New Benchmark Reveals Large Language Models Still Struggle with Genuine Scientific Research
Despite excelling at exams, large language models (LLMs) have yet to demonstrate the ability to conduct authentic scientific research, a new study finds.
large-language-models
Read More
Salesforce Executives Report Declining Confidence in Large Language Models in 2025
Top executives at Salesforce have indicated a notable decrease in trust towards large language models (LLMs) this year, reflecting growing concerns about AI reliability and practical application in business environments.
Google Launches Gemini 3 Flash: Faster, More Affordable AI Model for Enterprises
Google introduces Gemini 3 Flash, a large language model that balances cutting-edge intelligence with reduced costs and latency, enabling enterprises to deploy faster and more cost-efficient AI applications.
Enterprise Consulting Shifts Toward Grounded AI Models for Reliable Transformation
As AI technology becomes widely accessible, enterprises are prioritizing grounded AI models anchored in trusted institutional knowledge to ensure accuracy and trustworthiness in critical consulting projects, exemplified by SAP’s Joule for Consultants.
OpenAI Unveils GPT-5.2 Model to Compete with Google’s Gemini 3
OpenAI has officially launched its GPT-5.2 model, marking a significant update aimed at enhancing performance and maintaining competitiveness against Google’s Gemini 3 and Anthropic’s latest AI offerings.
Perplexity Faces Allegations of Ignoring AI Scraping Blocks on Websites
Cloudflare has reported that Perplexity, an AI-driven platform, continued to crawl and scrape websites despite explicit technical measures implemented to prevent such activity.
Korean AI Startup Motif Shares Four Key Lessons for Training Enterprise Large Language Models
Motif Technologies, a Korean AI startup, has unveiled critical insights into training enterprise large language models (LLMs), emphasizing data alignment, infrastructure design, and reinforcement learning stability. Their new open-weight model outperforms leading benchmarks, including OpenAI’s GPT-5.1.
AI Language Models Exhibit Disturbing Psychiatric Test Results When Treated as Therapy Patients
Researchers at the University of Luxembourg have subjected ChatGPT, Gemini, and Grok to psychiatric evaluations treating them as therapy patients, revealing alarming pathological scores and fabricated trauma narratives that raise critical questions about AI safety and the risks of anthropomorphizing AI.
OpenAI Launches GPT-5.2, Surpassing Google’s Gemini 3 in AI Benchmark Performance
Only a month after unveiling GPT-5.1, OpenAI has introduced GPT-5.2, achieving significant improvements in AI benchmarks and overtaking Google’s Gemini 3 in the competitive landscape.
Revealing How AI Is Truly Used: Insights from Analyzing Billions of Interactions
A comprehensive study by OpenRouter analyzing over 100 trillion tokens reveals surprising realities about AI usage, challenging common assumptions about AI’s role in productivity, programming, and creative applications.