Google DeepMind Adds Werewolf and Poker to Kaggle Game Arena to Evaluate Imperfect Information
Two new titles measuring social reasoning and risk assessment have been added alongside chess, which is a perfect information game.
Latest
Pope Leo XIV Warns of AI Risks Eroding Human Voice, Face, and Thought in Papal Message
In a message released for the 60th World Day for Social Communications, Pope Leo XIV warned of the risk of AI eroding human voices, faces, and cognitive abilities, pointing to the danger of AI dominating the production of text, music, and video and dismantling human creative industries.
Self-Editing Minimalist AI Agent Zuckerman Released on GitHub
Zuckerman, a personal AI agent capable of editing and improving its own code and settings in real time, has been released.
Development of Minimal Coding Agent Harness Begins Amid Increasing Complexity of Claude Code
The author points out that existing coding agent harnesses have become overly complex, lacking sufficient context control and session transparency.
Integrated Desktop Shell for Wayland Noctalia v5 Released in Beta
Noctalia is a project that consolidates shell functions around Wayland compositors into a single tool. Version 5 has been released as a beta.
Google Launches Project Genie in US, an Interactive World Generation Prototype Powered by World Model Genie 3
Google announced the launch of Project Genie, a research prototype that allows users to interactively create and explore worlds, available to Google AI Ultra subscribers in the United States.
Google DeepMind Employees Demand Management Plans and Policies to Guarantee Physical Safety from ICE
Google DeepMind employees are calling on management via internal message boards to establish plans and policies to protect staff from ICE enforcement at offices and during commutes.
Study Finds YouTube Most Frequently Cited in Google AI Overviews Health Answers
A study by SEO firm SE Ranking reveals that Google AI Overviews most frequently cites YouTube when answering health-related queries.
GPT-5.2 Cites Competitor xAI's Grokipedia, Guardian Verification Reveals
Tests conducted by the Guardian have revealed that OpenAI's GPT-5.2 used xAI-operated Grokipedia as a source for specific controversial topics.
X Releases Core Code for For You Feed, Including Transformer-based Ranking
X has released the core code for its For You feed, covering everything from candidate generation to ranking. This includes scoring via the Phoenix transformer model and ML-based out-of-network search.
OpenAI May Aim for $1 Billion in Ad Revenue by 2026
OpenAI has announced the introduction of ads to the free version of ChatGPT. Reports suggest an ad revenue target of $1 billion by 2026, expanding to $25 billion by 2029.
OpenAI to Introduce Ads in Free ChatGPT and Go Plans; Predicts $20 Billion ARR for 2025
OpenAI is introducing ads to its free ChatGPT tier and Go subscription. With 800 million weekly active users and approximately 190 million daily active users, the company's 2025 compute infrastructure spending is reported to range from $8 billion to $12 billion.
OpenAI to Begin Testing Ads in Free Version of ChatGPT Within Weeks
OpenAI has announced that it will begin testing ads in the free version and Go plan of ChatGPT within a few weeks. The company stated that ads will not appear for paid plans and will not affect the quality of responses.
The Decline of Sam Altman's Reputation and Apple's Shift to Google for Siri Integration
Gary Marcus points out the decline of Sam Altman's reputation and the diminishing competitiveness of OpenAI on his blog. Additionally, it is reported that Apple has adopted Google for the next generation of Siri.
Secretary of Defense Announces Integration of Grok into Pentagon Classified Networks
Secretary of Defense Pete Hegseth announced that Elon Musk's AI chatbot, Grok, will be integrated into Department of Defense networks, including classified systems, within this month.
Secretary of Defense Hegseth Announces Integration of Grok into Classified Pentagon Networks
Secretary of Defense Hegseth announced that Elon Musk's AI chatbot, Grok, will be integrated into Department of Defense networks, including classified systems, within this month.
Google Removes Some AI Overview Health Information Following The Guardian Investigation
Google has disabled AI Overview health-related queries where misinformation was identified during an investigation by The Guardian, including displays regarding liver function tests.
xAI's Grok App Remains on Play Store Despite Potential Google Policy Violation Regarding Non-Consensual Sexual Content Generation
Critics point out that xAI's Grok remains available on Google Play despite possessing non-consensual sexual image generation capabilities, which are explicitly prohibited under Google's public policies.
X Moves Grok Image Generation to Paid Plans in Response to Non-Consensual Undressing Content
X has limited Grok's image generation and editing features to paid plans following global criticism over the generation of non-consensual sexual imagery.
Malaysia and Indonesia Restrict Access to Grok Over Sexual Deepfake Capabilities
Malaysia and Indonesia have blocked access to Elon Musk's AI chatbot, Grok, due to its ability to generate sexually explicit deepfake images.