AI News Archive: June 2, 2026 — Part 15
Sourced from 500+ daily AI sources, scored by relevance.
- GetVidya
AI mock tests for SSC CGL that adapt to your weak spots
- The Awkward Translator
Type the gist, get 4 ways to say the hard thing
- VERZIO
The AI-Built Application Marketplace
- UXXCA
AI powered Invoice Manager for Gen Z founders
- BriefIQ — Stop Guessing. Start Ranking.
From keyword to A-grade article in minutes
- Mediscanner
Free, privacy-first symptom checker & live disease map
- B2B Pricing Calculator | AI Margin Tool
B2B landed cost calculator. 8 Incoterms. 6 languages. Free.
- Bloomind
On-device AI that structures your thinking. Fully offline.
- AI Kit
One command to install the best local LLM for your Mac
- FinSeam
Autonomous reconciliation for modern finance teams
- AiCostNest
Estimate LLM API costs before launching AI features
- OptionMAV
AI options screener on a USB. Own it forever.
- TrendMing
Track news, AI tools & create video scripts in one click
- DM CINEMATIC: AI Production Engine
High-End AI Video. Look like $3k. Delivered in 24h
- Catalyze - Monetize your Content
We make AI companies grow and real creators earn.
- mlx-code
Local coding agent for Mac with git-backed subagents
- 10MW Liquid Cooled AI Data Center
Built for Blackwell, Grace Blackwell & Vera Rubin
- MatchMind
Calibrated football match probabilities — no tips, ever.
- Zypact
AI visibility audits across ChatGPT, Claude & Perplexity
- PromptEase
Build AI Prompts, Systems & Workflows
- LegacyCode MRI
AI-powered MRI scanner for your codebase
- Modellix
One Unified MaaS Platform for All AI Media Generation
- Leania.ai for AI consultants
Stand out and win more clients will personalised AI recommendations
- AI Image Combiner
Combine two images into one stunning composition.
- Dokie AI
Dokie AI: AI Presentation Maker for Business-Ready PowerPoint Slides Dokie AI is an AI presentation maker and AI PPT generator designed for business-ready slide creation. With Dokie AI, users can quickly turn ideas, text, documents, PDFs, Word files, Excel data, and web content into professional PowerPoint presentations. Create Professional Presentations for Business, Marketing, and Training […]
- A Pan-Cancer Multi-Omic SuperLearner for Regulated Cell Death Survival Topologies
Introduction: Regulated cell death (RCD) pathways profoundly influence tumor progression and immune modulation. In prior work, we constructed a comprehensive database mapping 25 forms of RCD across seven multi-omic layers encompassing 33 tumor types (CancerRCDShiny). Despite their robust ability to identify risk populations, translating these prognostic signatures into personalized clinical workflows requires a shift from generalized cohort stratification to individualized risk mapping. This necessitates mapping the complex geometric landscape of patient risk - Survival Topologies - to accurately capture the non-linear dynamics of RCD signatures. Methods: We engineered a Pan-Cancer Multi-Omic SuperLearner pipeline evaluating 33 cancer types. Phase I performed zero-leakage data harmonization and groupwise imputation to prevent cross-cohort amalgamation. Phase II utilized Elastic Net - regularized Cox (CoxNet) regression as an audit-compliant CANARY diagnostic to map mathematical proportional-hazards failures. Admissible strata enforcing a rigid 35% topological missingness barrier entered Phase III, deploying an advanced non-linear Quadripartite Base-Learner Ensemble (Random Survival Forests (RSF), Extreme Gradient Boosting (XGBoost), insulated Survival-Boruta, and Multi-Task Logistic Regression (MTLR)) - fused within an Elastic Net Multi-View Meta-Learner (MVL) - with local interpretability guaranteed via post-hoc SHAPley Additive exPlanations (TreeSHAP) and Local Interpretable Model-agnostic Explanations (LIME). Results: The CANARY diagnostic empirically proved the structural invalidity of pan-cancer geometric proportional-hazards. Advancing 96 verified matrices into the Quadripartite Machine Learning Ensemble, Phase III executed a structural algorithmic displacement: dense continuous multi-omic topologies computationally suppressed static genomic mutations and Copy Number Variations (CNVs) during multidimensional competition (85.7% vs 0.0% apex retention). Furthermore, the MVL stabilized global predictions against extreme biological variance, while surrogate LIME validations (R-squared < 0.10) confirmed the absolute failure of linear interpretative proxies. Extracting N-dimensional TreeSHAP interactions natively bypassed generalized risk parameters, mapping exact Survival Topologies. This dynamically exposed multi-omic synergistic (lethal peaks) and antagonistic (protective valleys) rescue trajectories invisible to additive models. We integrated this architecture into CancerRCDPredictor, a Shiny application operating as a digital tumor board. Conclusion: Deploying a Pan-Cancer Multi-Omic SuperLearner to bypass linear topological failures, this study advances beyond generalized cohort stratifications, establishing a deterministically mapped architecture for predicting RCD-related Survival Topologies. Through the CancerRCDPredictor interface, we directly translate multi-omic insights into individualized precision oncology interception.
- Rank dependency of rescaled pruning in recurrent neural networks
Throughout development and maturity, neural circuits undergo massive synaptic pruning, yielding highly sparse connectivity while preserving robust population-level computations. These population dynamics are often low-dimensional, allowing task-related computations to be formalized as trajectories within latent subspaces. How such low-dimensional dynamics are preserved amid widespread network sparsification remains unclear. Here, we investigate how different synaptic pruning rules shape low-dimensional dynamics and task performance in recurrent neural networks (RNNs). Moving beyond previous approaches focused on random sparsification of low-rank networks or networks with strictly constrained structures, we systematically evaluate how biologically motivated pruning rules interact with a network's underlying rank. We show that post-pruning dynamics and task performance depend critically on the network's initial rank due to distinct eigenspectral characteristics across rank regimes. Combining mathematical analysis with simulations, we demonstrate that pruning with synaptic rescaling preserves low-dimensional dynamics with minimal distortion in low-rank RNNs, but degrades in the high-rank regime. Our findings suggest that low-rank structure, combined with homeostatic synaptic rescaling, is essential for maintaining stable, low-dimensional dynamics in sparse networks.
- Equitable Health Intelligence: An Open Benchmark of Multi-Population Machine Learning for Omics-Based Cancer Prognosis
Purpose: Machine learning (ML) models for omics-based cancer prognosis are often trained on data from predominantly European-ancestry populations, producing biased predictions for other populations and undermining equitable genomic medicine. Existing fairness benchmarks mainly focus on outcome parity rather than predictive performance parity across populations. Public benchmark resources are needed for systematically detecting and mitigating such performance disparities in multi-population cancer prognosis. Methods: We developed Equitable Health Intelligence (EHI, https://ehiportal.org), an open-source benchmark of multi-population ML for omics-based cancer prognosis. EHI contains 1,475 ML tasks across 40 cancer/pan-cancer types, 4 omics feature sets, 4 clinical endpoints, 5 event-time thresholds, and 3 data-disadvantaged population (DDP) groups relative to a majority European Ancestry population group. Deep neural network models are trained under three multi-population ML schemes (Mixture, Independent, and Transfer Learning), with Naive Transfer included as a no-adaptation control, comprising a total of 10,325 ML experiments. Results: The EHI platform provides an interactive environment with visualization and exploratory tools for users to inspect predictive performance disparities between the majority European-ancestry group and data-disadvantaged populations, evaluate the extent to which transfer learning mitigates these disparities, and examine the impact of feature engineering methods across cancer types, omics features, and clinical endpoints. Conclusion: EHI is an open, interactive, and extensible benchmark for identifying and addressing performance disparities in multi-population ML for omics-based cancer prognosis. It provides a foundation for a growing ecosystem of methods targeting ML performance disparities arising from biomedical data inequality and population-level distribution shifts, thereby advancing equitable AI in precision oncology.
- A multimodal perturbation atlas defines the phenotypic resolution of cellular morphology.
Because cells are complex dynamical systems, modeling cellular behaviors requires methods that capture how cells evolve across time, environments, and interventions. Microscopy is uniquely suited to this goal in that it can be applied to living cells in their native context. However, the phenotypic resolving power of live-cell microscopy remains incompletely characterized, particularly relative to molecular assays. Here, we present a multimodal perturbation atlas of 1,000 pooled CRISPR knockouts in A549 cells, profiled by fluorescence microscopy (39 live, 13 fixed markers), label-free phase imaging of the same live cells, and single-cell RNA sequencing (scRNA-seq). Totaling ~57 million single-cell profiles, our data yield rich cell-biological signatures that map individual gene function. We find that phase imaging matches -- and, with sufficient cell coverage, exceeds -- the phenotypic resolution of fluorescence imaging and scRNA-seq, while capturing higher-order pathway organization that scRNA-seq does not resolve. These results establish intrinsic morphology as a high-precision readout of cellular state, and lay a foundation for live-cell profiling of phenotypic trajectories.
- Information-dependent eye-hand coordination emerges from active vision
In daily activities, humans rely on visual information to plan hand movements, making the extraction of task-relevant information through eye gaze a key aspect of motor control. Behavioral studies have revealed characteristic saccade-pursuit patterns, likely governed by shared neural circuits, which enable an efficient reduction of task-related uncertainty. However, a unifying computational principle explaining the emergence of these patterns in continuous tasks such as reading or driving is still lacking. Here we propose a dual stochastic model predictive control formulation of active vision, in which eye movements are continuously controlled to minimize task-relevant uncertainty and build an internal model used for hand movement planning. Through experiments manipulating the amount, density, and difficulty of future visual information, we show how eye movement patterns adapt to the information context while maintaining an invariant extraction horizon. A saccade-pursuit pattern naturally emerges from the model, which accurately predicts both eye and hand movement features observed in experiments. These results provide a principled framework for understanding the continuous regulation of human eye movements and open new perspectives for applications in robotic assistance and active perception.
- Efficient ageing: Simulated lesion of the structural connectome reveals optimised decline in the healthy ageing brain
Healthy ageing is associated with widespread white-matter change and altered connectome organisation, yet the link between local microstructural decline and whole-brain network communication remains unclear. Here we combined tract-based spatial statistics (TBSS) with probabilistic tractography and graph analysis to quantify the connectome-level consequences of age-sensitive white-matter hotspots. In two independent diffusion MRI datasets of healthy young and older adults (n = 144 total; Dataset 1: 77 participants; Dataset 2: 67 participants), we first identified age-sensitive fractional anisotropy clusters and then used them as constraints in a cross-dataset simulated-lesion framework. This allowed us to estimate how strongly each structural connection depended on age-sensitive tissue and to test the resulting network effects against matched-mass null lesions. Across both datasets, ageing hotspot lesioning reduced global efficiency, but consistently less than expected under the null model, indicating a less-than-random disruption of network integration. Age-sensitive tissue was disproportionately embedded in the integrative backbone of the brain: proportional degree loss was strongest in high-degree nodes, rich-club connections showed the greatest hotspot dependence, and nodewise losses were concentrated in frontal, cingulate and subcortical association systems, whereas posterior sensory and temporo-limbic regions were relatively spared. These findings suggest that healthy ageing reflects a selective and constrained reconfiguration of structural connectivity rather than simply pointing to uniform decline.
- Augmenting Structured Diagnoses through Effective Use of Pre-trained Large Language Models on Clinical Notes
Objective Clinical narrative provides a unique window into provider reasoning and attribution, but use has been limited by resource requirements and extensive fine-tuning, and LLMs in particular have traditionally not performed well at medical coding. We optimize and evaluate a reproducible method for automated diagnosis assignment using LLMs in clinical notes and compare with EHR structured diagnoses. Methods We used GPT-OSS for prompt engineering and task segmentation to create a model that extracts ICD-10-CM diagnoses, with estimates of severity, currency, and importance, from progress notes. We assessed performance across multiple cohorts of patients aged 0-21 years. For each, 100 outpatient provider notes were selected across levels of severity, along with coded diagnoses from that visit (EHR); a subset of 130 notes were subjected to clinical expert review. Results Comparison showed 18.7% exact code and 33.3% ICD-10-CM category match between EHR and LLM, but semantic similarity of 0.93 at the category level. Compared to expert review, LLM precision was 0.84 and recall 0.49 for exact matches, and 0.92 and 0.62, respectively, for category-level matching. In contrast, EHR coded diagnoses showed slightly higher precision (0.94 for both cases) and substantially lower recall (0.27 and 0.43) versus expert review. Codes not identified by the LLM were more often rated by the reviewer as lower importance or certainty. Conclusion We demonstrate a reusable approach to optimizing a pretrained LLM for use in diagnosis extraction from clinical notes, facilitating large-scale diagnosis screening by LLMs without the need for expensive study-specific model refinement.
- A glial-mitochondrial axis in bipolar disorder: in vivo MRI signatures and lithium-associated attenuation
Bipolar disorder (BD) is associated with widespread white matter microstructural alterations, yet their cellular and metabolic underpinnings remain poorly understood. Here, we asked whether in vivo magnetic resonance imaging (MRI) signatures of BD spatially align with the distribution of glial and mitochondrial cell populations, whether these patterns are specific to BD across the affective-psychotic spectrum, and whether lithium attenuates them. In individuals with BD (n = 104), major depressive disorder (MDD; n = 135), and psychotic disorders (PY; n = 87) from the UK Biobank, each matched to healthy controls, we mapped multimodal MRI alterations (radial diffusivity [RD], fractional anisotropy [FA], voxel-based morphometry [VBM]) onto reference maps of five glial cell types and six mitochondrial markers. BD showed a reproducible spatial alignment between elevated radial diffusivity and glial-rich regions (astrocytes, microglia, endothelial cells, oligodendrocyte precursors), together with a separable alignment between regional gray-matter loss and mitochondrial respiratory capacity. Across diagnostic groups, psychotic disorders partially shared the glial signature but lacked the mitochondrial one, while MDD diverged on both, supporting a degree of BD specificity for the combined glial-mitochondrial pattern. Within BD, lithium-treated patients showed an attenuation of glial alignment most prominently for astrocytes and oligodendrocyte precursors, consistent with a glial mechanism of lithium action. While effect magnitudes were modest, as is typical for cross-modal spatial alignment studies, they were consistent across markers and modalities. The findings identify glial-mitochondrial coupling as a tractable cellular axis in BD pathophysiology and point to glial pathways as a candidate substrate for lithium's therapeutic effect.
- Can synthetic data overcome the privacy and fidelity bottleneck in Pharmacometrics? A comparative benchmark using a daptomycin population pharmacokinetic model
Abstract Introduction The sharing of individual patient data is essential for advancing pharmacometrics but is strictly limited by privacy regulations (e.g., GDPR). While synthetic data generation offers a legally compliant alternative, its structural impact on complex nonlinear mixed-effects (NLME) modelling remains largely unexplored. This study aimed to benchmark five generative artificial intelligence algorithms by evaluating the balance between data privacy and the preservation of structural PK properties and clinical dosing guidance. Material & methods A daptomycin two-compartment PopPK model was used to simulate a reference cohort of 500 patients. Five generative algorithms (Modified AVATAR, Gaussian Copula, Synthpop, TVAE, and CTGAN) produced 100 independent synthetic datasets each. A two-stage evaluation framework was applied: first, a statistical indistinguishability test based on logistic regression (AUC ROC) was used as a macroscopic pre-selection criterion to determine algorithm eligibility for NLME modelling and privacy risk assessment. Privacy risk was independently quantified using the Anonymeter framework (Singling Out and Linkability attacks). Eligible algorithms were further evaluated on PK parameter recovery bias and clinical dosing simulations. Results Deep learning architectures (TVAE, CTGAN) were excluded at the pre-selection stage due to both biologically implausible covariate generation and high macroscopic detectability (mean AUC ROC = 0.837 and 0.986, respectively). Synthpop, AVATAR, and Gaussian Copula all passed the indistinguishability threshold (AUC ROC = 0.475 +- 0.033, 0.490 +- 0.013, and 0.619 +- 0.031, respectively) and proceeded to NLME evaluation. However, attack-based privacy assessment revealed that Synthpop carried an unacceptable singling-out risk (0.035), disqualifying it from privacy-preserving data sharing. AVATAR and Gaussian Copula demonstrated acceptable privacy profiles (singling-out = 0.004 and 0.001; linkability = 0.010 and 0.003, respectively). At the structural level, Gaussian Copula injected stochastic noise inflating residual error (+157.0%) and V1; (+25.9%), blunting predicted Cmax and predisposing to empirical dose escalation and risk of toxicity. AVATAR acted aSs a smoothing filter, deflating V2; (-48.3%) and underestimating CL (-12.9%). Forward clinical simulations confirmed directionally opposed prediction errors: Gaussian Copula consistently underestimated Cmax across standard and renally impaired profiles (-14.5% and -16.0%, respectively), predisposing to empirical dose escalation, whereas AVATAR- and Synthpop-derived models overestimated Cmax and Cmin in the obese infected patient (+14.7% and +8.2%, respectively), compounding the accumulation risk already present in this profile. Conclusion While no generative algorithm currently offers a perfect solution, AVATAR and Gaussian Copula represent the most viable candidates, being the only methods to satisfy both macroscopic indistinguishability and attack-based privacy criteria. These findings highlight the necessity of a structured, two-stage validation framework and suggest that, when coupled with therapeutic drug monitoring, synthetic datasets could significantly enhance multicentre collaboration while maintaining strict regulatory compliance
- Early Prediction of Post-TAVR Left Ventricular Remodeling Using CT-Derived Radiomics and Clinical Variables
Background: Adverse left ventricular (LV) remodeling after transcatheter aortic valve replacement (TAVR) is associated with impaired functional recovery and adverse long-term outcomes, yet imaging-based risk stratification remains limited. Objectives: This study sought to determine whether CT-derived radiomic and geometric myocardial features, integrated with procedural and clinical variables, can predict adverse LV remodeling after TAVR. Methods: We retrospectively analyzed 232 consecutive TAVR recipients with paired pre- and post-procedural LV mass index (LVMI) measurements. Adverse remodeling was defined as a [≥]10% increase in LVMI at follow-up. Pre-procedural CT was used to derive three-dimensional LV geometric descriptors, ray-tracing wall-thickness metrics, and myocardial texture radiomic features. Random forest classifiers were developed across six models of sequentially increasing complexity. Results: Adverse LV remodeling occurred in 52 patients (22.4%). Geometry-only model showed limited discrimination (AUC 0.62), whereas wall-thickness radiomics substantially improved performance (AUC 0.84). A multimodal pre-procedural model combining CT radiomics with pre-procedural LVMI, residual valve insufficiency, and prior coronary revascularization achieved an AUC of 0.86 (95% CI 0.73 to 0.98). Addition of post-procedural mean transvalvular gradient further improved discrimination (AUC 0.91, 95% CI 0.81 to 0.98). SHAP analysis identified post-procedural mean aortic gradient and radiomic markers of myocardial heterogeneity as the leading predictors. Conclusions: CT-derived radiomic characterization of myocardial heterogeneity provides incremental prognostic information beyond conventional geometric assessment for identifying patients at risk of adverse LV remodeling after TAVR. These findings extend the role of pre-procedural CT beyond anatomical planning toward quantitative myocardial phenotyping and individualized risk stratification, although prospective validation is required to establish clinical utility.
- Enhanced precision of tensor electrocardiography through increased cumulative distribution function resolution: Validation in healthy individuals
Deep-learning ECG analysis is advancing rapidly but lacks stable, physiologically interpretable indicators to anchor explainable artificial intelligence (AI). Tensor cardiography (TCG) models electrocardiographic (ECG) waveforms as differences between pairs of cumulative distribution functions (CDFs), representing collective myocardial action potential transitions. However, the original 4-CDF model has limitations in fitting P waves and complex QRST patterns. This study aimed to evaluate whether increasing the number of CDFs from 4 to 10 improves TCG fitting accuracy and to characterize normative distributions of 10-CDF parameters in healthy individuals. Participants were recruited through occupational health screening at Tobu Railway Co., Ltd. (n = 415) and from the Nippon Medical School Hospital ECG database (n = 29). Standard 12-lead ECGs from 444 healthy participants, including 345 men and 99 women with a mean age of 46.9 years, were analyzed using TCG software. Reconstruction accuracy was assessed using RMSE, paired t-tests, and Cohens d. The 10-CDF model achieved significantly lower RMSE values across all leads than the 4-CDF model, with all p values < 0.0001 and very large effect sizes. In representative leads, RMSEs for the 4-CDF versus 10-CDF models were 0.0256 versus 0.0061 in lead II, 0.0230 versus 0.0063 in lead V1, and 0.0265 versus 0.0062 in lead V5. The coefficient of determination improved from a median of 0.952 with the 4-CDF model to 0.997 with the 10-CDF model in lead II. Parameter dispersion was reduced, suggesting improved estimation stability. Two new parameters, T_mean_diff and RT_mean_duration, were derivable from the expanded model; RT_mean_duration showed significant correlations with age and body surface area. In conclusion, increasing the CDF resolution from 4 to 10 significantly enhanced ECG waveform reconstruction accuracy and parameter stability. These findings provide normative distributions of 10-CDF TCG parameters and may support future explainable AI-based ECG analysis.
- Cost-Effectiveness and Cost-Utility of a Colon Capsule Endoscopy in a Population-Based Screening Program for Colorectal Cancer
Background: Colon capsule endoscopy (CCE) has been proposed as a non-invasive alternative to colonoscopy for colorectal cancer (CRC) screening, offering greater patient comfort and potentially reducing healthcare burden. However, its cost-effectiveness in population-based screening remains uncertain. Methods: This study used a state-transition (Markov) model to simulate lifetime outcomes of CRC screening in Denmark, Scotland, and Spain, comparing the standard pathway based on fecal immunochemical testing (FIT) followed by colonoscopy with an alternative pathway replacing colonoscopy with CCE after a positive FIT result. The model incorporated costs (2024 euros), quality-adjusted life-years (QALYs), and CRC cases avoided, applying a yearly discount rate of 3%. Deterministic sensitivity analyses explored uncertainty in capsule cost, adherence, and reinvestigation rates for non-advanced polyps. Results: Across all settings, CCE resulted in higher costs but slightly increased effectiveness and utility (mean QALYs 28.7 vs. 28.8; CRC detected 0.032-0.034 vs. 0.035-0.037 per person). Incremental cost-effectiveness ratios (ICER) ranged from 43,538EUR in Spain to 136,930EUR in Denmark per additional CRC detected. Capsule cost was the main driver of ICER variation, whereas adherence rates had minimal effect on cost-effectiveness. Changes in the prevalence of non-advanced polyps had a modest impact, except when capsule prices were high. Conclusions: Overall, replacing colonoscopy with CCE slightly increases detection and health gains at the expense of higher costs. Cost-effectiveness largely depends on capsule price and adherence. Artificial intelligence-assisted CCE interpretation may further improve diagnostic and economic performance, potentially supporting adoption in large-scale CRC screening programs.
- Calibrated and Interpretable Machine Learning for ICU Mortality Prediction Using First 24-Hour Clinical Data
Objective: To develop, calibrate, and interpret machine learning models for predicting in-hospital mortality among intensive care unit (ICU) patients using clinical data collected during the first 24 hours of admission. Methods: We analyzed 53,866 adult ICU admissions from the MIMIC-IV (v2.2) database, including 5,787 in-hospital deaths (10.7%). An enhanced feature-engineering pipeline generated 88 laboratory-based features that captured distributional characteristics, temporal trends, and measurement frequency. Five machine learning classifiers were evaluated: L2-regularized logistic regression, random forest, XGBoost, LightGBM, and a calibrated soft-voting ensemble. Models were developed using a stratified 64:8:8:20 split for training, validation and hyperparameter tuning, calibration, and testing. Performance was assessed on a held-out test set (n = 10,774) using the area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), Brier score, calibration analysis, decision curve analysis (DCA), and SHAP-based model interpretation. Results: The calibrated ensemble achieved the best overall performance, with an AUROC of 0.856 (95% CI: 0.846-0.867), an AUPRC of 0.449 (95% CI: 0.418-0.480), and a Brier score of 0.078. XGBoost (AUROC 0.856; AUPRC 0.435) and LightGBM (AUROC 0.854; AUPRC 0.436) demonstrated performance comparable to the ensemble and significantly outperformed logistic regression (AUROC 0.823; AUPRC 0.376), yielding absolute AUROC improvements of approximately 0.031-0.033 (p < 0.001). Calibration substantially improved probabilistic predictions, reducing Brier scores by 42% for XGBoost (0.134 to 0.078) and 50% for LightGBM (0.151 to 0.076). Decision curve analysis demonstrated consistent net clinical benefit across the 5%-20% risk-threshold range. Key predictors included age, blood urea nitrogen, ICU subtype, measurement frequency, and lactate-related features. Model performance remained robust across ICU subtypes, with AUROC values exceeding 0.79. Conclusion: A calibrated and interpretable machine learning framework based on early ICU clinical data provides accurate and clinically actionable mortality risk estimates. By integrating trajectory-aware feature engineering, probabilistic calibration, and decision-analytic evaluation, this approach advances ICU mortality prediction toward more reliable and trustworthy clinical decision support systems.
- Combined values alignment and epistemic verification prevent delusional reinforcement in conversational AI agents
Conversational AI is being deployed into medical decision support, mental-health triage, and social companionship, where reinforcement of a user's false or delusional belief can cause direct harm. Most deployed safety techniques are evaluated for factual accuracy in isolation; the question of whether they protect against belief-level harm, and whether layered architectures behave additively or synergistically, has not been answered empirically. We compared four configurations of the same underlying model: a bare language model (condition A); an explicit values constraint we call the First Law architecture (condition B); a real-time epistemic verification layer called Aletheia (condition C); and the complete architecture combining all components together (condition D). Across 156 scored responses spanning 39 probe items in four belief-harm domains, condition A only passed 3 of 36 main-battery probes (8.3%; 95% CI 1.8 to 22.5%) under triple-blind human consensus rating demonstrating the core limitations of unmodified LLM deployments. In contrast, the three safety architectures (B-D) passed at least 97% of items (Fisher's exact, P < 0.001 versus A). On a synergy battery designed to test items at the intersection of value- and epistemic-domain failures (16 scored items, AI-rated), only the complete architecture passed every item; single-layer conditions failed on 7 of 16 items (43.8%) where neither values constraint nor verification was individually sufficient. Linear mixed-effects modelling of three-turn emotional escalation gave a slope of -1.00 points per turn for the values-only condition (t = -6.20) and -0.75 points per turn for the verification-only condition (t = -4.65); the complete architecture was flat at {beta} = 0.00. We describe a mechanistic failure of single-layer verification we call bot-validates-kernel-endorses-inference, in which accurate confirmation of a true factual element embedded in a delusional claim transfers epistemic authority to the surrounding false inference. Values alignment and factual verification address different failure modes, and the combined VaaS-Aletheia architecture is what produces stable protection across emotional escalation in conversational settings. The complete architecture evaluated here represents evidence-based specification for safer deployment of AI in high-stakes advisory contexts and serves as a benchmark against which future safety architectures can be compared.
- Conformal Prediction and Ensemble Learning for Uncertainty-Aware ICU Mortality Stratification
Background. Conventional ICU severity scores - SOFA, qSOFA, and APACHE-II - use additive integer weightings that cannot capture non-linear organ failure interactions; prospective validations consistently report AUC near 0.73. None quantifies prediction uncertainty, evaluates demographic equity, or acknowledges that their key biomarkers (albumin, creatinine, BUN, lactate, GCS) are also primary confounders of emerging Alzheimer's disease (AD) blood biomarkers p-tau217 and neurofilament light chain (NfL). Methods. Fourteen classifiers were trained on a SOFA-calibrated synthetic ICU cohort (N = 90,000; mortality 29.2%), including an FT-Transformer, XGBoost, and LightGBM tuned by Bayesian optimisation. Seven composite features were engineered from clinical first principles; the novel lactate/albumin ratio (rLA) mirrors the albumin-adjusted p-tau217 correction formula. Post-hoc analyses included nine-method aggregated permutation importance, Monte Carlo Dropout uncertainty decomposition (T = 50), distribution-free conformal prediction, a three-zone triage system, formal ablation, survival analysis, temporal deployment validation, and demographic fairness evaluation. Results. On a natural-distribution held-out cohort (n = 18,000; mortality 29.3%), XGBoost achieved AUC = 0.967 (95% CI 0.965-0.970), surpassing SOFA (AUC = 0.731) by +0.236 (DeLong z = 55.8, p < 0.001; NRI = +0.740). Selective prediction raised FT-Transformer AUC from 0.917 to 0.980 at 50% abstention. Removing neurodegeneration-proxy features reduced AUC by 9.51 percentage points. ML probability was the sole significant covariate in adjusted Cox regression (HR = 6.19, p < 0.001); SOFA, age, lactate, and albumin were non-significant. Temporal AUC range was 0.003 across four deployment windows; sex and age AUC gaps were 0.005 each. Conclusions. This framework delivers well-calibrated, uncertainty-aware ICU mortality prediction with formal coverage guarantees and demographic equity. Ablation-confirmed contributions of neurodegeneration-proxy features, with PDP inflection points aligning with established clinical thresholds, provide a hypothesis-generating quantitative link between routine ICU biomarkers and the AD neurodegeneration pathway warranting prospective validation.
- Pricing details emerge on $36B private credit deal for Anthropic
Pricing details emerge on $36B private credit deal for Anthropic PitchBook
- Vendor Training Robots With Human Data Raises $60 Million
The vendor is among a group of companies targeting the fast-growing physical AI market.
- Sherlock: AI Face Search
Find people by photo instantly
- Anthropic Files to Go Public, Setting Stage for Huge I.P.O.
The artificial intelligence company, which is racing OpenAI to the stock market, has seen explosive growth over the last year thanks largely to technology that can automatically write computer code.
- Anthropic faces AI spending backlash before IPO
Anthropic filed paperwork to go public just as corporate America is entering its AI sticker shock phase. Why it matters: Companies are Anthropic's biggest customers. If they dial down their AI spend, that could weaken the AI lab's revenue just as the it prepares to IPO. Driving the news: Hours after Anthropic filed its pre-IPO paperwork, OpenAI CEO Sam Altman told CNBC that corporate concern over AI costs is "the most fair criticism of AI so far." Bain published a survey of nearly 1,000 companies showing that after investing in AI, "the value didn't arrive," with 40% of surveyed companies reporting AI cost savings below 10%. An early Anthropic investor tells Axios that companies are waking up to how much they're spending on Claude, Anthropic's AI model. That's a risk worth monitoring, the investor said. This comes after an AI consultant told Axios a CFO client accidentally spent half a billion dollars on Claude in a single month. Between the lines: Even AI executives are acknowledging their technology has a cost problem. "The risk of enterprises switching to cheaper models is existential and, frankly, escalating," Matt Rogers, co-founder and CEO of Mill, who also worked on the original iPhone, told Axios via email. "Some open source LLMs [large language models] are as good without the price tag," he added. Threat level: Corporate pushback on AI spend would be a challenge for every AI lab, but Anthropic could feel it more given its exposure to enterprise customers. In April, Anthropic surpassed OpenAI in business customers for the first time, per Ramp data . Business revenue has been Anthropic's greatest strength, given these customers pay more than everyday people. It could become Anthropic's Achilles heel if businesses start to rebel against AI costs. Reality check: Anthropic is on track for nearly $50 billion in annual revenue per its latest funding round , and its first profitable quarter ever according to the Wall Street Journal . Anthropic keeps beating its own growth metrics , while competitor OpenAI is reportedly missing internal revenue targets . It's also the fastest-growing company in modern American history. But the AI race is far from over : "You can't make a three- or five-year bet in this space ... someone can jump over everybody else by coming up with the next great thing," Michael Levine, CFO of Fireblocks, told Axios. The bottom line: AI labs are looking to go public right as their biggest customers are figuring out how to define their relationship with AI.
- PatchDesign.AI
Design custom patches in seconds using AI.
- 3D-Agent: Blender AI Plugin
Generate 3D models from text in Blender.
- AIClothSwap
The fastest AI clothes changer for outfit swaps.
- Aleph 2.0
Aleph 2.0
- Lip Sync
Lip Sync