The government is investing in scanning and diagnostic equipment for 36 new and expanded Community Diagnostic Centres (CDCs).
The Download: the state of AI, and protecting bears with drones
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.
Want to understand the current state of AI? Check out these charts.
If you’re following AI news, you’re probably getting whiplash. AI is a gold rush. AI is a bubble. AI is taking your job. AI can’t even read a clock. Stanford’s 2026 AI Index—the field’s annual report card—cuts through the noise.
The data reveals a technology evolving faster than we can manage. From the China-US rivalry and model breakthroughs to public sentiment and the impact on jobs, here are the index’s key findings on the state of AI today.
—Michelle Kim
Why opinion on AI is so divided
Stanford’s 2026 AI Index is full of striking stats. It also reveals a field riddled with inconsistencies, most notably in the gap between experts and non-experts.
On jobs, 73% of US experts view AI’s impact positively, compared to just 23% of the public. Similar divides emerged on the economy and healthcare. What’s driving this disconnect?
Part of the answer may lie in their diverging experiences. Those using AI for coding and technical work see it at its best, while everyone else gets a more mixed bag. The result is two very different realities. Read the full story on what they are—and why they matter.
This story is from The Algorithm, our weekly newsletter on AI. Sign up to receive it in your inbox every Monday.
—Will Douglas Heaven
Job titles of the future: Wildlife first responder
Grizzly bears have made such a comeback across eastern Montana that in 2017, the state hired its first-ever prairie-based grizzly manager: wildlife biologist Wesley Sarmento.
For seven years, Sarmento worked to keep both bears and humans out of trouble. He acted like a first responder, trying to defuse potentially dangerous situations. He even got caught in some himself, which led him to a new wildlife safety tool: drones. Find out the results of his experiments in digital ecology.
—Emily Senkosky
This article is from the next issue of our print magazine, which is all about nature. Subscribe now to read it when it lands on Wednesday, April 22.
The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 Human scientists still trounce the top AI agents at complex tasks
The best agents perform only half as well as experts with PhDs. (Nature)
+ Can AI really help us discover new materials? (MIT Technology Review)
2 OpenAI is escalating its fight with Anthropic while pulling away from Microsoft
A leaked memo exposes plans to attack Anthropic. (Axios)
+ And says Microsoft “limited our ability” to reach clients. (The Information $)
+ While touting a budding alliance with Amazon. (CNBC)
3 Carbon removal technology is stalling—and that may be good news
Better solutions could now emerge. (New Scientist)
+ Here are three that are set to break through. (MIT Technology Review)
4 AI is finding bugs faster than we can fix them—and hackers will benefit
Welcome to the bug armageddon. (WSJ $)
+ AI may soon be capable of fully automated attacks. (MIT Technology Review)
5 A Texas man has been charged with the attempted murder of Sam Altman
He allegedly threw a Molotov cocktail at the OpenAI CEO’s home last Friday. (NPR)
+ The suspect reportedly had a list of other AI leaders. (NYT $)
6 AI is beginning to transform mathematics
It’s proving new results at a rapid pace. (Quanta)
+ One AI startup plans to unearth new mathematical patterns. (MIT Technology Review)
7 Students are turning away from computer science
It’s had a massive drop in enrollments. (WP $)
+ AI coding tools have diminished the degree’s value. (NYT $)
8 India’s bid to become a data center hub is sparking a fierce backlash
Farmers are protesting Delhi’s courtship of hyperscalers. (Rest of World)
9 Meta is set to overtake Google in advertising revenue this year
And become the world’s largest digital ad platform for the first time. (WSJ)
10 AI influencers are taking over Coachella
Synthetic content creators are “everywhere” at the festival. (The Verge)
Quote of the day
“These people are almost nothing like you. They are most likely sociopathic/psychopathic and, in the case of Altman, consistently reported to be a pathological liar.”
—The alleged firebomber of Sam Altman’s home shares his distrust of AI leaders in a blog post.
One More Thing
We’ve never understood how hunger works. That might be about to change.
A few years ago, Brad Lowell, a Harvard University neuroscientist, figured out how to crank the food drive to the maximum. He did it by stimulating neurons in mice. Now, he’s following known parts of the neural hunger circuits into uncharted parts of the brain.
The work could have important implications for public health. More than 1.9 billion adults worldwide are overweight, and more than 650 million are obese. Understanding the circuits involved could shed new light on why these numbers are skyrocketing.
—Adam Piore
We can still have nice things
A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line.)
Top image credit: Stephanie Arnett/MIT Technology Review | Getty Images
+ Someone built a mechanical version of Tony Hawk’s Pro Skater from Lego.
+ Enjoy this wholesome clip of toddlers discovering the existence of hugs.
+ This interactive body map shows exactly which exercises you need.
+ Jon McCormack’s photos of nature’s patterns are breathtaking.
A calibration-aware hierarchical CNN-SWIN fusion framework for robust Cross-Dataset brain MRI analysis
IntroductionDeep learning approaches have become central to brain MRI analysis; however, their reliability under dataset shift remains a critical barrier to safe and scalable deployment in neuroscience and clinical research. While convolutional neural networks (CNNs) provide strong locality-driven inductive biases for robust feature extraction, they lack global contextual awareness. Conversely, transformer-based architectures capture long-range dependencies but often exhibit reduced robustness and miscalibrated confidence when applied to heterogeneous medical imaging data, particularly in Cross-Dataset settings.MethodsIn this work, we propose a calibration-aware hierarchical CNN-Transformer fusion framework designed for robust brain MRI analysis under dataset shift. The architecture integrates a pretrained multi-scale CNN backbone with a hierarchical transformer branch and performs scale-aligned fusion through cross-attention mechanisms. By allowing local convolutional features to selectively query global contextual representations, the proposed design maintains stable feature contributions during fusion and mitigates overconfident reliance on transformer features when generalization degrades across datasets. The framework is evaluated using a strict Cross-Dataset protocol, where models are trained on one dataset and tested on a distinct dataset.ResultsExperimental results demonstrate that the proposed fusion model achieves competitive classification performance while substantially improving probabilistic calibration relative to both CNN-only and transformer-only baselines. Specifically, the model attains an average accuracy of 99.20% and achieves lower Expected Calibration Error (ECE = 0.0041), Brier score (0.0028), and Negative Log-Likelihood (NLL = 0.0277) compared to a standalone Swin Transformer and a strong ResNet50 baseline.DiscussionThese findings demonstrate that calibration-aware hierarchical CNN-Transformer fusion enhances both predictive reliability and robustness under Cross-Dataset evaluation. By improving the alignment between predictive confidence and empirical correctness, the proposed method supports safer large-scale analysis of heterogeneous brain MRI data, with important implications for multi-center neuroscience studies and trustworthy clinical decision support.
Toward a dual-pathway model of neuroplastic adaptation in sport: neural efficiency and neural optimization
Mitochondrial dysfunction, neuroinflammation, and associated mechanisms in sepsis-associated encephalopathy: from pathogenesis to emerging therapeutics
Sepsis-associated encephalopathy (SAE) is a devastating neurological complication of sepsis, leading to diffuse brain dysfunction, long-term cognitive deficits, and increased mortality. Its pathogenesis is complex, with mitochondrial dysfunction and neuroinflammation emerging as central, interconnected drivers. This review systematically elucidates the pathogenic crosstalk between these two processes. We detail how dysregulated mitochondrial dynamics (e.g., Drp1-mediated fission), impaired biogenesis (via the proliferator-activated receptor-gamma coactivator-1α axis), oxidative stress, and the activation of mitochondria-dependent cell death pathways (ferroptosis, pyroptosis) contribute to neuronal injury. Concurrently, microglial activation, particularly through the NOD-, LRR- and pyrin domain-containing protein 3 (NLRP3) inflammasome, creates a vicious cycle that exacerbates mitochondrial damage and synaptic loss. Furthermore, we summarize emerging therapeutic strategies that target this mitochondrial-neuroinflammatory axis, including molecular hydrogen, mitochondria-targeted peptides (SS-31), natural compounds, and specific inhibitors (e.g., Mdivi-1, MCC950). The integration of recent insights on the gut-brain axis and cerebral metabolomics further expands the therapeutic landscape. Ultimately, targeting this core axis offers a promising paradigm for developing effective interventions to improve neurological outcomes in septic patients.
Correction: The effect of musical sensory orientation training in improving consciousness level in patients with disorders of consciousness: a pilot study
Immersive virtual reality as a novel approach to improve social cognition in multiple sclerosis: an EEG-based pilot study
IntroductionMultiple sclerosis (MS) affects different cognitive domains, including social cognition. Immersive Virtual Reality (VR) may provide a novel rehabilitative approach to treat motor and cognitive symptoms of MS. This exploratory pilot study evaluated the effects of immersive VR rehabilitation on social cognition in MS patients and explored related cortical neurophysiological signatures.MethodsSeven MS patients underwent immersive VR rehabilitation with the CAREN system (3 sessions/week, approximately 45 min of active training per session, about 1 h including preparation, 8 weeks), while seven healthy controls (HC) did not undergo any intervention. Patients were evaluated at baseline (T0) and post-treatment (T1) with standardized measures of cognitive, emotional, and motor functioning. EEG data were acquired from all participants, and, after artifact removal, spectral parameterization decomposed signals into aperiodic (exponent, offset) and periodic oscillatory components (alpha and beta power). Power spectral density was analyzed using group comparisons and Pearson correlations with neuropsychological measures.ResultsCompared with HC, MS patients showed reduced alpha-band power, mainly over frontal and parieto-occipital regions, whereas aperiodic parameters did not differ between groups. In patients, alpha and beta power correlated with the Positive Emotions Self-Efficacy Scale (alpha: r = 0.92, p = 0.003; beta: r = 0.83, p = 0.020). Alpha power is also correlated with RAO SRT–LTS (r = 0.85, p = 0.016), and beta with EQ-CE (r = 0.82, p = 0.023). Overall, alpha and beta power were correlated with emotional self-efficacy, balance, memory, and empathy, suggesting that oscillatory markers are potential indicators of clinical outcomes.DiscussionRehabilitation via immersive VR has shown promising clinically significant effects in the cognitive, emotional, and motor domains, supported by convergent EEG spectral signatures. Future studies employing predictive modeling approaches will be required to assess their prognostic value.
How stressful life events are associated with depression: the mediating pathway of security in a clinical adolescent sample
BackgroundStressful life events are well-established risk factors for adolescent depression; however, the psychological mechanisms underlying this association remain insufficiently understood, particularly regarding which types of stress and which dimensions of security are most closely linked to depression. This study aimed to investigate whether security and its two sub-dimensions statistically mediated the association between stressful life events and depression among clinically diagnosed adolescents, while also examining the relative strength of indirect associations across specific stress types.MethodsA cross-sectional study was conducted with 284 adolescents (70.1% female; mean age = 15.82 ± 1.86 years) diagnosed with major depressive disorder according to the DSM-5 criteria at a tertiary psychiatric hospital in Western China. Participants completed the Adolescent Self-Rating Life Events Checklist (ASLEC), Self-Rating Depression Scale (SDS), and Security Questionnaire (SQ) questionnaires. Simple mediation, parallel mediation, and dimension-specific analyses were performed using the PROCESS macro (Model 4) with 5,000 bootstrap resamples, controlling for gender and parental marital status.Resultsstressful life events were significantly positively correlated with depression (r = 0.491, p < 0.001) and negatively correlated with security (r = −0.464, p < 0.001). Simple mediation analysis revealed that security demonstrated a significant indirect association through security (indirect effect = 0.176, 95% CI [0.126, 0.232]), accounting for 53.8% of the total association. Parallel mediation analysis further indicated a dual-pathway model: both Interpersonal Security (indirect effect = 0.083, 95% CI [0.037, 0.133]) and Certainty in Control (indirect effect = 0.093, 95% CI [0.043, 0.152]) functioned as significant statistical mediators of comparable magnitude, with no significant difference between them (Contrast = −0.010, 95% CI [−0.065, 0.042]). Furthermore, dimension-specific analyses revealed that Interpersonal Stress (standardized indirect effect = 0.266) and Academic Stress (standardized indirect effect = 0.231) showed the strongest indirect associations with depression through the security pathway. Exploratory subgroup analyses revealed a gender-crossed pattern: for male adolescents (n = 85), the indirect association was significant only through Interpersonal Security (effect = 0.116, 95% CI [0.048, 0.199]); for female adolescents (n = 199), it was significant only through Certainty in Control (effect = 0.136, 95% CI [0.067, 0.212]).ConclusionSecurity functions as a significant statistical mediator in the association between stressful life events and adolescent depression. The findings are consistent with a “dual-pathway” model wherein stress is concurrently associated with lower levels of both relational security (Interpersonal Security) and personal agency (Certainty in Control). Exploratory analyses suggest that the relative importance of these two pathways may differ by gender. If confirmed by future longitudinal research, clinical interventions may benefit from an integrated approach that addresses both dimensions, with particular attention to interpersonal conflicts and academic pressure as the stressors most strongly associated with depression through security pathways.
A longitudinal analysis of the prevalence of restrictive interventions involving women with mental health conditions, learning disabilities or autism in mental health services in England
IntroductionRestrictive interventions, including physical restraint, seclusion, chemical restraint, and segregation, continue to be used within mental health services, despite sustained policy efforts to promote least-restrictive and trauma-informed care. However, little is known about national trends affecting women, for whom restrictive interventions often carry heightened risks of re-traumatisation and stigma.MethodsWe conducted a longitudinal secondary analysis of publicly available administrative data from the Mental Health Bulletin covering NHS-funded mental health services in England between 2017 and 2025. Annual counts of restrictive interventions involving women were examined relative to the number of women detained under the Mental Health Act to estimate annual rates per 1,000 detained. Regression modelling was used to assess temporal trends overall, by age group and type of restrictive intervention, and interrupted time-series analyses to examine changes following implementation of the Mental Health Units (Use of Force) Act 2018 (“Seni’s Law”). Trends were also examined alongside available national data on restrictive interventions involving men.ResultsRates of restrictive interventions involving women increased by approximately 12 percent per year over the study period, with no evidence of a reduction following the introduction of Seni’s Law. Increases were most pronounced for chemical restraint, seclusion, and segregation, while physical and mechanical restraint remained stable. Restrictive interventions declined among women under 18 but increased consistently across all adult age groups, indicating a widening age-related divergence. Although overall trends broadly mirrored those observed among men, the types of restrictive interventions used and their potential impact may differ, highlighting gendered dimensions in how restrictive practices are experienced and applied.DiscussionDespite extensive national initiatives, restrictive interventions involving women have continued to rise in England, highlighting a persistent gap between policy intent and practice. The findings suggest that legislative frameworks alone are insufficient to achieve meaningful reductions without operational changes in clinical practice, organisational culture, and monitoring systems. Internationally, the study contributes rare gender-disaggregated longitudinal evidence and highlights the need for comparable monitoring systems and coordinated research to inform rights-based, trauma-informed strategies to reduce restrictive interventions in mental health services.
Self images: an empirical enquiry into Rembrandt’s self-portraits
Many have speculated that events of personal and financial loss in the life of Rembrandt van Rijn (Rembrandt) caused depression and that this is revealed by examination of his work particularly self-portraits painted in old age. Some report detecting various physiological diseases associated with aging, including vision impairment, which may have affected his mood and work. Aging and neurodegenerative disease which often accompanies it, are both associated with depression. Depression is characterised by visual deficits including perception of reduced contrast and colour. Age-related neurological disorders are associated in artists with reduced complexity. Recent advances in imaging and computer technology make it possible to empirically examine changes in artistic style which can contribute to understanding artists’ physical and mental health. Previous studies have identified associations between adverse events in artists’ lives and altered contrast and colour in their self-portraits. In the current study changes in contrast, colour and fractal dimension were measured in the entire corpus of Rembrandt’s painted self-portraits and portraits to determine whether changes in style indicate depression, cognitive decline, or neurological disease and whether differences in style can be detected between self-portraits and portraits of related and unrelated others. Productivity was also examined as an indirect indicator. The results suggest that it is unlikely that Rembrandt suffered from unipolar or bipolar depression, age-related cognitive decline, or neurodegenerative disease. The data are consistent with someone experiencing episodes of low mood associated with normal grieving and adversity followed by resilient recovery. There is evidence of a gradient in saliency and complexity between self-portraits and related and unrelated portraits and of a ‘late’ style identified by leading art historians consistent with macular degeneration.