Recently, the Hong Kong Generative AI Research Center (HKGAI), funded by the Hong Kong Special Administrative Region government's "InnoHK Innovation Hong Kong R&D Platform," officially launched its HKGAI V1 large language model. This is Hong Kong's first homegrown AI large language model, marking a new stage in the development of artificial intelligence in Hong Kong. Image Note: Image generated by AI, image authorized by Midjourney. HKGAI V1 is based on DeepSeek full parameter fine-tuning technology and has undergone continuous training, successfully achieving the first DeepSeek...

Today, the first AI short film starring Zhou Hongyi, chairman of 360 Group, officially went live, meeting the audience's anticipation. The film features renowned actress Huang Yi, adding to its appeal. The plot includes exciting fight scenes between Huang Yi and Zhou Hongyi. Zhou Hongyi mentioned on his social media account that he was honored to collaborate with Huang Yi, but also admitted that acting opposite her put a lot of pressure on him, so he used AI technology to "cheat".

Database engine developer MongoDB recently announced the acquisition of AI startup Voyage AI for $220 million. This acquisition aims to help businesses reduce the potential for 'hallucinations' in their AI applications, enhancing the reliability and accuracy of their products. Image note: Image generated by AI, image licensing service Midjourney. MongoDB President and CEO Dev Ittycheria stated in a press release...

Ping An Healthcare and Technology Company Limited has officially launched its AI-powered doctor assistant service, "Ping An Xin Yi." This service will be fully available on the Ping An Good Doctor app, providing users with around-the-clock health consultations. The digital assistant leverages advanced digital human technology and big medical data, combined with natural language processing and machine learning, to efficiently simulate the professional knowledge and experience of a real doctor. Image note: Image generated by AI, image authorization service provider Midjourney

Tesla's China website announced that its FSD (Full Self-Driving) package has been officially renamed "FSD Intelligent Assisted Driving" in the Chinese market, along with an update to its capabilities description. Despite the name change, the price remains at RMB 64,000. Before the renaming, the FSD feature was directly translated as "Full Self-Driving Capability".

Along with the restoration of the recharge service, DeepSeek has also adjusted the pricing for some model calls. The promotional period for the deepseek-chat model has ended, and the pricing for its underlying V3 model has been adjusted to ¥2 per million input tokens and ¥8 per million output tokens. The pricing for the deepseek-reasoner model, based on the R1 model, is ¥4 per million input tokens and ¥16 per million output tokens.

Tesla has officially launched its City Driving Autopilot feature in China. The news quickly sparked discussion on X, marking another significant advancement for Tesla in China's intelligent driving sector. According to a statement released via WeChat, the Autopilot feature intelligently guides vehicles along navigation routes. User @imxiaohu posted on X that the function enables automatic navigation through ramps and intersections.

Baidu announced a significant upgrade to its education and search services. According to recent X platform posts, Baidu Education's 'Photo Search Questions' and 'AI Essay Writing' features have been integrated with the full-fledged DeepSeek-R1 model, providing users with standard answers and detailed solution demonstrations. Simultaneously, Baidu Search has fully integrated DeepSeek and Wenxin large language models, significantly improving the search experience. Since launching the relevant access points on the Baidu homepage, over 10 million users utilized the services within just one hour, sparking widespread attention. X user @TechBitD

Amidst the ongoing success of the Call of Duty franchise, Activision Blizzard has confirmed the use of generative AI in the development of its games. While the player community had already suspected this, evidenced by some unusual character designs, the official statement has drawn significant attention. The Steam page for Call of Duty: Black Ops 6 explicitly mentions "Our team uses generative AI tools to help develop in-game assets." However, specific details about its implementation remain undisclosed. Based on available information...

Baidu Education's integration of the DeepSeek-R1 model aims to enhance the intelligence of its learning tools through deep learning technology. In the photo search for questions feature, DeepSeek-R1 not only provides standard answers but also leverages its powerful reasoning capabilities to demonstrate the problem-solving process step-by-step, helping users truly grasp problem-solving methods. This upgrade addresses learners' need for answers while focusing on cultivating their problem-solving thinking, embodying the educational philosophy of "teaching people to fish, not just giving them fish."

On February 25th, JD Retail's technology team announced the official launch and full opening of its self-developed JingDianDian AIGC content generation platform. This platform aims to provide e-commerce merchants with a highly efficient and cost-effective content generation solution using AI technology. Currently, the platform covers over 20 core scenarios, with daily AI capability calls exceeding 10 million times. It assists over 350,000 JD merchants in generating high-quality product images, marketing copy, and soon-to-be-launched main video, significantly improving content production efficiency and reducing production costs.

Chinese e-commerce giant Pinduoduo has announced the formation of a large model team focused on e-commerce recommendations, further strengthening its commitment to AI in the sector. According to recent posts on X, the team is led by a former core member of Baidu's Phoenix Nest (Baidu's advertising system), whose salary is reportedly multiple times higher than at Baidu, demonstrating Pinduoduo's significant investment in talent acquisition. While Pinduoduo has not officially disclosed the leader's identity, an X user @TechBitDaily revealed in a February 23rd post that the team has already recruited numerous individuals from Baidu.

In US cities and suburbs, felled trees often end up being chipped instead of being made into lumber. This frustrated Ben Christensen, who grew up in New Mexico with a family deeply rooted in the lumber industry – his father was a carpenter – instilling in him a deep respect for trees. Image Note: Image generated by AI, image licensing service Midjourney Christensen points out that the main reason for wood waste is...

A leaked iOS 18.4 beta suggests Apple is preparing to integrate Google's Gemini AI model into its ecosystem, offering users a third-party AI option.

Following its latest financial report, Alibaba Group announced a significant forum to be held on February 28th. Hosted by its research arm, DAMO Academy, the event, titled "2025 XuanTie RISC-V Ecosystem Conference," will feature a keynote speech by DAMO Academy's chief scientist on "Opportunities for RISC-V from DeepSeek Innovation." The conference will also see the launch of the XuanTie RISC-V series of chips. Notably, Alibaba Group CEO, Wu Yongming, announced yesterday that the company plans to invest significantly over the next three years.

vivo recently announced a gray-scale update (version 5.2.1.22) to its "Blue Heart Little V" voice assistant, incorporating DeepSeek large model capabilities. This highly anticipated update enables "Little V" to offer more intelligent and in-depth thinking, assisting users in better information processing. User feedback indicates that with the deep thinking (R1) function enabled, "Blue Heart Little V" can conduct more comprehensive and thorough analysis based on user input commands.

DeepSeek announced its second-day product of the open-source week: the first open-source EP communication library for MoE models, enabling full-stack optimization for mixed-expert model training and inference. DeepEP is a high-efficiency communication library specifically designed for Mixture-of-Experts (MoE) and Expert Parallelism (EP). It aims to provide high-throughput and low-latency many-to-many GPU kernel communication, commonly known as MoE routing and aggregation. DeepEP not only supports low-precision operations such as FP8 but also integrates with the DeepSeek-V3 paper.