Artificial Intelligence
Models, agents, and their real-world impact.
Reason No topic weight computed yet.
Articles
20-
Deep Learning Reconstructs 35 Years of Global Human Migration Flows
For the first time, researchers have a consistent, annual picture of who moved where across all 230 countries from 1990 to 2024 — and deep learning built it from sources that previously couldn't talk to each other.
Reality 72Hype 35 -
Trapped-Ion Processor Cuts Logical Error Rates Up to 800× via Combined Error Correction
An 800-fold reduction in logical error rates on a real quantum processor isn't a simulation or a theoretical bound — it's a measured result published in Nature. That number moves the goalposts on what fault-tolerant quantum computing requires in practice.
Reality 78Hype 35 -
Brick-Composer Trains MLLMs to Assemble Physical Objects Step by Step
AI that can read a design and build it from physical parts has been a fantasy — Brick-Composer makes it measurably less so, lifting assembly success from under 1% to ~15% per step, with a single 8B model correctly handling 42% of steps end-to-end.
Reality 72Hype 45 -
Nature's June Images: Milky Way, Selfie Cameras, and AI Astrophysics
A smartphone's front camera can now track your heart rate — no wearable required. That's buried in Nature's June briefing alongside a galaxy portrait and a pointed question about AI's effect on astrophysics.
Reality 62Hype 35 -
China's EV Fleet Linked to 260,000 Fewer Premature Deaths
China's electric vehicle rollout has prevented an estimated 260,000 premature deaths — but the air-quality story is messier than the headline suggests, with some pollutants actually rising.
Reality 72Hype 45 -
AI Matches But Doesn't Beat Headache Specialists in Literature Summarization
Ten headache specialists preferred their own writing over AI — but couldn't reliably tell the two apart. That gap between preference and detection is where this study gets interesting.
Reality 72Hype 25 -
China's Multi-Corresponding-Author Inflation Exposed — and Partly Fixed
Nearly one in three papers from China listed multiple corresponding authors between 2016 and 2020 — a rate anomalous enough to distort global research metrics and prompt a policy crackdown.
Reality 72Hype 25 -
Dual-Use Military-Civilian Research Consistently Outperforms Citation Benchmarks
Science that feeds both weapons labs and civilian markets doesn't just survive the dual-use stigma — it dominates citation rankings. A new Nature analysis finds military-applicable research pulls measurably more scientific impact than civilian-only work.
Reality 72Hype 45 -
Spiralling Ebola Outbreak Tests Lessons From Past Epidemics
The current Ebola outbreak is escalating fast enough that Nature's editors are pulling the emergency brake — invoking hard-won lessons from prior epidemics as the primary containment playbook.
Reality 72Hype 28 -
Trial Cracks "Undruggable" Cancer, Signaling Broader Therapeutic Shift
A landmark clinical trial has produced unprecedented results against a cancer type long considered beyond the reach of targeted therapy — the kind of outcome that rewrites treatment roadmaps, not just protocols.
Reality 55Hype 75 -
Blue Origin Rocket Explosion Threatens NASA's Lunar Timeline Against China
A Blue Origin rocket has exploded, and the blast radius extends well beyond the launch pad — it now threatens NASA's already razor-thin margin to land humans on the Moon before China does.
Reality 65Hype 72 -
South African Labs Run Science Through Rolling Blackouts Daily
Most research infrastructure assumes uninterrupted power. South African scientists have quietly built a parallel playbook for doing rigorous work without it — and the rest of the world hasn't been paying attention.
Reality 72Hype 35 -
Cardiac Gene Therapy Rebounds After Years of Failed Trials
Heart failure gene therapy is back — and this time the field has more than hope to sell. After a decade of high-profile stumbles, new treatments targeting the heart's contractile machinery are entering serious clinical contention.
Reality 55Hype 65 -
Science Sleuths Flag 100+ Suspect Images in Thermo Fisher Antibody Catalogue
More than 100 product images in Thermo Fisher's antibody catalogue have been flagged as suspicious — and if the findings hold, a significant slice of the world's most-used research reagents may be misrepresented at the point of sale.
Reality 45Hype 65 -
Nature Argues Human Judgment Remains Essential for Scientific Literature Reviews
Nature isn't hedging: AI-generated scientific reviews aren't just imperfect — they're structurally unfit for the job. The argument isn't about hallucinations. It's about judgment.
Reality 65Hype 35 -
Superconducting Qubits Deliver Certified Perfect Randomness From Weak Sources
True randomness — not just hard-to-predict, but mathematically certified — has been demonstrated on a real quantum device for the first time. The trick: take slightly random input, run it through a Bell test, and get output that is provably free of any classical bias or adversarial tampering.
Reality 72Hype 45 -
Nature Calls Out Neuroscience's Broken Computer-Brain Metaphor
Neuroscience's dominant framework — the brain as a biological computer — is being called out in Nature as a theoretical dead end. The field isn't just stuck; it may be stuck in the wrong direction.
Reality 72Hype 25 -
Acute Stress Disrupts Brain's Memory-Linking Circuitry, Blocking Insight
Stress doesn't just make you anxious — it actively breaks the neural wiring that lets you connect dots. New imaging data shows why a single stressful episode is enough to kill inferential thinking.
Reality 72Hype 55 -
Wearable Robot Helps Children With Spinal Muscular Atrophy Regain Knee Strength
A robotic exoskeleton is doing what physical therapy alone rarely can: rebuilding functional knee strength in children with spinal muscular atrophy (SMA), a disease that progressively strips away motor neurons and has no reliable mechanical fix.
Reality 55Hype 65 -
Causal Head Imbalance Found to Drive Multimodal Hallucination, Targeted Fix Proposed
When a vision-language model ignores what it sees and trusts a wrong text prompt instead, the culprit isn't the whole network — it's a structural imbalance between a few dozen attention heads. Researchers have now mapped that imbalance causally and built a surgical fix that outperforms every inference-time baseline tested.
Reality 72Hype 45