropical zone could help warn of major hurricanes and other dangerous storms. Microsoft announced that the upcoming version of Windows will allow Android apps to run on Windows PCs and enable developers to use their own payment systems. This Agency Wants to Figure Out Exactly How Much You Trust AI The National Institute of Standards and Technology (NIST) is a federal agency best known for measuring things like time or the number of photons that pass through a chicken. Now NIST wants to put a number on a person's trust in artificial intelligence, Khari Johnson writes. The researchers say they want to help businesses and developers who deploy AI systems make informed decisions and identify areas where people don't trust AI. NIST views the AI initiative as an extension of its more traditional work establishing reliable measurement systems. Trust will be measured two ways: A user trust potential score may consider things about a person using an AI system, such as their age, gender, cultural beliefs, and experience with other AI systems. The second score, the perceived system trustworthiness score, will cover more technical factors such as whether an outdated user interface makes people call AI into doubt. Why Put a Number on It? The adoption of AI will slow or halt if users don't trust it, says NIST cognitive psychologist Brian Stanton. The agency's effort is a start, but researchers say future studies should look at factors like the impact of human emotion or whether information about an AI model can lead people to trust AI too much. Find out more about efforts to quantify trust in AI here.     Need to Know   French authorities indicted four spyware executives for aiding torture, accusing them of selling tools to Libya and Egypt that were used against activists and dissidents. Human rights advocates hope such indictments can help slow the global trade of surveillance technology. Andreessen Horowitz is looking to cash in on crypto with a new $2.2 billion fund to invest in cryptocurrency startups. In an op-ed, Nathan Schneider and Morshed Mannan suggest giving users partial ownership of the tech companies they helped build. These startups are betting on a remote-first world, and reimagining everything from housing to education. A constellation of nanosatellites monitoring the Earth's tropical zone could help warn of major hurricanes and other dangerous storms. Microsoft announced that the upcoming version of Windows will allow Android apps to run on Windows PCs and enable developers to use their own payment systems.       So, This Happened Twitter hired some of Big Tech's biggest critics for its ethical AI team. (Protocol) The US military and defense contractors are clients of Premise, which runs an app that pays users, many in the developing world, to do basic reporting and information gathering. (The Wall Street Journal/paywall) Fashion brands like Shein are rolling out gamified loyalty programs to get customers hooked. (BoF Professional/paywall) The LAPD launched an investigation into officers' relationships with Amazon's Ring after a report that at least 100 officers received free or discounted Ring cameras and were encouraged to promote the product. (Los Angeles Times) Jay-Z is auctioning an NFT based on the cover of Reasonable Doubt to mark the 25th anniversary of his debut album. (Rolling Stone)     Until Next Time   "Does Siri work for the CIA?" "Is Google Home always listening?" Siri, Alexa, and Google Home answer the web's most searched questions.     More Essential Reads Andreessen Horowitz Goes Ham on Crypto with a New $2.2B Fund Let Users Own the Tech Companies They Help Build These Startups Are Betting on a Remote-First World Tiny Satellites Could Help Warn of the Next Big Hurricane GET WIRED Fourth of July Sale! Ending Soon: Get WIRED for $10 $5. You can't see into the future, but you can subscribe to WIRED. Subscribe now.             This e-mail was sent to you by WIRED. To ensure delivery to your inbox (not bulk or junk folders), please add our e-mail address, wired@newsletters.wired.com, to your address book.   View our Privacy Policy Unsubscribe or manage your newsletter subscriptions   Copyright © Condé Nast 2021. One World Trade Center, New York, NY 10007. All rights reserved.