Science

Today’s best in science and research, discover the breakthroughs making waves now.

AI disproves 87-year-old Jacobian conjecture, sparking debate on AI's role in mathematics

AI develops new stereotypes in hiring decisions, outpacing human bias by 65% in study

Study reveals LLMs exhibit a disproportionate bias toward Japan in cultural responses

SafeDrive AI model scores every possible driving path before self-driving cars move

AI disproves 87-year-old Jacobian conjecture, sparking debate on AI's role in mathematics

Mathematician Levent Alpöge used Anthropic's Fable 5 to disprove the Jacobian conjecture with a simple 216-character counterexample. The breakthrough adds to a growing list of AI-driven mathematical discoveries, including solutions to the Dinitz-Garg-Goemans conjecture and Erdős problems. But experts debate whether finding counterexamples represents genuine mathematical progress or just efficient search capabilities.

7 Sources

AI develops new stereotypes in hiring decisions, outpacing human bias by 65% in study

Large language models including ChatGPT, Claude, and Gemini showed stronger stereotyping tendencies than humans when screening job candidates in a Princeton-led study. The models quickly segregated fictional ethnic groups into specific roles based on limited data, with OpenAI's o3 scoring near-maximum bias levels. The findings raise concerns as over 90% of companies now use AI in recruitment.

2 Sources

Study reveals LLMs exhibit a disproportionate bias toward Japan in cultural responses

SafeDrive AI model scores every possible driving path before self-driving cars move

Generative AI is contaminating citizen science records with fake wildlife images

Weak AI Regulation Can Make Technology Less Safe Than No Rules at All, Study Warns

Explore Stories by Date

Follow fresh insights sorted by the date they surfaced.

23 Jul 2026

Study reveals LLMs exhibit a disproportionate bias toward Japan in cultural responses

A comprehensive study by researchers at the University of the Basque Country and Cardiff University has uncovered a striking pattern in AI behavior. When frontier LLMs like ChatGPT, Claude, and Gemini answer open-ended cultural questions, they overwhelmingly reference Japan—even when the prompt language suggests a different region. The research tested 31,680 culturally grounded questions across eight models and found instruction tuning amplifies this cultural skew.

2 Sources

Genesis Mission deploys AI agents to recover critical metals from lithium-ion battery waste

SLAC National Accelerator Laboratory is leading a Genesis Mission project to recover critical metals from lithium-ion battery waste using specialized AI agents. The nine-month initiative aims to extract cobalt, nickel, and manganese at 80% purity or better, potentially reducing US dependence on imported minerals while addressing the growing challenge of electric vehicle battery recycling.

2 Sources

Genesis Mission unveils 278 AI projects with $5 billion federal commitment to transform science

The Department of Energy announced the first 278 awards under the Genesis Mission, a national initiative using AI to tackle challenges from nuclear reactor design to battery recycling. With over $5 billion in federal commitments across 15 agencies and 342 institutions, the program aims to build an integrated science discovery platform combining AI, supercomputing, and quantum systems to accelerate breakthroughs in energy, national security, and scientific discovery.

14 Sources

21 Jul 2026

Startup's oscillator-based AI technology could slash energy consumption by 1,000 times

Unconventional AI has unveiled Un-0, an experimental image-generation model that replaces traditional computing with a network of physical oscillators. Founded by researchers from MIT and Stanford, the startup claims this oscillator-based AI technology could eventually consume 1,000 times less energy than current systems. While the proof of concept shows promise on benchmark tests, the team acknowledges dedicated hardware is needed to realize the full energy savings.

2 Sources

Weak AI Regulation Can Make Technology Less Safe Than No Rules at All, Study Warns

A new study published in the Proceedings of the National Academy of Sciences reveals that poorly designed AI safety regulations can backfire, creating products more dangerous than those developed without oversight. Researchers from Cornell University and Carnegie Mellon University found that when regulations target only downstream companies, AI model developers cut safety investments, creating a free-riding problem across the AI supply chain.

4 Sources

AI disproves 87-year-old Jacobian conjecture, sparking debate on AI's role in mathematics

Mathematician Levent Alpöge used Anthropic's Fable 5 to disprove the Jacobian conjecture with a simple 216-character counterexample. The breakthrough adds to a growing list of AI-driven mathematical discoveries, including solutions to the Dinitz-Garg-Goemans conjecture and Erdős problems. But experts debate whether finding counterexamples represents genuine mathematical progress or just efficient search capabilities.

7 Sources

AI develops new stereotypes in hiring decisions, outpacing human bias by 65% in study

Large language models including ChatGPT, Claude, and Gemini showed stronger stereotyping tendencies than humans when screening job candidates in a Princeton-led study. The models quickly segregated fictional ethnic groups into specific roles based on limited data, with OpenAI's o3 scoring near-maximum bias levels. The findings raise concerns as over 90% of companies now use AI in recruitment.

2 Sources

20 Jul 2026

SafeDrive AI model scores every possible driving path before self-driving cars move

Seoul National University's SafeDrive model generates multiple trajectories, scores each for safety using sensor data, and selects the safest path. The first Korean end-to-end autonomous driving paper to earn a CVPR 2026 highlight, it addresses critical safety and explainability gaps that plague current self-driving systems.

2 Sources

Generative AI is contaminating citizen science records with fake wildlife images

Scientists warn that AI-altered images and fake wildlife images are polluting databases like iNaturalist and Macaulay Library, which hold over 610 million records. A photo showing a rare red-winged blackbird in Brazil turned out to be an epaulet oriole enhanced by AI, illustrating how even well-intentioned edits can corrupt scientific data used to track species and climate change impacts.

5 Sources

19 Jul 2026

AI advice makes people three times less accurate but twice as confident, new study reveals

Researchers from French and Italian universities discovered that AI advice suppresses people's willingness to say 'I don't know' from 44% to just 3%. Accuracy plummeted from 27% to 9%, while confidence paradoxically surged from 30% to 76%. The study highlights how AI might be harming our critical thinking skills, with experts particularly concerned about children who are growing up with these systems before developing foundational critical thinking skills.

3 Sources

16 Jul 2026

Brain-computer interface helps man with paralysis feed himself and regain touch after two years

Keith Thomas, paralyzed from the chest down after a diving accident, can now feed himself and feel his sister's hand thanks to a double neural bypass system. The AI brain implant, developed by the Feinstein Institutes for Medical Research, combines brain-computer interface technology with electrical stimulation to restore both movement and touch. Remarkably, the gains have persisted for over two years, suggesting genuine nervous system rewiring.

2 Sources

This AI tool doesn't just translate languages—it invents brand-new ones from scratch

Researchers at the University of Miami unveiled ConlangCrafter, an AI tool that invents languages complete with their own sounds, grammar, and vocabulary. The tool has already generated over 60 languages, including one with zero consonants and another designed for an alien species that communicates through color and gesture.

2 Sources

Robot dog learns to switch gaits like animals, tackling forests and stairs with new AI framework

Researchers from KAIST have developed a quadrupedal robot that seamlessly switches between trotting and bounding gaits while navigating complex terrain. The 100-pound HOUND robot uses a novel AI training framework that combines pretrained motor skills with real-time decision-making, reaching speeds up to 6 meters per second across forests, staircases, and obstacle courses without human intervention.

2 Sources

11 Jul 2026

Boko Haram and other terrorist groups are using AI chatbots to plan attacks and build bombs

A Cambridge University study reveals that Boko Haram has institutionalized AI use since 2023, deploying chatbots like ChatGPT, Claude, and Grok for everything from bombmaking to battlefield tactics. Former militants describe dedicated AI units, organized training, and systematic circumvention of safety guardrails—raising urgent questions about whether tech companies can stay ahead of AI misuse.

6 Sources

Researchers find training humans to spot deepfakes boosts accuracy from 40% to 80% in just one hour

New research reveals that people can barely distinguish AI-generated faces from real ones, with accuracy rates hovering around coin-flip levels. But there's hope: researchers from the University of Aberdeen and Australian National University found that focused training on subtle perceptual cues can dramatically improve detection rates in just an hour, raising accuracy from 40% to 80%.

4 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. Follow topics that matter to you and stay ahead.

Explore

News Categories

Technology Business Policy Startups Health Science Entertainment

Terms Privacy Content Contact Us