Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about how AI is saving hospital patients, offline-first connectivity for physical AI, and more!
AI in Action: AI is rewriting the rules of triage
Hospitalized patients can crash fast, often with few visible warning signs beforehand. Researchers from RWJBarnabas Health and Rutgers Robert Wood Johnson Medical School wanted to know whether an AI tool already sitting inside their electronic health records could change that. They tracked 23,132 high-risk patients across 11 hospitals over several years, rolling out the Epic Deterioration Index system-wide after refining it at one academic medical center. In-hospital mortality among those patients dropped from 23.1% to 18.6%, an 18% reduction in the risk-adjusted odds of death, according to results published in NEJM AI.
The index pulls from data hospitals already collect: vital signs, lab results, nursing assessments, age. It recalculates a patient's risk score every 15 minutes and automatically pages the rapid response team once someone crosses into the highest-risk band. After rollout, those activations rose from 25.3% of high-risk stays to 37.5%, yet ICU transfers didn't climb along with them. Dr. Thomas Nahass, the study's lead author, put it plainly: an earlier score means a critical-care physician can lay eyes on a patient before intervention gets harder.
Senior author Stephen O'Mahony was careful to credit more than the software. Staff training, EHR alerts, and automatic notifications all worked together, and the mortality gain reflects that whole system rather than any single score. Because the underlying tool is already built into Epic, one of the most widely used EHR platforms in the country, the approach doesn't require new infrastructure to try elsewhere. Researchers are now testing whether catching patients whose scores are climbing quickly, rather than waiting for them to hit a threshold, buys even more time.
🔥 Rapid Fire
Commentary: What Happens If OpenAI Dies?
Analysis: How Much Money Does AI Need?
OpenAI’s sales show tepid growth, disappointing investors
There is no correlation between AI use and revenue per employee
European Central Bank warns that AI crash is looming
How Big Tech’s AI debt is driving up bond yields
AI data center backlash is showing up everywhere
AI was supposed to win people over by now — it hasn’t
Understanding the limitations of mathematical reasoning in LLMs
Google helps publishers combat AI-driven traffic loss caused by Google
📖 What We’re Reading
Imagine an autonomous inspection robot making its way through a solar farm hundreds of miles from the nearest city. Equipped with cameras, environmental sensors, and AI models trained to identify faults before they become failures, it’s designed to inspect hundreds of panels without human intervention. However, halfway through its route, cellular coverage drops. Does the robot stop working? It can’t.
