AI News

iQIYI launched the 'AI Celebrity Library' plan, using its self-developed platform to create digital avatars of celebrities, aiming to improve the efficiency of film and television production. However, after the plan was announced, multiple participating celebrities quickly spoke out to deny their involvement, sparking a wide public discussion on the boundaries of AI technology application and protection of celebrities' rights.

iQIYI launched an 'AI Actor Library,' allowing actors to authorize the creation of digital virtual identities. Some top-tier actors have already joined. This technology digitizes actor appearances, using them for short dramas and interactive videos while protecting image rights.

The domestic AI model Qwen3.6-35B-A3B is officially open-sourced, using a hybrid expert architecture. It has a total of 35 billion parameters but activates only 3 billion during inference, achieving 'winning with small strength' high efficiency performance, significantly reducing computing costs.

Alibaba Cloud announced that starting from April 28, 2026, the multi-modal interaction development kit of the "BaiLian" large model service platform will be subject to rate limiting, with the newly established connection count uniformly adjusted to 10 QPS. This move aims to optimize resource scheduling and ensure service stability.

Aliyun has adjusted the default API rate limiting for the BaiLian Multimodal Interaction Development Kit, effective April 28, 2026. The new limit is set to 10 calls per second (QPS) to optimize resource scheduling and service stability. It also supports 600 new sessions per minute and 36,000 new sessions per hour, meeting most development testing and daily business needs.

Alibaba launches the preview version of Qwen3.6-Max-Preview model, which can be accessed through QwenStudio for conversation or via Alibaba Cloud BaiLian API. Compared to Qwen3.6-Plus, the new model shows significant improvements in agent programming, world knowledge, and instruction following, and performs well in six major programming benchmarks.

世界のAI光受発信モジュール市場は急速に拡大しており、2026年には260億ドルに達すると予測され、年間成長率は57%以上となる。この成長は主にAIデータセンター建設の加速に起因し、光通信サプライチェーンの構造的な再編を促進し、特に800G以上の高速モジュール需要が急増している。

Huawei launched the XMAGE Smart Shot function at the new product launch event of the Pura series, integrating technologies such as AI pose recommendation, AI assisted composition, and 3D dynamic photos to enhance the convenience and intelligence of photography. The AI pose recommendation is trained on 20 million data points and can analyze user needs and provide the best shooting suggestions within 300 milliseconds.

Tongyi Lab launches the Fun-ASR1.5 speech recognition large model, achieving a balance between versatility and accuracy through a unified architecture. The model supports 30 mainstream languages globally and is deeply adapted to the seven major Chinese dialects and over 20 regional accents, demonstrating outstanding performance in multilingual, multi-dialectal, and complex contexts.

Microsoft's CEO announced the early launch of the Fairwater AI data center in Wisconsin, USA. With an investment of approximately 22.5 billion yuan, it spans 127.6 hectares, houses hundreds of thousands of NVIDIA GB200 superchips, and offers performance ten times that of the world's fastest supercomputer for cutting-edge AI model training.....

The high-end smart speaker Huawei Sound X 5 starts pre-sale on April 20. The core highlight of the new product is the first deep integration of an AI large model, achieving an upgrade from "auditory interaction" to "smart companionship." In terms of acoustics, it continues the family's high specifications and is expected to adopt a multi-unit configuration.

The efficiency of large language model inference has made a breakthrough. Tsinghua University and Moonshot AI jointly proposed a new architecture called "Prefill-as-a-Service," which splits the inference process into two stages: prefilling and decoding, and optimizes the allocation of computing resources, effectively solving hardware limitations and significantly improving model service performance.

At the Beijing International Film Festival, Tim from Film Turbulence shared their AI video creation journey, including backlash after labeling AI-generated content. They used AI in drone reviews to highlight product features.....

Ant Lingguang App launched the 'Lingguang Flash Application Creator Incentive Program', investing 100 million yuan to support high-quality flash applications and creators. The program will start on April 20th, offering up to 1 million yuan per day to reward 10,000 high-quality applications, and up to 1 million yuan per week to reward 100 outstanding creators.

Anthropic engineers introduced Claude Buddy, an open-source hardware device that transforms the AI coding tool Claude Code into a palm-sized desktop companion. It turns code approval into interactive fun, addressing frequent manual authorizations for sensitive operations, sparking discussions among developers.....

Musk proposed the concept of 'universal high income' to address potential large-scale unemployment caused by AI and automation. He suggested that the US government should distribute funds to make people's lives more prosperous, believing that in the future, AI and robots will greatly increase the production of goods and services.

UK AI Minister Kendall stated that she does not use AI at work but actively uses it in her private life, such as analyzing the causes of skin allergies from cosmetics. She recently announced the establishment of a £500 million fund to support UK AI companies, emphasizing its importance for national security and the economy.

iQiyi CEO Gong Yu announces a shift towards decentralization, empowering creators to upload content independently. AI technology will reduce production costs and cycles, significantly increasing creator and content numbers, moving away from traditional centralized decision-making.....

Claude Design is an experimental product launched by Anthropic. Users can generate professional design drafts, prototypes, presentations, and more through text, images, or documents, greatly lowering the barrier to visual content creation. This marks Claude's accelerated evolution from a single dialogue model to a multi-modal productivity tool.

Ant Lingguang App upgrades and launches the "Lingguang Circle," creating a consumer-level Coding Agent. Building on the "30-Second App Generation" feature, it enhances multi-agent collaboration, full-modal generation, and mobile integration, becoming the first platform to allow users to create, distribute, use, and iterate AI applications on their smartphones using natural language, achieving 0-code, 0-deployment, and 0-barrier creation. Currently, users have created over 30 million flash applications.