Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.
A systematic review and meta-analysis up to September 2025 synthesized 83 studies involving at least 136,840 patients to evaluate machine learning models predicting hematoma expansion, poor functional outcome, and mortality after spontaneous intracerebral hemorrhage. Pooled analyses found that models combining clinical and radiomics features achieved the highest discrimination, with C-indexes of 0.822, 0.850, and 0.860 respectively, largely from internal validation sets.
Using longitudinal EHR data from the All of Us Research Program, a prospective cohort study tracked adults after a first opioid-related diagnosis to predict overdose within 30 days. Classical and machine learning survival models were tested, with 560 overdoses observed among 14,737 individuals and best test-set C-indices between 0.708 and 0.745.
Researchers prospectively collected 3676 smartphone images from 1016 atopic dermatitis patients at 16 Japanese institutions and trained a ConvNeXt model to score four EASI signs against dermatologist ratings. The model achieved AUCs of 0.825 to 0.860 for scores >=2 and over 0.900 for score 3, with erythema most accurately detected.
Investigators developed and externally validated a machine learning model to stratify risk of postoperative hydrocephalus after posterior fossa tumor resection using data from 1,073 patients treated at five tertiary centers from 2013 to 2024. After screening 30 variables, they built a three-variable preoperative model using Evans index, tumor-fourth ventricle relationship, and preoperative CSF diversion status, with SVM showing AUC 0.877 in the external cohort.
By June 2026, researchers had tested whether people can tell real from synthetic faces and voices and whether that skill transfers across senses. In a preregistered study, participants classified both types of stimuli, and performance for each modality was significantly above chance when measured with signal detection theory.
By April 2026, researchers studied 150 mastere28099s-level preservice teachers at a German university as they collaboratively solved three authentic mathematical modeling problems with large language models, analyzing interaction worksheets, surveys, and interviews.
What happened: By April 2026, researchers completed a year-long study of AI use among cancer specialists and found that initial operational gains hid intuition rust, described as the gradual dulling of expert judgment, which later evolved into chronic harms such as skill atrophy and identity commoditization.
By June 2026, a law review article described how generative AI had been used to create a viral 2023 track, 'Heart on My Sleeve,' that imitated Drake and The Weeknd. The label confirmed the vocals were not the artists but an AI imitation trained on their recordings, after millions had been convinced it was real.
ace