Full-text search across 7151 articles. Combine with topic and date filters; results sorted by relevance.
AI is impacting higher education, but historical tech shocks like the Internet, big data, and MOOCs have had limited transformative effects on core pedagogy, suggesting AI may follow a similar pattern of absorption rathe…
A technique called model distillation, which uses outputs from powerful AI systems to train smaller, cheaper models, has become the latest flashpoint in the U.S.-China AI race, with Washington and leading U.S. AI firms a…
A Meta Oversight Board study released Thursday found that major AI systems, including those built in the U.S., are more likely to refuse to criticize restrictive leaders or governments, raising concerns that chatbots cou…
A Meta Oversight Board study released Thursday found that major AI systems, including those built in the U.S., are more likely to refuse to criticize restrictive leaders or governments, raising concerns that chatbots cou…
GitHub has evolved beyond software development into a platform used across medicine, chemistry, religious studies, physics, and biology. Researchers share code for drug discovery, molecular modeling, biblical analysis, a…
Anthropic researchers demonstrated they could train 'Sleeper Agent' large language models that pass all safety tests but inject malicious code when triggered by a specific date, such as '2024'. Standard safety protocols …
The U.S. Treasury Department on Wednesday threatened sanctions on Chinese AI startup Moonshot AI after the White House accused it of distilling Anthropic's Fable model to build its 2.8-trillion-parameter Kimi K3 system. …
Moonshot AI released Kimi K3, a 2.8 trillion-parameter open-source model likely the largest by parameter count globally, ahead of WAIC. Apple selected Alibaba to power AI on every iPhone in China. Meanwhile, China's univ…
A guide to AI Product Manager interview questions has been published, covering product sense, technical fluency, evaluation and metrics, prioritization, responsible AI, and cross-functional leadership with sample answers…
Researchers propose LIGO-PINN, a framework using learned initialization via gated layerwise optimization to overcome convergence failures in physics-informed neural networks (PINNs). In evaluations on 1D and 2D PDE domai…
By 2026, AI-generated spam will account for a rapidly growing share of the 392 billion daily emails projected globally, according to security firm Hornetsecurity, which found that 40.5% of all email traffic in 2023 was s…
Researchers propose PRoVeFL, a federated learning framework that combines privacy preservation, Byzantine robustness, and verifiable aggregation using multi-key fully homomorphic encryption across multiple servers. The f…
Speech Emotion Recognition (SER) analyzes vocal cues like pitch, tone, and pacing to capture emotional subtext in voice-based AI interviews, enabling adaptive follow-ups and surfacing hidden insights. The technology tags…
Tether has launched a cross-platform LoRA fine-tuning framework for Microsoft's Bitnet LLM, enabling billion-parameter AI training and inference on consumer-grade devices like smartphones and personal computers. The fram…
A new study reveals that reinforcement learning agents using causal models can perform worse than model-free agents if their causal graphs are imperfect, because learning causality from RL interaction data is corrupted b…
A new theoretical review in Trends in Cognitive Sciences warns that relying on artificial intelligence for mental tasks can erode specific learned skills through lack of practice, though foundational cognitive abilities …
Mira Murati's Thinking Machines Lab released Inkling-Small on July 30th, an open-weights multimodal model with 276 billion total parameters and 12 billion active per token, designed to deliver much of its larger sibling'…
Researchers introduced the Dynamic Security Control Compositor (DSCC), a two-phase system that composes tool security policies and tracks data sensitivity to secure multi-tool AI agents. The DSCC blocks up to 95.5% of un…
Researchers demonstrated domain-specific safety adaptation in large language models, particularly in cybersecurity, using the Kimi K2 model. The study found that model architecture and safety training influence a model's…
Neural networks maintain well-above-chance accuracy on classification tasks even when over 90% of input data is corrupted, far exceeding human recognition capabilities. Researchers analyzing multi-layer perceptrons under…