AI News

UK AI firm Synthesia secures $200M funding at a $4B valuation, doubling from a year ago. Led by GV with NVIDIA's participation, funds will advance interactive AI video assistants.....

JetBrains announced that starting from the 2025.3 version, its full range of IDEs will natively integrate the OpenAI Codex model. Developers can directly call this model in the AI chat function to handle complex development tasks. The integration supports multiple activation methods: users can use the JetBrains AI native subscription, bind a ChatGPT account, or enter an OpenAI API key. It covers IntelliJ IDEA, PyCharm, WebStorm, and Rider.

On Jan 26, Clawdbot gained popularity by integrating MiniMax M2.1 model, surpassing 30k GitHub stars. User reviews highlight M2.1's high accuracy in full-stack tasks and tool calls, boosting productivity at low cost, praised by developers.....

Alibaba's Tongyi Qianwen app introduces an AI 'call to book restaurants' feature, praised for its natural and realistic voice, challenging stereotypes of robotic AI communication.....

The AI boom drives soaring demand for computing power, causing hardware supply-demand imbalance. Packaging and testing sectors raise prices due to AI chip demand and cost increases, while the memory industry faces supply shortages and price fluctuations, signaling profound industry transformation.....

Tencent Hunyuan launches Image 3.0, an 800B-parameter image-to-image model with a Mixture of Experts architecture, enabling intelligent, instruction-based image editing on Yuanbao Assistant and its official website.....

Apple had discussed a collaboration with AI company Anthropic, planning to completely upgrade Siri using its large model, involving an annual fee of billions of dollars, but the negotiations hit a stalemate last summer.

Tencent Hunyuan launches 80B-parameter Image 3.0 model with semantic understanding and reasoning, enhancing image editing and multi-image fusion, now available on Tencent Yuanbao across all platforms.....

Meta has globally disabled the "AI Role" feature for minors due to compliance risks with AI chatbots. Previously internal reports showed that its AI failed to effectively filter sensitive topics when communicating with minors, sparking controversy.

Google's AI search relies heavily on YouTube over medical websites for health queries, raising concerns about information reliability. A study of 50,000 German health searches found YouTube cited 4.4% of the time, surpassing professional medical sites, intensifying doubts about AI-generated health information accuracy.....

At Tencent's annual meeting, Ma Huateng announced that the AI application 'Yuanbao' will launch a cash distribution activity of 1 billion yuan during the Spring Festival on February 1st, with a maximum of 10,000 yuan per person, aiming to replicate the success of WeChat red envelopes. At the same time, Tencent revealed its previously confidential social AI project 'Yuanbao Party', officially integrating AI into its core social field. The project aims to create a multi-person social space with deep AI involvement, where AI can summarize group chats and take on roles such as fitness and reading companions.

StepFun AI secures over 5 billion yuan in Series B+ funding, setting a domestic record for single-round financing in the large model sector. Led by Shanghai Guotou XianDao Fund and China Life Equity, with Tencent and 5Y Capital participating, the company advances amid industry IPO pursuits.....

Mistral CEO claims China's DeepSeek-V3 uses its architecture, sparking debate over AI innovation attribution, focusing on whether sparse mixture-of-experts technology is borrowed or independently developed.....

JD Technology and Rokid have jointly launched JoyGlance, the world's first smart glasses shopping application. It combines AI large models with waveguide technology to enable voice-activated shopping and simplify user operations.

Tencent's AI assistant 'Yuanbao' has launched an internal test for its social function 'Yuanbao Party', exploring the application of AI in multi-person social scenarios. It aims to create a social space where AI and users can entertain and collaborate together, marking Tencent's AI expansion from efficiency tools to social interaction.

Shanghai Step Star completed a B+ round funding of over 5 billion yuan, setting a new record for the largest single investment in the large model sector in the country in the past year. This round was participated by multiple top-tier industry investment institutions, indicating that the competition in domestic large models has entered a new stage with increased capital and talent investment.

The National Research Center and Weitai Robot jointly launched the world's first large-scale cross-entity vision-touch multimodal dataset, Baihu-VTouch, aimed at solving the challenges of tactile perception in robotic fine manipulation, filling the data gap in this field and providing an important resource for developers worldwide.

The 'White Tiger-VTouch' dataset, jointly released by Guodi Center and Weitai Robotics, is the world's first cross-ontology visual-tactile multimodal dataset for robotic manipulation. It features the largest scale and most complete modalities, pioneering a data collection paradigm for real cross-ontology interactions, including high-precision visual-tactile sensor data, RGB-D depth vision, and joint poses, advancing robotic perception from single....

Shanghai Jiao Tong University launches Optics GPT, a vertical optical model, marking a breakthrough in China's 'AI + hard tech' sector. It acts as a 'digital brain' for precise insights in optical R&D, focusing as a 'senior specialist' in optics, unlike the generalist ChatGPT.....

Qwen officially refutes rumors that AI ordering is actually operated by real people, emphasizing that its AI assistant identifies over 50 complex emotions and matches empathetic responses within 100 milliseconds through a real-time emotion and intent recognition engine, achieving highly realistic natural interaction.