Recently, Teslaun and UBTECH made a joint announcement declaring that they have entered into a strategic partnership. Leveraging Teslaun's technologically advanced closed - loop system for spatial intelligence perception, reasoning, and action, as well as UBTECH's comprehensive in - house research and development capabilities in the realm of humanoid robots, the two entities are set to embark on a profound collaborative journey. This cooperation will revolve around the innovative fusion of "Spatial Intelligence + Embodied Intelligence", aiming to unlock new frontiers in technological advancements.

On September 17, iFLYTEK unveiled multiple AI new products in Nanning based on the Spark ASEAN Multilingual Large Model Foundation, including a translation SaaS platform, a dual-screen translation device 2.0, and more.

The head of the Department of Employment, Income Distribution, and Consumption at the National Development and Reform Commission revealed that forthcoming actions will center on elevating the quality and efficiency of service provision through three key strategies: spurring corporate innovation by adopting the 'Artificial Intelligence Plus' approach to drive the convergence of diverse sectoral consumption; fostering industry refinement by setting up training hubs in partnership with educational bodies and businesses, rolling out the 'Credit Plus' program in crucial sectors like childcare and elderly care, and enhancing the mechanisms for rewarding honesty and penalizing dishonesty; and invigorating market dynamism by accelerating the development of a unified, large-scale market, dismantling entry barriers, and leveraging communication channels with private enterprises to bolster their growth.

On September 17, Feizhu AI unveiled a new photo-explanation capability within its "Ask" function. This innovative feature allows users to snap photos of cultural relics or architectural structures at popular tourist destinations, including museums and historical landmarks. Subsequently, they can receive professional voice-guided explanations about these items. At present, the feature is accessible to a select group of users, with plans for a wider release in the near future.

As reported by The Verge, YouTube unveiled a suite of AI-driven functionalities at the "Made On YouTube" event in New York, designed to streamline and enhance the video creation and promotion process for creators. Among these innovations is the Ask Studio chatbot, which leverages data analytics to provide creators with insights into video performance, audience feedback, trending clip identification, and actionable suggestions for content refinement.

Additionally, YouTube has introduced an A/B testing feature for thumbnails and titles, enabling creators to experiment with multiple combinations simultaneously to gauge their impact. The platform's automatic dubbing technology has undergone significant enhancements, ensuring more natural lip-syncing. Furthermore, AI-generated dubbed content will be distinctly labeled beneath the video title and within the description, ensuring transparency.

YouTube also announced plans to integrate a tailored version of Google's Veo 3 text-to-video generative AI model, facilitating the rapid generation of short video clips for creators. The "Inspiration" tab in YouTube Studio will now offer AI-assisted video concept suggestions, delivering personalized concepts, inspiration, and ideas tailored to the creator's channel and specific queries.

Moreover, YouTube has launched community features that foster direct interaction between creators and their fans. The platform is also set to introduce a "Hype" feature, aimed at amplifying the visibility of creators with smaller channels. These features will be progressively rolled out to creators globally, enhancing both content production efficiency and dissemination effectiveness.

During institutional research sessions, iFLYTEK clarified that the revenue figures for intelligent hardware reported in its main business financial statements do not encompass learning machines. As the capabilities of the iFLYTEK Spark large-scale model continue to enhance, its proprietary range of intelligent hardware—including translation devices, voice recorders, and office notebooks—are becoming increasingly adept at professionally and securely addressing user needs in real-world scenarios. Consequently, it is anticipated that the annual revenue derived from intelligent hardware will demonstrate steady growth, with particularly promising growth prospects for overseas consumer-end hardware.

In the wee hours of September 17, Alibaba made its first deep research Agent model, Tongyi DeepResearch, available as open-source. This model delivered state-of-the-art (SOTA) performance on several authoritative benchmarks, including HLE, BrowseComp-zh, and GAIA, outperforming models like OpenAI's Deep Research and DeepSeek-V3.1. Presently, the model, along with its framework and solution, is fully open-sourced, enabling users to download both the model and its code from the Github, Hugging Face, and ModelScope communities.

To tackle the challenges of "cognitive overload" and "irreversible noise interference" that current deep research models encounter when dealing with long-term tasks, the Tongyi team has developed a comprehensive training pipeline powered by synthetic data. This pipeline covers both pre-training and post-training stages, substantially boosting the model's iteration speed and generalization abilities. During the inference stage, the team introduced two modes: ReAct and Heavy (built on their proprietary IterResearch), facilitating high-quality reasoning even in extended tasks.

His journey with deep learning began in high school. During his undergraduate years, he not only ventured into entrepreneurship and took part in robotics projects but also interned, playing a pivotal role in the development of Gemini 2.0. His expertise seamlessly bridges the realms of AI and philosophy. Remarkably, he wrapped up his Ph.D. studies at MIT and aced his thesis defense in under four years. Looking ahead, he's set to propel the frontier technology of 'world models' forward at OpenAI, a move that could potentially redefine the trajectory of artificial general intelligence.

While on his visit to the UK, U.S. President Trump witnessed the UK's Department for Science, Innovation and Technology unveil a landmark "Prosperity in Science and Technology Agreement" between the two nations. Under this pact, American tech behemoths have earmarked over £31 billion (equivalent to $42.3 billion) for cutting-edge technology initiatives, spanning artificial intelligence and quantum computing. Trump was accompanied on this diplomatic mission by NVIDIA's visionary leader Jensen Huang and OpenAI's CEO Altman. Notably, OpenAI is set to introduce its ambitious "Stargate" project to British shores.

Deepdub has released its latest voice AI model, Lightning 2.5, which is the company's fastest and most scalable voice solution to date, designed specifically for real-time production-grade voice experiences. Built on its proprietary foundational model, it supports multiple languages while preserving the speaker's identity, tone, and emotion. The model is equipped with a redesigned inference engine, delivering a 2.8x increase in throughput, a 5x enhancement in concurrency, and reducing latency to 200 milliseconds. Additionally, the model has been optimized for various NVIDIA GPUs, supporting both on-premises and cloud deployment. Lightning 2.5 is suitable for multiple industries, and Deepdub plans to conduct a live demonstration at the 2025 International Broadcasting Convention.

The AI-driven code review startup CodeRabbit has recently wrapped up a substantial Series B funding round, securing $60 million and achieving a valuation of $550 million. This latest funding effort was spearheaded by Scale Venture Partners, with additional support from NVentures and CRV. This brings CodeRabbit's cumulative funding to an impressive $88 million. To date, CodeRabbit has assisted more than 8,000 companies in boosting their code review efficiency, effectively cutting their manpower requirements in half. The company's overarching aim is to furnish developers with more thorough and insightful code review solutions.

According to reports, OpenAI has outlined a share - allocation plan as part of its restructuring strategy, with the company now valued at a staggering $500 billion. Let's delve into the distribution of these shares among various stakeholders.

Microsoft, acting as the largest external shareholder, has secured around 28% of the shares. At the current valuation, these shares are worth approximately $140 billion. This substantial stake reflects Microsoft's significant investment and strategic partnership with OpenAI.

The non - profit board of directors of OpenAI has been allocated roughly 27% of the shares, which are valued at about $135 billion. This arrangement underscores the unique governance structure of OpenAI, where the non - profit arm plays a crucial role in guiding the company's long - term vision and ethical considerations.

Current and former employees have a collective holding of approximately 25% of the shares, amounting to around $125 billion. This demonstrates OpenAI's commitment to rewarding and retaining its talent pool, as employees are directly invested in the company's success.

Investors participating in the new financing round, led by SoftBank, jointly hold around 13% of the shares, with a value of approximately $65 billion. Their involvement indicates confidence in OpenAI's future growth prospects and the potential of its cutting - edge AI technologies.

Finally, early investors have been granted approximately 1% of the shares, valued at $5 billion. Their early support has played a vital role in OpenAI's initial development, and this share allocation serves as a recognition of their contribution.

On September 16th (local time), entertainment behemoths Disney, Universal Pictures, and Warner Bros. Discovery collectively initiated legal proceedings against the Chinese artificial intelligence enterprise, MiniMax, in the Federal District Court of California. They accuse MiniMax's Conch AI service of flagrant copyright infringement. The lawsuit contends that MiniMax, without proper authorization, utilized renowned characters from these three major studios for promotional activities, disseminating them via generated content. Prior to this legal action, the studios had already urged MiniMax to cease such infringements. However, MiniMax neither replied to their requests nor implemented a reasonable safeguard mechanism.

On Tuesday (local time), Matthew Lane, the father of 16-year-old Adam, who took his own life this spring, delivered emotional testimony before Congress. In a heartfelt and impassioned address, he laid the blame for his son's tragic death on OpenAI's ChatGPT, asserting that the company had placed greater importance on product development and market expansion than on the well-being and safety of young users. Several weeks prior, Matthew Lane and his wife had already initiated legal proceedings against OpenAI and its CEO, Sam Altman. Earlier on the same Tuesday, Altman shared in a blog post that OpenAI is set to roll out a series of new safety initiatives aimed at protecting teenagers. These measures include age-estimation technology, parental-controlled "restricted time windows" (a more idiomatic English phrase replacing the literal translation of "disabled time periods" for clarity and naturalness), and limitations on discussions centered around suicide and self-harm. Furthermore, during the hearing, the AI firm Character.AI also came under scrutiny and faced criticism for its handling of similar issues.

On the afternoon of September 17, 2025, the Hang Seng TECH Index surged by over 4%, and the Hang Seng Index rose by over 1%. Among them, Baidu Group surged nearly 20%, NIO rose by over 10%, and JD.com and Alibaba increased by over 5%.

On September 17, John Lee, the Chief Executive of the Hong Kong Special Administrative Region, announced in the latest policy address that the Innovation, Technology and Industry Bureau of Hong Kong is actively facilitating the setup of two pilot production lines at the Hong Kong Microelectronics Research and Development (R&D) Institute. Moreover, it is anticipated that the establishment of the Life and Health R&D Institute, as well as the Hong Kong Artificial Intelligence R&D Institute, will be finalized next year.

In the meantime, Hong Kong is set to expedite the development of the third "InnoHK Innovation Hong Kong R&D Platform," which will zero in on sustainable development, energy, advanced manufacturing, and materials. The corresponding R&D centers are slated to be established in the first half of 2026.

The much - anticipated "Venture Capital Fund for Innovation and Technology Industries" is projected to kick off in the 2026 - 2027 fiscal year. Through this initiative, the government aims to steer market investments towards strategic emerging and future industries.

Even with the ongoing enhancements in the performance and cost-efficiency of large-scale models, three significant challenges continue to plague their practical implementation in business settings: the vague articulation of requirements, the indistinct delineation of model capabilities, and the complexities involved in dynamically adjusting to contextual changes. To tackle these issues head-on, Volcano Engine has rolled out the PromptPilot platform. This innovative solution leverages a systems engineering methodology that integrates "data-driven strategies with interactive collaboration" to cultivate three fundamental capabilities: guided exploration of requirements, automated refinement of prompts, and a closed-loop analysis of problematic cases. The platform presents a structured functional architecture that spans the entire spectrum of requirements, catering to a wide array of scenarios and ensuring compatibility across various models. Real-world application scenarios have shown that it can markedly boost efficiency and diminish the hurdles faced by both business professionals and AI developers. Moving forward, Volcano Engine is poised to propel the platform's capabilities in multiple directions, nurturing an ecosystem conducive to the widespread adoption of large-scale model applications.

On September 17, John Lee, the Chief Executive of the Hong Kong Special Administrative Region, announced in the latest policy address that the government is set to boost the development of the AI and data science sectors. The government's previously unveiled HK$3 billion 'Frontier Technology Research Support Scheme' is now open for applications. This initiative aims to assist local institutions in drawing international research experts in areas like AI and to advance fundamental research in cutting-edge technologies.

In addition, the government has allocated HK$1 billion to strategize the establishment of the 'Hong Kong Artificial Intelligence Research and Development Institute' in 2026. The institute's mission is to streamline upstream AI R&D efforts, facilitate midstream and downstream technology transfer (i.e., the commercialization of research outcomes), and broaden the scope of AI applications.

Furthermore, the Hong Kong Investment Corporation Limited, a long-term investment institution wholly owned by the government, has already invested in numerous AI ventures spanning sectors such as AI large language models, cloud computing, and AI-driven drug research and development. The corporation plans to further ramp up its investments in the AI industry down the road.

John Lee, the Chief Executive of the Hong Kong Special Administrative Region (HKSAR), articulated in his most recent Policy Address that Hong Kong is set to capitalize on the unique benefits of the "One Country, Two Systems" framework. This strategy will facilitate the secure and compliant cross-border transfer of data from the mainland to the Hetao Hong Kong Data Park. The overarching goal of this initiative is to bolster scientific research endeavors and foster innovation and testing in the realm of artificial intelligence (AI) applications.

Previously, Cyberport had already taken proactive steps by establishing an AI supercomputing center. Additionally, the government rolled out the "Artificial Intelligence Funding Scheme" to further support the sector. Looking ahead, within the current year, Hong Kong plans to allocate approximately 10 hectares of land in Sha Ling, situated in the North District, for the development of a data park. This land will be made available through a market tendering process, with the objective of equipping the park with state-of-the-art computing facilities. Such a move is anticipated to significantly propel the growth of the data and AI industries in the region.

On the night of September 16, 2025, Jack Ma turned up at the HHB Bar, which is situated right next to Alibaba's campus in Hangzhou. He was accompanied by Alibaba executive Shao Xiaofeng. Clad in a dark - colored T - shirt, white shorts, and a baseball cap, Jack Ma engaged with the crowd, exchanging pleasantries and waving to greet everyone.

Insiders reveal that Jack Ma has been a regular sight at Alibaba's campus lately. His direct involvement in business affairs has hit a peak not seen in nearly five years. He has developed a keen interest in the growth of the AI business. In fact, on one particular day, he asked the management team about the business progress no less than three times.