A controversial AI startup, Cluely, raised $5.3M despite backlash over its unethical service enabling users to cheat on job interviews and dates using AI-generated responses, with the CEO proudly boasting over $3M in annual revenue from the cheating tool.
Google has regained its dominance in the AI field, with its powerful models like Gemini 2.5 outperforming rivals like OpenAI and Anthropic, thanks to its unparalleled AI capabilities, talent, infrastructure, and resources, allowing it to reclaim a commanding lead after initially stumbling with ChatGPT.
IBM VP David Cox warns that while language models appear intelligent, they lack true understanding, and the real risks of current AI systems are being overlooked amidst the hype around artificial general intelligence (AGI).
OpenAI develops X-like social network prototype centered on AI image generation, with CEO Sam Altman seeking external feedback on the platform that would directly compete with Elon Musk and Meta while providing valuable real-time data for training AI models.
Apple is pioneering private AI by leveraging synthetic data and differential privacy to analyze user data securely, generating artificial datasets mimicking real data to enhance AI models for features like Genmoji, Image Playground, and email summaries while preserving privacy.
Anthropic's AI assistant Claude gains autonomous research capabilities to provide comprehensive answers rapidly, and integrates with Google Workspace for secure data access, promising enhanced productivity through intelligent analysis of users' emails, calendars, and documents.
Meta plans to utilize public content and user interactions from EU users to train its AI models for better local language and cultural understanding, while offering an opt-out option to comply with regulations and enhance transparency.
Meta's open-source Maverick AI trails rivals like GPT-4 and Claude on chat benchmarks, but the company hopes developers will customize the model for diverse applications, leveraging its conversational capabilities.
Researchers developed a multimodal AI tool that combines language models and graph-based models to efficiently design new molecules with desired properties from natural language queries, outperforming existing methods and potentially automating the entire molecule design and synthesis process.
BrowseComp, a novel AI benchmark, evaluates web browsing skills by posing 1,266 complex queries requiring synthesis of information across multiple websites, with a deep research model achieving 51.5% accuracy, surpassing other models' performance on this challenging task.