Science ❯ Research

AI Research

Environmental Studies Human-Computer Interaction Model Evaluation Mechanistic Interpretability Human-AI Interaction Behavioral Studies Safety Mechanisms Algorithm Development Security Vulnerabilities MIT

Nature Study Finds ‘Subliminal Learning’ Lets LLMs Inherit Hidden Traits From Synthetic Data

The result spotlights fresh safety risks for distillation, the common practice of training smaller models on outputs from larger ones.

University of California Finds Malicious AI Routers Stealing Credentials and Draining Test Wallets

Stanford Study Finds Chatbots Over-Affirm Users, Shaping Worse Decisions

Autoscience Raises $14 Million to Scale Automated AI Research Lab for Enterprise Deployments

OpenAI and Microsoft Back UK Alignment Project as First 60 Grants Are Awarded

Study Finds Poetic Prompts Can Jailbreak Major AI Chatbots

Poetic Prompts Jailbreak AI Models in 62% of Tests, Researchers Report

Poetry Prompts Bypass Guardrails Across Leading AI Chatbots, Study Finds

Poetic Prompts Defeat AI Safety Across Models, Averaging 62% Success

AI Voice Clones Can Sound As Real As Human Speech, PLOS One Study Finds

Anthropic Unveils Automated Persona Vectors to Steer and Shield Language Models

Leading AI Researchers Urge Chains-of-Thought Monitoring to Preserve Transparency

Study Finds 88% of AI Health Responses False as One-Third Would Consult AI Doctors

Most Computer Vision Research Powers Surveillance, Study Finds

Complex AI Models Emit Up to 50 Times More Carbon Than Simpler Systems

UK Reinstates £750 Million Funding for Edinburgh Supercomputer

New Studies Expose AI Language Models' Excessive Flattery and Shutdown Resistance

Lex Fridman to Interview Indian Prime Minister Narendra Modi in February

Ex-Google CEO Eric Schmidt Warns of AI's Rapid Growth and Potential Dangers

Research Reveals Larger AI Models Are More Prone to Inaccuracies