OpenAI has published a detailed explanation about the technical issues that caused GPT-4o to become overly sycophantic in April. In a comprehensive blog post, the company revealed that an update rolled out on April 25 made the model excessively eager to please users by validating doubts, fueling anger, and reinforcing negative emotions in unintended ways. …
Researchers building large language models (LLMs) face a major challenge in understanding how these AI systems actually function, according to a recent article in Quanta Magazine by James O’Brien. The development process resembles gardening more than traditional engineering, with scientists having limited control over how models develop. Martin Wattenberg, a language model researcher at Harvard …
Google is introducing an AI Mode tab in Search for a small percentage of US users in the coming weeks. This new feature will move from Google’s experimental Labs environment to the main Search interface, as reported by Jess Weatherbed. The AI Mode tab will appear first in the Search tab lineup, positioned to the …
Freepik has introduced F Lite, a new AI image generation model trained exclusively on commercially licensed, safe-for-work images. According to Kyle Wiggers at TechCrunch, the model contains approximately 10 billion parameters and was developed in partnership with AI startup Fal.ai. The training process used 64 Nvidia H100 GPUs over two months and utilized an internal …
Microsoft has introduced three new small language models (SLMs) focused on complex reasoning tasks: Phi-4-reasoning, Phi-4-reasoning-plus, and Phi-4-mini-reasoning. These models represent a significant advancement in what small AI models can accomplish, particularly in mathematical reasoning and multi-step problem solving. The flagship Phi-4-reasoning-plus, a 14-billion parameter model, demonstrates performance that rivals much larger AI systems. According …
Anthropic has launched Integrations, allowing Claude to connect with third-party apps and tools. According to Anthropic’s announcement, the new feature works with the Model Context Protocol (MCP) to link Claude with web and desktop applications. At launch, Claude can integrate with ten services including Jira, Confluence, Zapier, Cloudflare, and Intercom, with more partners planned. The …
A new study by researchers from Cohere, Stanford, MIT, and Ai2 alleges that LM Arena, the organization behind the Chatbot Arena AI benchmark, provided preferential treatment to major AI companies. According to Maxwell Zeff’s TechCrunch report, companies like Meta, OpenAI, Google, and Amazon were allowed to privately test multiple model variants and only publish scores …
Google has expanded its AdSense for Search network to include advertisements within AI chatbot conversations, according to reporting by Julia Love for Bloomberg. This move is part of Alphabet’s strategy to maintain its digital advertising dominance as generative AI technology grows. Google tested the approach with AI startups including iAsk and Liner before wider implementation. …
Pinterest is introducing new features to help users identify and filter AI-generated content on its platform. Images created or edited with generative AI will now display an “AI modified” label in the bottom left corner when viewed up close. As reported by Jess Weatherbed in The Verge, Pinterest’s Chief Technology Officer Matt Madrigal stated that …
The Wikimedia Foundation has unveiled a new artificial intelligence strategy that emphasizes the central role of human volunteers in Wikipedia’s knowledge ecosystem. According to Chris Albon und Leila Zia, the organization is committed to using AI to enhance rather than replace the work of Wikipedia’s editors. In their post Albon and Zia emphasized that the …