Background: In recent years, artificial intelligence (AI) has ushered in a promising era in medicine, particularly in medical education. However, studies assessing the knowledge, attitudes, and practices related to AI among medical students in Vietnam remain limited. Objective: This study aimed to evaluate AI knowledge, attitudes, and practices among Vietnamese medical students in learning and research, and to identify factors associated with their AI practices. Methods: A cross-sectional study was conducted among medical students at Thai Binh University of Medicine and Pharmacy from November to December 2025. Data were collected using an online structured questionnaire covering demographic characteristics and AI knowledge, attitudes, and practices. The main outcome of interest was AI practices in learning and research. Descriptive statistics and multivariable linear regression were used to examine associated factors. Regression coefficients (β), 95% CIs, and values are reported. Results: A total of 1002 medical students (mean age 21.00, IQR 19.00-23.00 years; n=596, 59.5% female) were included. The median percentage of maximum possible (POMP) score of AI knowledge was 66.67 (IQR 33.33‐83.33), with a high level of familiarity with common tools (n=798, 79.6%). AI attitudes were generally positive (median POMP score 70.00, IQR 53.33‐76.67). AI-related practices were lower (median POMP score 50.00, IQR 46.88-71.88), with AI being used primarily for information retrieval and literature research support. In the multivariable analysis, knowledge POMP score (β=0.12, 95% CI 0.08-0.16) and attitudes POMP score (β=0.42, 95% CI 0.34-0.51) were significantly associated with AI practices POMP score (<.001). Age, gender, major, grade point average classification, and having participated in an AI seminar or training were not associated with AI practices. Conclusions: Medical students showed favorable knowledge and positive attitudes, but their AI practices remained limited. Integrating AI into medical curricula, including fundamentals, applications, and ethical aspects, is essential to prepare future physicians for AI-driven health care.
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
The UK’s generational tobacco ban might not work. I’m supporting it anyway.
—Jessica Hamzelou
As the parent of two little girls, I often think about how their childhood is different from mine. The seven-year-old is learning about AI at school. The five-year-old is given internet-based homework every week. And they are both absolutely repulsed by the idea of smoking.
That was not the prevailing sentiment when I was young. Smoking was a central part of our culture. Which is why the UK’s recent passing of a generational sales ban on tobacco products feels like such a big deal.
This is what’s described as an “endgame” approach. While many tobacco control strategies—such as taxation or gory imagery—aim to reduce consumption, policies like the UK’s are designed to eliminate it entirely. It’s a new approach, and no one knows whether it will work. But it’s an enticing prospect—and it’s starting to look a lot less radical.
This story is from The Checkup, our weekly biotech newsletter. Sign up to receive it in your inbox every Thursday.
You do your own time
—You do your own time is a short story by Elizabeth Bear, an award-winning speculative fiction author.
There we were, a regular murderers’ row of librarians. Turning around in the nave of our library to greet the sound of footsteps, pistols leveled in case whoever was coming in didn’t respect sanctuary.
I pulled down a solid-state drive full of biographies and case studies of people who had spent time—and sometimes their whole lives—in labor camps or chattelhood. It was illegal to possess, and the feds used smart agents to track down and obliterate any copies. Which was why we were sending one to the stars.
What’s left behind when a name is erased from the system? No legacy, no memory—that is the point of media and narrative control. So that was our plan: to preserve it, for later generations, or just as a silent record of our existence.
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 An EU lawmaker investigating spyware was hacked by that spyware Citizen Lab found Pegasus spyware on Stelios Kouloglou’s phone. (Wired $) + It said the EU “looks the other way” on spyware abuses. (Guardian) + Meet the director of Citizen Lab. (MIT Technology Review)
2 Anthropic is closing loopholes that allow Chinese access to Claude It’s targeting VPNs, relay services, and overseas accounts. (FT $) + Users in China keep finding new workarounds. (Wired $)
3 A Tesla driver has been charged with manslaughter after a fatal crash Court records show he was using automated driver-assistance. (WSJ $) + Tesla sales have surged 25% after a rebound in Europe. (NYT $)
4 Trump bought lots of tech stock the day he unveiled his AI Action Plan He acquired up to $5 million in stock from Amazon and others. (Engadget) + His AI Action Plan was a distraction. (MIT Technology Review)
5 Companies are throttling employees’ AI use because it’s too expensive They’re pleading with workers to use less powerful models. (404 Media) + Tesla has capped their AI spending at $200 per week. (The Information $)
6 The Energy Dept wants data centers on backup power in heat waves It wants them to free up power for AC. (NYT $) + People near data centers are dreading heat wave pollution. (Politico $) + No one wants a data center in their backyard. (MIT Technology Review)
7 A Meta glasses feature just went from free to a subscription service “Conversation Focus” will now cost $19.99 per month. (BBC) + The move heralds a new era of consumer tech subscriptions. (Wired $)
8 Random wobbles in time could solve gravity’s greatest mystery A new idea could reconcile gravity and quantum mechanics. (New Scientist $)
9 Peter Thiel claims the pope is “working for the Chinese Communists” By pushing for stricter AI rules that may benefit Chinese interests. (CNN) + Pope Leo XIV said AI must be “disarmed” in his first major teaching. (BBC) + His encyclical offered a template for steering AI. (MIT Technology Review)
10 Supersonic flight over land could finally be legal again Regulators want to lift a ban—so long as the planes are quiet. (Ars Technica)
Quote of the day
“We don’t have robots that are nearly as good at understanding the physical world as a rat.”
—Yann LeCun, the founder of AMI Labs and Meta’s former chief AI scientist, tells the BBC that AI isn’t as smart as many think.
One More Thing
MARCO GIANNAVOLA
How two brothers became go-to experts on America’s “mystery drone” invasion
On a Friday evening in December, every tier of US law enforcement was dispatched to a military research installation outside Boston after a squadron of 15 to 20 drones was spotted violating restricted airspace. The culprits could not be found.
It was the latest in a series of purported drone sightings along the US East Coast. Lacking coordination or clarity from the White House, the Pentagon, and the intelligence community, law enforcement officers turned to an unlikely source: twin brothers from Long Island who hunt UFOs.
The Tedescos have built a mobile field lab to investigate unexplained aerial phenomena. Now members of the FBI want their support.
High altitude exposes the brain to heterogeneous hypoxic, hemodynamic, rheological, inflammatory, and healthcare-access conditions. This heterogeneity makes altitude biologically informative for stroke research, but it does not justify treating natural altitude exposure as a single protective or harmful state. In this structured narrative review, we searched and organized the literature to ask which altitude-associated hypoxic cues resemble or reveal mechanisms compatible with cerebral ischemic tolerance, and what level of evidence supports that claim. We separate long-term adaptation, short-term acclimatization, chronic or excessive environmental hypoxia, and experimental hypoxic conditioning; define direct, supportive, and indirect evidence tiers; and integrate neurovascular-unit biology with multi-omics and stroke pathophysiology. Experimental hypoxic preconditioning remains the clearest direct evidence that a defined sublethal hypoxic stimulus can induce a time-limited tolerant state. In contrast, human high-altitude epidemiology, physiology, and genetics mainly constrain the clinical context and nominate candidate pathways rather than prove stroke-specific protection. We also emphasize that chronic hypoxia can be maladaptive through endothelial dysfunction, oxidative stress, erythrocytosis, thrombogenicity, blood–brain barrier impairment, and microvascular injury. Across neurovascular-unit cell types, a transparent evidence-weighting framework prioritizes endothelial biology because of its direct connection to BBB stability, effective reperfusion, hemorrhagic transformation risk, and no-reflow, while neurons, astrocytes, microglia, oligodendrocyte-lineage cells, and pericytes require different degrees of causal and human validation. We argue that the most productive path forward is not to label altitude as protective, but to use altitude-related biology to prioritize testable, stroke-facing hypotheses regarding BBB stability, microvascular patency, metabolic support, inflammatory thresholds, white-matter resilience, and biomarker-defined conditioning windows.
Transducer architecture is an increasingly important design consideration in transcranial focused ultrasound, where focal precision, electronic steering, skull-aberration correction, workflow complexity, and cost must be balanced. Under matched aperture, transmit frequency, focal depth, effective source geometry, and equivalent aberration correction, subdividing an aperture into multiple independently driven elements does not by itself reduce the diffraction-limited focal volume. Phased arrays provide important capabilities that single-element transducers generally lack, including electronic steering, dynamic phase correction, and flexible control of energy deposition. These capabilities can improve the realized intracranial field, particularly when skull-induced aberrations would otherwise broaden, shift, or split the focus. Single-element systems may also be extended through patient-specific acoustic holograms or lenses, which impose spatially varying phase delays to compensate for aberrations, steer the focus, or shape the pressure field. These approaches may reduce system complexity but are typically static and less adaptable than phased arrays. We therefore frame transducer selection as an application-driven design-matching problem: the relevant question is not which architecture is inherently “more focused” under matched conditions, but which best meets a study’s requirements for steering, skull-aberration correction, temporal reconfigurability, spatial degrees of freedom, workflow burden, cost, positioning robustness, and safety.
BackgroundThe mammalian suprachiasmatic nucleus (SCN) serves as the master circadian pacemaker, which coordinates daily behavioral and physiological rhythms through functionally diverse neuronal subtypes. Cholecystokinin (CCK) is expressed in a subset of SCN neurons; however, its role in locomotor activity rhythms remains poorly understood.MethodsTo study the functional contribution of SCN CCK-expressing (SCNCCK) neurons, we selectively blocked synaptic transmission by injecting a Cre-dependent tetanus toxin (TeNT) viral vector into the SCN of CCK-IRES-Cre mice. Before and after the virus injection, spontaneous locomotor activity was continuously recorded under a 12:12 h light–dark (LD) cycle. Subsequently, we used Cre-dependent fluorescent reporter (mYongHong) to label SCNCCK neurons and performed whole-brain projection mapping to characterize their downstream connectivity.ResultsSynaptic inhibition of SCNCCK neurons significantly attenuated the strength of locomotor rhythmicity, resulting in reduced rhythm organization and a more uniform distribution of activity. This disruption was mainly driven by a significant decrease in dark-phase locomotor activity, while light-phase activity remained unchanged. Anatomically, SCNCCK neurons are widely projected along the anterior and posterior axes to multiple hypothalamic, thalamic, and limbic regions, including the medial preoptic area (MPA), paraventricular thalamic nucleus (PVT), paraventricular hypothalamic nucleus (PVH), anterior hypothalamic area (AHC), dorsomedial hypothalamic nucleus (DMH), ventromedial hypothalamic nucleus (VMH), and medial amygdala nucleus (MeA). Quantitative analysis revealed projections to these downstream regions, with moderate variation in projection density across targets.ConclusionTogether, these findings identify SCNCCK neurons as an important neuronal subpopulation, which contributes to the robustness and consolidation of spontaneous locomotor rhythms, likely through a wide range of downstream circuits.
Neurogenic pulmonary edema (NPE) is a life-threatening complication of acute central nervous system (CNS) injury, characterized by the rapid onset of hypoxemia and pulmonary fluid accumulation in the absence of underlying cardiopulmonary disease. In recent years, emerging integrative frameworks such as the “neuroimmunoaxis” and “brain-lung axis” have provided new perspectives on how CNS injury leads to systemic immune dysregulation and pulmonary dysfunction. However, critical questions remain regarding the interplay between excessive sympathetic activation, immune homeostasis disruption, and lung tissue injury. This narrative review proposes a neurotransmitter-immune-inflammatory model that integrates mechanical, adrenergic, and inflammatory pathways across the spatiotemporal evolution of NPE. We identify four progressive stages involving sympathetic storm initiation due to central autonomic network disinhibition, pulmonary vascular barrier disruption through Piezo channel activation and angiotensin II-norepinephrine synergy, inflammatory amplification from loss of the cholinergic anti-inflammatory reflex, and systemic progression involving gut-lung axis dysregulation. The model generates three testable predictions. Lesions disrupting the nucleus tractus solitarius-ventromedial hypothalamus-intermediolateral column projection should produce more severe NPE. And selective activation of TRPA1+ dorsal root ganglion neurons should attenuate sympathetic outflow and pulmonary edema. Enhancing α7 nicotinic acetylcholine receptor signaling should mitigate systemic inflammation. These predictions offer experimental avenues for validating the hijacking hypothesis. Translational implications include stage-specific interventions, early sympathetic blockade, mid-phase anti-inflammatory and neuro-modulatory strategies, and late-stage lung-protective ventilation. This study aims to offer a comprehensive analysis of NPE by exploring its pathological mechanisms—from central sympathetic signaling to peripheral lung damage. Emphasis is placed on examining the interactions between neural signals, neurotransmitters, and immune responses to uncover the spatiotemporal dynamics of NPE. By identifying potential pathways for early diagnosis and targeted therapies, the research seeks to improve disease management and contribute to better clinical outcomes for affected patients.
ObjectiveBody dysmorphic disorder (BDD) is a prevalent concern among young adults. However, the underlying mechanisms of BDD development remain elusive. This study aims to investigate the intricate relationship between attachment styles and BDD symptoms, with appearance-based rejection sensitivity (ARS) as a mediating factor and gender as a moderator.MethodsA total of 815 young adults participated, completing a battery of questionnaires including the Revised Adult Attachment Scale (RAAS), Appearance-Based Rejection Sensitivity Scale (ARSS), and Scale of Body Image (SBI).ResultsData indicated a positive association between attachment anxiety and BDD symptoms, with ARS found to mediate this link. Furthermore, gender differences were observed to moderate the relationship between ARS and BDD symptoms.ConclusionThis study sheds light on the foundational mechanisms of BDD, tracing its origins to early caregiver-infant bonds and highlighting the enduring impact of ambivalent care on body image perceptions. Additionally, the identification of ARS as a specific contributing factor to BDD onset underscores its significance in understanding and addressing this disorder. By considering the influence of social norms and cultural context, gender differences in the association between ARS and BDD symptoms are elucidated.
BackgroundExercise is increasingly discussed as part of lifestyle-based and multimodal care for mood and anxiety disorders, but review-level evidence often mixes formally diagnosed clinical populations with symptom-defined or medically mixed samples.MethodsWe conducted an umbrella review of systematic reviews, meta-analyses, and network meta-analyses of structured exercise interventions for adults with depressive or anxiety disorders. Six databases were searched from inception to 1 March 2026. Primary outcomes were depressive and anxiety symptom severity, remission, and response; secondary outcomes were acceptability and tolerability. Review quality was appraised with AMSTAR 2, and primary-study overlap was quantified with corrected covered area (CCA), including overall and symptom-cluster analyses. The synthesis was designed to summarize review-level credibility and clinical interpretability rather than to generate a second-order pooled efficacy estimate.ResultsNine reviews met eligibility criteria; four supplied directly extractable primary overall review-level estimates for core psychiatric symptom outcomes. AMSTAR 2 appraisal rated one review as high, three as low, and five as critically low. Recalculated overall overlap was slight (112 primary-study occurrences, 89 unique primary studies; CCA = 3.23%), although cluster-level analyses identified localized redundancy, particularly within anxiety-disorder-specific reviews. In major depressive disorder, one clinically focused review reported a large reduction in depressive symptoms for aerobic exercise versus non-exercise comparators (Hedges’ g = -0.79, 95% CI -1.00 to -0.57; I² = 21%). Across diagnosed depressive and/or anxiety disorders, broader review-level estimates also favored exercise for depressive symptoms (SMD = -0.97, 95% CI -1.28 to -0.66) and anxiety symptoms (SMD = -0.66, 95% CI -1.09 to -0.23), but heterogeneity was high. Anxiety-disorder-specific evidence was less secure: the primary DSM-IV anxiety-disorder pooled estimate showed no clear benefit over selected controls (SMD = 0.02, 95% CI -0.20 to 0.24). Acceptability estimates were close to null, and adverse-event reporting was too sparse to support confident safety conclusions.ConclusionExercise is best supported as an adjunctive, patient-centered component of care for depressive disorders. Anxiety-disorder-specific efficacy remains uncertain when comparator rigor, diagnostic heterogeneity, and localized overlap are considered, and safety reporting needs substantial improvement.Systematic Review Registrationhttps://www.crd.york.ac.uk/PROSPERO/, identifier CRD420261364264.
ObjectivesThis study examined whether socioeconomic status was associated with anxiety symptoms, depressive symptoms, BMI, waist-to-hip ratio, and quality of life among Brazilian adults living in areas of moderate social vulnerability. In addition, we described anxiety and depressive symptoms, BMI and waist-to-hip ratio, and quality of life in individuals living in moderate social vulnerability.MethodsThis is a cross-sectional study. In a socially vulnerable cohort, interviews captured demographics, comorbidities, medications, anxiety and depressive symptoms, and quality of life, followed by measurement of anthropometric characteristics.ResultsAmong 299 socially vulnerable adults, 8% had moderate–severe depressive symptoms and 7% had moderate–severe anxiety symptoms; ~50% showed increased risk of cardiometabolic diseases (i.e., waist-to-hip ratio greater or equal to 0.90 for men and 0.85 for women, respectively). Poor quality of life affected 4–12% across domains. Mental health, anthropometrics (waist-to-hip ratio, BMI), increased risk of cardiometabolic diseases and quality of life in physical, social and environmental domains did not differ by socioeconomic status (B, C, DE; all P>0.05). Poor psychological quality of life was more frequent among participants in higher socioeconomic status (B: 8%; C: 6%; DE: 4%, P = 0.0157). Linear regression analyses showed no statistically significant differences across socioeconomic status in depressive symptoms, anxiety symptoms, BMI, waist-to-hip ratio, or quality of life scores in any domain (all P>0.05).ConclusionsOur findings suggest that, among Brazilian adults living in moderate social vulnerability and classified within socioeconomic status B, C, and DE, mental health, BMI, waist-to-hip ratio, and quality-of-life indicators were similar across socioeconomic strata. However, these results should be interpreted as reflecting intra-group socioeconomic differences within a moderately vulnerable population and should not be generalized to individuals from the highest socioeconomic status.
Background: Medical interview training faces limitations in both implementation and evaluation. While generative artificial intelligence (GenAI) offers a potential solution, it remains unclear whether reasoning-oriented models improve evaluation, particularly for the Japanese language. Objective: We assessed scoring patterns to evaluate the agreement between reasoning-oriented GenAI model scores and clinical educator consensus ratings in Japanese medical interview training. Methods: This study was conducted at a medical university in Japan using original Japanese-language text data derived from medical interview training. Postgraduate year 1 and 2 residents were involved. Two blinded human clinical educators independently evaluated the transcripts and reached consensus through discussion. These consensus ratings were used as a practical reference standard, while preconsensus agreement was also assessed to characterize interhuman variability. Two GenAI models, GPT-5.2 Thinking (OpenAI) and Gemini 3.0 Pro (Google LLC), independently evaluated the same transcripts directly. Each GenAI model generated a single zero-shot evaluation per transcript using default settings. All evaluations used a standardized 6-domain Objective Structured Clinical Examination rubric (patient care, history taking, physical examination, accuracy and organization of clinical information, clinical reasoning, and management) scored on a 1‐6 Likert scale, where 1 indicates inferior, and 6 indicates excellent. We compared mean evaluation scores using the Wilcoxon signed-rank test and assessed interrater reliability using intraclass correlation coefficients between the GenAI models and the clinical educators. Results: Clinical educators and both GenAI models rated the entire dataset of 40 transcripts by 20 included residents. Clinical educator consensus ratings yielded the highest overall mean scores (5.18, 95% CI 5.06 to 5.30). Comparatively, both GenAI models demonstrated significantly lower scores: GPT-5.2 Thinking assigned the lowest overall score (3.68, 95% CI 3.62 to 3.72; <.001), followed by Gemini 3.0 Pro (4.09, 95% CI 3.97 to 4.21; <.001). This discrepancy was most pronounced in the management domain, where GPT-5.2 Thinking assigned 2.93 (95% CI 2.79 to 3.06) compared with the clinical educator consensus mean score of 5.20 (95% CI 4.91 to 5.49). Agreement between the GenAI models and the clinical educator consensus ratings was poor across all domains, with overall intraclass correlation coefficients of 0.04 (95% CI 0.00 to 0.09) for GPT-5.2 Thinking and 0.22 (95% CI 0.10 to 0.35) for Gemini 3.0 Pro. Conclusions: In this preliminary, single-center, transcript-based Japanese-language study, single-run zero-shot evaluations by GPT-5.2 Thinking and Gemini 3.0 Pro showed lower scores and poor agreement with the clinical educator consensus ratings. These findings should be interpreted cautiously because multiple outputs, prompt-sensitivity analyses, local validation, and model-parameter comparisons were not performed. Under the specific conditions tested, these models should not be used as standalone evaluators for Japanese Objective Structured Clinical Examination medical interview transcripts. Whether these models can provide useful formative feedback remains a hypothesis. Trial Registration: UMIN-CTR Clinical Trial UMIN000053747; https://center6.umin.ac.jp/cgi-open-bin/ctr_e/ctr_view.cgi?recptno=R000061336
