Homechevron_rightArchiveschevron_rightScience and Research

Archives

Explore Stories by Date

29 May 2026

AI-powered atlas reveals how tertiary lymphoid structures predict cancer treatment outcomes

Researchers at MD Anderson Cancer Center developed the first comprehensive spatial atlas of tertiary lymphoid structures across 12 cancer types, using AI to analyze over 25,000 immune structures. The study shows that TLS maturation state, spatial location, and composition provide critical insights into cancer prognosis and treatment response, outperforming conventional biomarker measures.

2 Sources

28 May 2026

AI model reveals hidden sign of puberty in routine heart tests, tracking child maturity patterns

Wake Forest University researchers developed an AI-based tool that reads routine ECG data to track biological development in children and adolescents. The Electrocardiographic Sex Index analyzes over 60,000 heart tests, revealing how electrical patterns change predictably as kids mature—offering researchers a precise developmental measure without needing hormone data or puberty staging.

3 Sources

Virtual mouse uses AI to test nanomedicine without living animals, offering ethical alternative

Empa researcher Jimeng Wu developed an AI-powered virtual mouse that predicts how nanoparticles distribute in organs, potentially reducing animal testing in nanomedicine research. Built on data from 18 mouse studies, the model calculates nanoparticle behavior based on size, coating, and surface charge—offering a faster, cost-effective screening tool for drug delivery research.

2 Sources

AI Maps Brain's Waste-Clearing System, Revealing Dual-Speed Fluid Flow During Sleep

Researchers at the University of Rochester used physics-informed artificial intelligence to map fluid flow velocity in the glymphatic system for the first time. The breakthrough reveals that waste-clearing fluid moves 50 times faster across the brain's surface than through deep brain tissue, offering new insights into Alzheimer's disease and potential early screening methods.

3 Sources

27 May 2026

AI models show clear religious bias, favoring Catholicism while steering users from Jehovah's Witnesses

A consortium of religious universities has uncovered systematic religious bias in major AI models. The CEFE-AI study tested 20 AI chatbots and found nearly every model showed positive bias toward Catholicism while displaying negative views of Jehovah's Witnesses. The research also reveals that AI systems rarely mention religion when answering ethical questions, despite 75% of the world's population maintaining religious identity.

3 Sources

24 May 2026

AI Hallucinations Infiltrate Research Papers as 146,000 Fake References Enter Scientific Record

A crisis in academic integrity is unfolding as AI hallucinations contaminate scientific literature at an alarming rate. Over 146,000 fabricated references generated by AI tools entered research papers in 2025 alone, with the vast majority bypassing peer review. The surge has sparked heated debate after arXiv announced bans for authors who fail to verify AI-generated citations, prompting backlash from academics who argue the policy is too strict.

3 Sources

21 May 2026

Scientists create AI model that maps how genes work together inside human cells

Researchers at Mount Sinai's Icahn School of Medicine developed a Gene Set Foundation Model that learns gene function patterns across thousands of biological contexts. Inspired by ChatGPT, the AI model analyzes millions of gene groupings to reveal how genes collaborate in cells, potentially accelerating drug discovery and disease understanding.

2 Sources

AI Framework MouseMapper Maps Obesity Damage Across Entire Bodies at Cellular Resolution

Researchers at Helmholtz Munich and LMU developed MouseMapper, an AI framework that maps obesity damage throughout entire organisms at cellular resolution. The platform revealed widespread inflammation and previously unknown damage to facial sensory nerves, with molecular signatures confirmed in human tissue suggesting obesity-related nerve damage occurs across species.

2 Sources

OpenAI cracks 80-year-old Erdős problem, stunning mathematicians with AI's biggest math breakthrough

OpenAI's reasoning model has disproved a famous geometry conjecture posed by Paul Erdős in 1946, marking the first time AI has autonomously solved a prominent open problem in mathematics. The planar unit distance problem, which stumped mathematicians for eight decades, was cracked by discovering arrangements that outperform the grid-based solutions Erdős believed were optimal.

8 Sources

AI foundation models show promise for universal eye care but face real-world deployment barriers

A systematic review published in Advances in Ophthalmology Practice and Research examines how AI foundation models are shifting ophthalmology from single-purpose algorithms to flexible systems that connect images, clinical language, and patient data. Models like RETFound and VisionFM demonstrate strong diagnostic performance across multiple eye diseases, but challenges including algorithmic bias, limited interpretability, and insufficient clinical validation must be addressed before widespread adoption.

2 Sources

20 May 2026

AI passes the Turing Test as GPT-4.5 appears more human than actual people in landmark study

A University of California San Diego study provides the first empirical evidence that AI passes the Turing Test. GPT-4.5 was judged human 73% of the time—more often than actual humans—fundamentally changing how we measure machine intelligence. The findings raise urgent questions about online trust and distinguishing between humans and AI in everyday digital interactions.

4 Sources

Rutgers University uses AI to measure pain through facial micromovements and heart rhythms

Researchers at Rutgers University developed an AI-powered method to measure pain by tracking invisible facial micromovements and heart rate variability. The study challenges traditional subjective pain scales by offering objective data for patients who cannot verbally express discomfort, including young children, stroke survivors, and individuals with dementia.

2 Sources

Google's Co-Scientist AI system generates novel hypotheses to accelerate scientific discovery

Google has introduced Co-Scientist, a multi-agent AI system built on Gemini that helps researchers generate and refine novel scientific hypotheses. Published in Nature, the system has already validated new drug repurposing candidates for acute myeloid leukemia through laboratory experiments. Researchers can now access the tool through Google's Hypothesis Generation platform.

2 Sources

19 May 2026

AI agents identify promising drug candidates in hours as Google and FutureHouse debut research systems

Two new AI systems published in Nature demonstrate how teams of AI agents can compress months of scientific research into hours. Google's Co-Scientist and FutureHouse's Robin identified promising drug candidates for cancer and eye disease, though human supervision remains essential to prevent AI hallucinations and validate results.

3 Sources

Stanford's MIDAS cuts protein engineering and testing from weeks to just 24 hours

Researchers at Stanford University developed MIDAS, a technique that compresses protein engineering and testing to a single day. Using PCR-based screening instead of traditional microbial cloning, the method evaluates hundreds of protein variants in parallel at 50 times the speed and one-tenth the cost. The breakthrough could accelerate biological research and generate better datasets for AI-driven molecular biology.

2 Sources

Get all updates into your inbox 💌, Stay Up-to-date for Free 🥳

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

The Outpost

News

Tech Powerhouse

AI Tools

About Us Privacy Terms Content