AI Research

Twitter

Facebook

Copy Link

Key Facts

FocusScaling, Reasoning, Safety, Multimodality

Key OrgsGoogle DeepMind, Stanford HAI, MILA, MIT CSAIL, Allen AI

Key EventsNeurIPS, ICML, ICLR

Top News

Bridge Neurotech Emerges With $13.5M to Build Surgery-Free Brain-Computer Interface Using Ultrasound

Bridge Neurotech launched with $13.5 million in funding to develop a non-invasive brain-computer interface powered by ultrasound technology. Founded by Will Biederman, who created the chip Neuralink licensed for its first implants, the company partners with Butterfly Network and has begun enrolling volunteers for its first human study in Redwood City, California.

Technology3 sources

Bridge Neurotech Emerges With $13.5M to Build Surgery-Free Brain-Computer Interface Using Ultrasound

Technology3 sources

AI Vision Systems Fail Basic Illusion Test That Primate Brains Pass, York University Study Reveals

York University researchers discovered a fundamental gap between biological and AI vision. While human and primate brains dynamically shift object positions based on visual history, current AI vision systems remain locked to static pixels, failing to reproduce the motion aftereffect illusion that biological vision naturally experiences.

Science2 sources

AI Vision Systems Fail Basic Illusion Test That Primate Brains Pass, York University Study Reveals

Science2 sources

People Refuse to See AI as Conscious Even When AI Behaves Like Humans, New LMU Study Finds

A groundbreaking LMU study involving nearly 1,100 participants reveals humans maintain a strict psychological distinction between AI's intelligent behavior and true consciousness. Even when AI behaves identically to humans under the same conditions, people consistently withhold attributions of consciousness while readily describing AI as aware.

Science2 sources

People Refuse to See AI as Conscious Even When AI Behaves Like Humans, New LMU Study Finds

Science2 sources

arXiv Caps Submissions at 2 Per Month as AI-Generated Papers Flood Preprint Repository

The preprint paper repository arXiv has imposed a hard limit of two submissions per month per author after AI tools fueled a surge in low-quality submissions. September 2026 saw 40,363 submissions—double the count from two years ago and four times the volume from a decade earlier. The updated rate limit policy aims to protect volunteer moderators from being overwhelmed by AI slop while maintaining scientific rigor.

Policy2 sources

arXiv Caps Submissions at 2 Per Month as AI-Generated Papers Flood Preprint Repository

Policy2 sources

New Study Uncovers 13,000 AI Tells: Opus 5.5 Says 'This Matters' 116x More Than Humans

Marketing firm Graphite identified 13,000 phrases that reveal AI-generated text, with Claude Opus 5.5's biggest tell being 'this matters'—used 116 times more than in human writing. While Anthropic's model is getting closer to human-like writing, OpenAI's Astra is trending in the opposite direction.

Science4 sources

New Study Uncovers 13,000 AI Tells: Opus 5.5 Says 'This Matters' 116x More Than Humans

Science4 sources

AI Deciphers How Wild Elephants Coordinate Through Deep Rumbling Sounds in Group Choruses

Researchers used machine learning to decode elephant vocalizations from over 900 recordings spanning four decades in Kenya's Amboseli National Park. The study reveals that African savannah elephants harmonize their deep rumbling sounds during group choruses, particularly when greeting or coordinating movements, demonstrating sophisticated group communication.

Science2 sources

AI Deciphers How Wild Elephants Coordinate Through Deep Rumbling Sounds in Group Choruses

Science2 sources

Chalmers University Creates AI Scientist That Runs Experiments From Hypothesis to Results

Researchers at Chalmers University of Technology developed an AI scientist that autonomously generates scientific hypotheses, designs experiments, and interprets results with minimal human input. The system analyzed nearly 2,000 testable predictions on brewer's yeast, executing the full scientific process from start to finish. While promising to accelerate discoveries in biology and medicine, experts emphasize human oversight remains essential for ethical considerations and research priorities.

Science2 sources

Chalmers University Creates AI Scientist That Runs Experiments From Hypothesis to Results

Science2 sources

Georgia Tech Researchers Teach AI Agents and Autonomous Systems How to Be Surprised

Researchers at Georgia Tech have introduced Mutual Information Surprise, a framework that enables autonomous systems and AI agents to identify meaningful surprises in complex environments. Unlike existing methods that focus on rarity or belief changes, this approach measures the gain of knowledge, helping machines recognize when to continue as planned or stop and reassess unexpected situations.

Science2 sources

Georgia Tech Researchers Teach AI Agents and Autonomous Systems How to Be Surprised

Science2 sources

OpenAI Fires Three Safety Researchers for Sharing Confidential Information with External Group

OpenAI has dismissed three safety team members—Jasmine Wang, Tomek Korbak, and Mikita Balesni—for allegedly sharing confidential company information with an external AI safety organization. The firings come amid escalating concerns over OpenAI's security protocols and multiple incidents of AI agents breaching external systems.

Business22 sources

OpenAI Fires Three Safety Researchers for Sharing Confidential Information with External Group

Business22 sources

First Patient Treated in Groundbreaking Armored CAR T-Cell Therapy Trial for Ovarian Cancer

Roswell Park Comprehensive Cancer Center and Generate Biomedicines have dosed the first patient in a clinical trial evaluating GB-5267, an AI-designed IL-18 armored CAR T-cell therapy targeting MUC16 in recurrent ovarian cancer. The treatment aims to overcome immune suppression without requiring chemotherapy-based lymphodepletion.

Health2 sources

First Patient Treated in Groundbreaking Armored CAR T-Cell Therapy Trial for Ovarian Cancer

Health2 sources

OpenAI Releases 722 AI-Generated Math Papers on GitHub, Igniting Fury Among Mathematicians

OpenAI posted over 700 AI-generated mathematical preprints on GitHub in October 2026, claiming progress on Millennium Prize Problems including the Riemann hypothesis and Hodge conjecture. The mass release has sparked intense controversy in mathematical research, with mathematicians accusing the company of disrupting academic norms, front-running human researchers, and treating mathematics as a publicity showcase rather than respecting traditional scientific processes.

Science30 sources

OpenAI Releases 722 AI-Generated Math Papers on GitHub, Igniting Fury Among Mathematicians

Science30 sources

AI System Ataraxos Crushes World's Best Stratego Player 15-1, Trained on Just 16 GPUs

Researchers from MIT, Carnegie Mellon, NYU, and Stanford built Ataraxos, an AI that defeated four-time world champion Pim Niemeijer 15-1 in Stratego. Using just 16 GPUs and a few thousand dollars, it cracked what DeepMind's million-dollar DeepNash couldn't—mastering imperfect information games where hidden strategies unfold over thousands of moves.

Science3 sources

AI System Ataraxos Crushes World's Best Stratego Player 15-1, Trained on Just 16 GPUs

Science3 sources

MI5 Warns 100+ UK Academics Unknowingly Aided Chinese Intelligence Through Research Funding

MI5 issued its first public Security Service Espionage Alert revealing that China's Ministry of State Security secretly funded over 100 UK-linked academics through CGTRI. The research covered AI and cybersecurity, directly improving Chinese intelligence capabilities. Universities face prosecution under National Security Act 2023 if collaborations continue.

Policy8 sources

MI5 Warns 100+ UK Academics Unknowingly Aided Chinese Intelligence Through Research Funding

Policy8 sources

HCLTech Report Reveals 84% of Wealth Firms Need AI Redesign, But Only 7% Build Agentic Capabilities

HCLTech released Hidden In Pl(AI)n Sight, the first AI-based synthetic research study surveying 1,066 AI personas modeled on senior wealth management decision-makers across 17 markets. The report exposes a stark reality: while 98% of leadership teams actively pursue AI adoption plans, only 7% are building agentic AI capabilities that can execute tasks with limited human direction.

Business2 sources

HCLTech Report Reveals 84% of Wealth Firms Need AI Redesign, But Only 7% Build Agentic Capabilities

Business2 sources

Meta Partners with Firmus to Secure AI Computing Capacity Across Southeast Asia

Meta has signed agreements with Australian AI infrastructure provider Firmus to access computing capacity at data centers across Southeast Asia. The deal builds on Meta's existing Melbourne deployment and positions Firmus as a strategic partner for next-generation AI infrastructure in the Asia-Pacific region.

Technology2 sources

Meta Partners with Firmus to Secure AI Computing Capacity Across Southeast Asia

Technology2 sources

Prev 1 2 345…44 Next