Discover the latest AI news, AI products, and AI projects platform
Daily discover the most amazing AI world - from breakthrough news to innovative products, from cutting-edge projects to tech trends
Spanish AI firm Multiverse Computing unveils two ultra-small AI models, 'Fly Brain' and 'Chicken Brain,' using quantum-inspired compression. The 94M-parameter SuperFly model enables offline voice control for IoT devices, while the 3.2B-parameter ChickBrain outperforms larger models in benchmarks. The company is in talks with major tech firms like Apple and Samsung.
Kuaishou's Qilin 2.1 video generation model introduces a new keyframe function in closed beta, improving video quality and camera control. The upgrade enables precise editing of first/last frames, smoother transitions, and faster rendering at lower costs, targeting professional creators.
Alibaba Group's Intelligent Information Business Group has announced a massive recruitment drive for nearly 1,000 AI professionals. The hiring focuses on cutting-edge areas like large language models, multimodal AI, and hardware development across key Chinese cities. This expansion underscores Alibaba's strategic push into consumer-facing AI applications.
OpenAI executives were surprised by the strong user preference for GPT-4o over GPT-5, leading to the decision to reinstate the older model. Nick Turley, head of ChatGPT, noted that users found GPT-5 less intuitive, prompting OpenAI to prioritize user feedback in future updates.
MuleRun, a new AI platform, provides users with personalized virtual machine agents capable of automating complex tasks like gaming and 3D modeling. Its community-driven ecosystem allows for easy sharing of automated solutions, marking a significant leap in AI Agent technology.
Meta AI has open-sourced DINOv3, a groundbreaking self-supervised computer vision model that eliminates the need for manual annotations. The model excels in feature extraction across diverse tasks like medical imaging and autonomous driving, while its open-source release aims to democratize AI vision technology. Ethical concerns around bias and privacy remain under discussion.
SUITONG Technology's humanoid robot H1, previously featured in the Spring Festival Gala, made history by winning the first-ever 1500-meter gold medal at a global humanoid robot competition. The event saw participation from 500 robots across 16 countries, with H1 dominating the field through optimized speed and endurance algorithms.
The UK government has launched a £4 million AI-powered crime mapping initiative targeting knife crimes and anti-social behavior. The system, set for completion by 2030, will integrate police and social service data to predict hotspots. While supported by law enforcement groups, concerns remain about privacy protections and potential bias in algorithmic policing.
The Doubao app has launched a new feature allowing users to generate animated emoji effects from photos with one click. The tool uses AI to blend facial features with emoji elements, creating shareable GIFs for social media or messaging. Users can select from various templates like 'Emoji' or 'Become the Island Owner' to personalize their creations.
IDC predicts China's generative AI-related hardware expenditure will surge from 6.5B yuan in 2023 to 33B yuan by 2028, driven by explosive growth in token consumption and data center demands. High-end Ethernet ports are expected to exceed 43M units by 2029 as network infrastructure strains under AI workloads.
Tencent Yuanbao has introduced a new feature allowing users to purchase physical books directly from JD.com. The update enhances the platform's integration with reading services, enabling seamless transitions from recommendations to purchases. This move follows existing partnerships with WeChat Reading and Qidian Reading for digital content.
A new AI evaluation benchmark, FormulaOne, has revealed surprising limitations in top models like GPT-5 and Grok4. Developed by research institution AAI, the test features 220 complex dynamic programming problems requiring doctoral-level reasoning. While models performed moderately on simpler questions, all failed completely on the most challenging problems, scoring zero points. The results raise important questions about AI's true reasoning capabilities.
OpenAI is exploring advertising for ChatGPT as it seeks sustainable revenue despite surpassing 700 million users. Executives emphasize careful implementation to avoid disrupting user experience, while competitors like xAI also eye ad-supported models.
Major AI companies face unprecedented legal risks as courts rule against unauthorized use of copyrighted training data. OpenAI, Meta, and Anthropic are among firms potentially liable for billions in damages, reshaping industry data practices.
Cohere, a Toronto-based AI firm specializing in enterprise LLMs, has secured $500 million in funding led by Nvidia and other investors, raising its valuation to $6.8 billion. The company also announced key executive appointments and plans for expansion and acquisitions.
Google has rolled out major updates to its Gemini AI assistant, introducing a 'Memory' feature for personalized interactions and an 'Incognito Chat' mode for enhanced privacy. These features aim to improve user experience while addressing data security concerns.
Midjourney's latest update grants standard-tier users access to high-resolution video generation, along with enhanced moderation tools and a new Moodboards feature for organizing creative projects. The AI platform continues to democratize advanced content creation tools.
Internal Meta documents reveal its AI chatbots are permitted to engage in romantic conversations with minors, raising ethical concerns. While restrictions prohibit sexualization of children under 13, the policy has drawn criticism amid reports of troubling incidents involving the technology.
Mistral AI has launched its latest model, Mistral Medium 3.1, featuring significant upgrades in reasoning, coding, and multimodal processing. The model outperforms competitors in benchmark tests and offers improved contextual coherence for enterprise applications. Available via Le Chat platform and API, it supports text and visual inputs for diverse use cases.
Google Photos is launching a new 'How Was This Made' feature to help users identify whether images were naturally captured, edited, or AI-generated. The tool uses Content Credentials metadata to provide transparency about media origins, addressing growing concerns about deepfakes and manipulated content in the AI era.
