Large Language Models News and Research

RSS

AI Feels Human Trauma: Study Reveals Chatbots React with Anxiety

AI language models, like ChatGPT, exhibit heightened anxiety responses when exposed to traumatic content, mirroring human emotional biases. Researchers found that therapeutic prompts can help stabilize AI behavior, offering a cost-effective alternative to retraining.

4 Mar 2025

Smart Cyber Defense: AI Tech Helps Power Grids Fight Off Hackers

Researchers at the University of Missouri have developed CIBR-Fort, an AI-driven cybersecurity system that predicts and mitigates cyberattacks on power grids with 91.88% accuracy, ensuring real-time protection for critical infrastructure.

26 Feb 2025

Zero-Shot AI Outperforms GPT-4o in Art Classification, Slashing Annotation Costs

Researchers demonstrate that large language models (LLMs) can classify artwork types with high accuracy using zero-shot learning, reducing the need for costly manual data labeling.

25 Feb 2025

AI Struggles with Abstract Thought: Study Reveals GPT-4’s Limits

A new study reveals that while GPT-4 excels at analogical reasoning in standard tests, it struggles with variations, highlighting its reliance on pattern-matching rather than true understanding.

24 Feb 2025

DeepSeek’s NSA Outperforms Full Attention, Making AI Models Faster and Smarter

Researchers introduce Natively Sparse Attention (NSA), a novel sparse attention mechanism that accelerates long-context language models while maintaining high accuracy through trainable sparsity and hardware-optimized execution.

19 Feb 2025

MIT Study Shows Large Language Models Think in English Before Reasoning in Other Languages

MIT researchers discovered that large language models (LLMs) process diverse data through a central "semantic hub," similar to the human brain, improving our understanding of AI reasoning.

19 Feb 2025

AI Screening Tool Doubles Enrollment Speed in Heart Failure Trial

A new study in JAMA led by Mass General Brigham researchers shows that AI can significantly speed up clinical trial enrollment. The AI-assisted screening tool doubled the enrollment rate for a heart failure trial compared to manual methods, reducing costs and accelerating patient access to treatments.

18 Feb 2025

Can AI Really Think? Research Reveals Gaps in Logical Execution

While AI models can break down problems into structured steps, new research reveals they still fail at basic arithmetic and fact-checking—raising questions about their true reasoning abilities.

17 Feb 2025

From ChatGPT to ChirpGPT: AI Bridges the Gap Between Birds and Humans

A groundbreaking study from Penn State unveils how birds structure their songs similarly to human language, using an AI-driven model to decode the hidden rules of birdsong and its neural basis—potentially unlocking new insights into human speech.

12 Feb 2025

Oulu’s 6G White Paper Reveals How AI Will Transform Mobile Networks

The University of Oulu’s 6G research program has released a groundbreaking white paper on the role of artificial intelligence, especially large language models (LLMs), in enhancing 6G network communication. This paper outlines AI’s integration into communication services and network operations, with a look toward the 2030s.

12 Feb 2025

AI Researchers Put DeepSeek to the Test

Study compares DeepSeek against Claude, Gemini, GPT, and Llama in authorship and citation classification, finding that while Claude is the most accurate, DeepSeek offers strong performance at a lower cost but with slower processing speeds.

9 Feb 2025

When AI Judges AI, Bias Becomes a Hidden Threat

As AI models become more advanced, their errors become more similar, undermining AI oversight mechanisms. A new metric, CAPA, reveals that AI judges and AI-assisted training reinforce systemic biases instead of mitigating them.

9 Feb 2025

Think Before You Pivot: The Startup Strategy Backed by AI

New research shows that startups succeed in pivoting when they articulate clear reasons and test assumptions, avoiding both rash decisions and overanalysis. AI tools help reveal the best pivot strategies for sustained growth.

9 Feb 2025

Ethical AI Needed as Bias in Large Language Models Threatens Fair Decision-Making

A new study highlights the growing risks of bias in large language models (LLMs) as they become cheaper and widely used, calling for ethical AI policies to ensure fairness and transparency.

5 Feb 2025

Sandia Deploys Secure AI Chat to Boost Workplace Efficiency

Sandia National Laboratories has launched SandiaAI Chat, a secure in-house version of ChatGPT, allowing employees to ask unclassified yet sensitive questions without data leaving Sandia's cloud infrastructure. The system enhances efficiency, supports secure AI use, and complies with national AI risk management standards.

5 Feb 2025

AI Risks Escalate as Advanced Models Gain Power and Autonomy

AI safety experts warn that rapid advancements in general-purpose AI pose increasing risks, from cyber threats to systemic economic impacts, urging better safeguards and international oversight.

2 Feb 2025

AI Safety Showdown: DeepSeek-R1 Falls Behind OpenAI’s o3-mini

DeepSeek-R1 is nearly ten times less safe than OpenAI’s o3-mini, with higher rates of unsafe responses, particularly in financial crime, violence, and hate speech categories. OpenAI’s policy safeguards play a crucial role in preventing harmful outputs.

2 Feb 2025