Latest news with #BAAI
Yahoo
09-06-2025
- Business
- Yahoo
China's AI lab unveils RoboBrain 2.0 model to accelerate humanoid robot development
In a move that will further assert China's bid to scale robotics industry, the Beijing Academy of Artificial Intelligence (BAAI)—a not for profit research laboratory—unveiled last week a series of new open-source artificial intelligence (AI) models, dubbed RoboBrain 2.0, that will function as the 'brain' of robots. According to BAAI head Wang Zhongyuan, the use of powerful AI models in China's booming robotics market could accelerate the development and adoption of humanoids, as the sector works to overcome key challenges such as limited model capabilities and a lack of high-quality training data. Wang further explained that BAAI is actively seeking collaboration across the embodied intelligence industry, emphasizing the importance of joint efforts to accelerate progress. He noted that the institute is working with more than 20 leading companies in the sector and is looking to expand its network of partners to drive continued growth. Unveiled as part of China's broader push to advance intelligent machines, RoboBrain 2.0 was described by Wang as the world's most powerful open-source AI model designed to enhance a wide range of robots, including humanoids. Its debut positions BAAI as a potential key player in the evolving sector, the South China Morning Post reported. Furthermore, RoboBrain 2.0 introduces major improvements in spatial intelligence and task planning, delivering a 17% boost in speed and a 74% increase in accuracy compared to the previous version launched just three months earlier. With enhanced spatial intelligence, robots can now perceive distances from surrounding objects more precisely, while advanced task planning enables them to autonomously deconstruct complex activities into manageable steps, significantly improving overall performance. The RoboBrain model is part of the Wujie series, which also includes RoboOS 2.0—a cloud platform for deploying robotics AI models—and Emu3, a multimodal system capable of interpreting and generating text, images, and video. BAAI is one of China's early developers of open-source large language models, the technology behind generative AI chatbots. Several former employees have used their experience at BAAI to start their own AI companies, helping to grow the AI startup community in China. China's push to lead in robotics AI involves multiple players, with BAAI joined by the Beijing Humanoid Robot Innovation Centre, which earlier this year launched Hui Si Kai Wu—a general-purpose embodied AI platform. The center is also known for developing the Tien Kung humanoid robot, which made headlines after completing a half-marathon in Beijing in April. The Chinese institution aims to have its platform become the "Android of humanoid robots," serving as a standard operating system much like Google's Android does in the smartphone industry. Moreover, this year's edition of the BAAI Conference attracted over 100 AI researchers from around the world and more than 200 industry experts, including leaders from major Chinese tech companies such as Baidu, Huawei Technologies, and Tencent Holdings. Additionally, the Chinese academy also announced a strategic partnership with the Hong Kong Investment Corporation to collaborate on talent development, technology advancement, and capital investment aimed at fostering innovation and entrepreneurship in the country's AI sector.
Express Tribune
07-06-2025
- Business
- Express Tribune
Beijing launches AI ‘RoboBrain' to power next generation of humanoid robots
Humanoid robot "Tiangong" participates along with human runners in the E-Town Half Marathon & Humanoid Robot Half Marathon in Beijing, China April, 19 2025. PHOTO: REUTERS Listen to article China has taken a significant step forward in its race to lead global robotics innovation with the unveiling of RoboBrain 2.0 — an open-source artificial intelligence model designed to serve as the cognitive core of humanoid robots. The model was launched on Friday by the Beijing Academy of Artificial Intelligence (BAAI), a leading non-profit research institution, during its annual conference in the Chinese capital. The release marks a milestone in China's efforts to build smarter and more autonomous machines for its fast-growing robotics sector. Described by BAAI director Wang Zhongyuan as the world's most powerful open-source AI model tailored for robotics, RoboBrain 2.0 promises to enhance a robot's spatial perception and task-planning capabilities. Compared to its earlier version introduced just three months ago, the upgraded model is reported to operate 17% faster and with 74% higher accuracy. 'Currently, we are partnering with over 20 leading companies and are open to new collaborations to drive growth in the embodied intelligence industry,' Wang told attendees at the Zhiyuan Institute — BAAI's local moniker. The enhanced spatial intelligence enables robots to more accurately interpret their surroundings and judge distances, while improved planning algorithms allow them to autonomously deconstruct complex tasks into simpler, executable steps. RoboBrain 2.0 is part of BAAI's broader Wujie model suite, which also includes RoboOS 2.0 — a cloud-based distribution platform for robotics AI — and Emu3, a multimodal system capable of processing and generating text, images, and video. The model's launch comes amid intensifying competition in China's robotics landscape. Earlier this year, the Beijing Humanoid Robot Innovation Centre introduced its own general-purpose platform, Hui Si Kai Wu, which it envisions as the Android equivalent for humanoid robots. The centre made headlines in April after its Tien Kung humanoid robot completed a half-marathon in Beijing. BAAI has emerged as a pioneer in open-source large language models and has played a key role in China's generative AI boom. Many of its alumni have gone on to found successful AI start-ups. Despite being added to the US Entity List in March — which restricts access to American technology — the institute continues to push forward, with Wang calling the decision 'a mistake' and lobbying for its reversal. In a show of resilience and ambition, BAAI also announced a new strategic partnership with the Hong Kong Investment Corporation to promote innovation through joint initiatives in talent development, technology exchange, and venture capital. This year's BAAI Conference drew more than 100 global AI researchers and over 200 industry experts, including leaders from major Chinese tech firms such as Baidu, Huawei, and Tencent, alongside robotics-focused start-ups like Unitree Robotics, Zhipu AI, and Shengshu AI. As China accelerates its quest to develop commercially viable humanoid robots, BAAI's open-source approach could help democratise access to high-performing AI systems — and set the foundation for a new generation of intelligent machines.
South China Morning Post
06-06-2025
- Business
- South China Morning Post
Beijing academy unveils open-source ‘RoboBrain' AI model for China's humanoid robots
The Beijing Academy of Artificial Intelligence (BAAI), a non-profit research laboratory in China, launched on Friday a series of new open-source artificial intelligence (AI) models designed to be the 'brain' of robots, as the country rushes to build smarter machines. Advertisement The use of powerful AI models in China's booming robotics industry could accelerate the development and adoption of humanoids, as the sector addresses challenges such as limited model capabilities and a lack of training data, according to BAAI head Wang Zhongyuan during the institute's annual conference in Beijing. Wang described BAAI's RoboBrain 2.0 as the world's most powerful open-source AI model designed to improve various types of robots, including humanoids. The launch of this general-purpose AI model coincides with the Chinese robotics industry's rapid growth, positioning BAAI as a potential major player in the local sector. Beijing Academy of Artificial Intelligence director Wang Zhongyuan speaks at the institute's annual conference on Friday. Photo: Handout 'We sincerely hope that various stakeholders in the embodied intelligence industry will collaborate with the Zhiyuan Institute,' Wang said, referring to the local name for BAAI. Advertisement 'Currently, we are partnering with over 20 leading companies in the sector and are looking for additional collaborators to drive growth.'
South China Morning Post
06-06-2025
- Business
- South China Morning Post
Beijing academy unveils open-source ‘RoboBrain' AI model for China's humanoid robots
The Beijing Academy of Artificial Intelligence (BAAI), a non-profit research laboratory in China, launched on Friday a series of new open-source artificial intelligence (AI) models designed to be the 'brain' of robots, as the country rushes to build smarter machines. The use of powerful AI models in China's booming robotics industry could accelerate the development and adoption of humanoids, as the sector addresses challenges such as limited model capabilities and a lack of training data, according to BAAI head Wang Zhongyuan during the institute's annual conference in Beijing. Wang described BAAI's RoboBrain 2.0 as the world's most powerful open-source AI model designed to improve various types of robots, including humanoids. The launch of this general-purpose AI model coincides with the Chinese robotics industry's rapid growth, positioning BAAI as a potential major player in the local sector. Beijing Academy of Artificial Intelligence director Wang Zhongyuan speaks at the institute's annual conference on Friday. Photo: Handout 'We sincerely hope that various stakeholders in the embodied intelligence industry will collaborate with the Zhiyuan Institute,' Wang said, referring to the local name for BAAI. 'Currently, we are partnering with over 20 leading companies in the sector and are looking for additional collaborators to drive growth.' According to Wang, RoboBrain 2.0 features significant upgrades in spatial intelligence and task planning, achieving 17 per cent faster performance and 74 per cent greater accuracy compared to its predecessor, which was introduced three months ago.
Yahoo
15-05-2025
- Business
- Yahoo
GOSIM AI Paris 2025 Concludes with Resounding Success! 80+ Global Luminaries Delve into Model Frontiers, Tackle Computing Challenges, and Chart Implementation Pathways
HANGZHOU, China, May 15, 2025 /PRNewswire/ -- Fueled by the open-source wave, the AI field is experiencing explosive global growth, forging new pathways for developers, researchers, and entrepreneurs to construct next-generation intelligent systems. Against this dynamic backdrop, GOSIM AI Paris 2025 took center stage in Paris, France, on May 6–7. Co-hosted by GOSIM (Global Open-Source Innovation Meetup), CSDN (Chinese Software Developer Network), and the conference united over 970 technical experts, open-source contributors, AI practitioners, and industry professionals from 28 countries and regions for two days of intensive dialogue and intellectual synergy. A testament to the surging global momentum in AI open source, the conference boasted a diverse and illustrious lineup of speakers and projects. The agenda featured 2 Keynotes and 6 thematic forums, with 89 distinguished guest speakers from world-renowned institutions such as Meta, NVIDIA, Alibaba Group Holding Limited, Hugging Face, Beijing Academy of Artificial Intelligence (BAAI), MiniMax, Neo4j, MetaGPT, Zhipu AI (Beijing Zhipu Huazhang Technology), Eigen Technologies, Docker, InfiniFlow, Peking University (PKU), Fraunhofer Institute (Germany), University of Oxford, and the OpenLLM France community. Together, they delivered 82 compelling technical deep dives focused on pivotal topics including AI models, infrastructure, applications, embodied intelligence, PyTorch, and OpenHarmony. Meanwhile, several groundbreaking achievements were unveiled for the first time during the conference. Notably, Beijing Academy of Artificial Intelligence (BAAI) released its high-quality bilingual dataset, CCI4.0, poised to significantly boost global AI open-source innovation. Additionally, 12 exceptional innovative open-source projects, selected for the Spotlight segment, showcased the latest strides by global developers in AI innovation practices and cutting-edge technological exploration. While championing the fusion of open source and technology, the conference also placed immense value on cultivating a robust community ecosystem and fostering deep developer engagement. Interactive hubs like the PyTorch Day France track, the 2025 European OpenHarmony Technical Forum track, and the Seeed Embodied AI Workshop hands-on experience zone catered to diverse needs—from technical sharing and practical operations to immersive learning—creating a rich, high-impact exchange experience for frontline tech professionals and learners alike. This edition of GOSIM AI Paris 2025 also drew robust engagement from multiple prestigious global technical communities and organizations. These included heavyweight partners such as Huawei Technologies Co., Ltd., Association des Jeunes Chinois pour la Technologie et l'Innovation Scientifique en France (AJTIS), Association d'Intelligence Artificielle France-Chine (AIFC), The Apache Software Foundation (ASF), Eclipse Foundation, The Khronos Group, WasmEdge (Runtime), LF AI & Data Generative AI Commons, Linux Foundation Research, OpenWallet Foundation (OWF), Open Source Initiative (OSI), Software Heritage, and K8SUG (Kubernetes & AI User Group). Their collective contributions were instrumental in staging this landmark international technology summit, celebrating the synergy of AI and open source, and the convergence of academia and industry. Global Open Source Collaboration Shapes the Future, AI Open Source Evolution Reaches New Depths Michael Yuan, Co-founder of GOSIM, underscored in his opening Keynote: "Future AGI should not be monopolized by the company with the deepest pockets but should be developed through global collaboration, co-building an ecosystem network covering models, knowledge bases, robotics, and execution systems." Daniel Goldscheider, Executive Director of the OpenWallet Foundation (OWF), shared his insights from promoting global digital collaboration: the strategy involves first uniting diverse stakeholders to forge consensus, paving the way for specialized bodies to then spearhead the development of specific standards and technical frameworks. Zhipeng Huang, a board member of the LF AI & Data Foundation, systematically delineated six pivotal development trends in the contemporary large model landscape demanding attention. These encompass unified multimodal architectures, the evolution of efficient attention mechanisms, second-order optimization maturing into practical applications, the paradigm shift from AI4S to Equivariance4ALL, the construction of "data oil wells," and the post-training stage emerging as a critical juncture. He firmly avowed that as large models continue their evolutionary trajectory, only by embracing open source, fortifying collaboration, and championing standards development can a truly sustainable technological ecosystem be forged. To unlock AI's full potential, Matt White, Executive Director of the PyTorch Foundation and General Manager of AI at the Linux Foundation, offered several key recommendations: select appropriate open-source licenses for AI tools and models; actively engage in global collaboration and resist the protectionist undercurrents of "AI nationalism"; oppose excessive regulation of open source to prevent the stifling of innovation; support neutral, reliable open standards; and consistently build trustworthy AI with responsibility, ensuring ethical and engineering integrity from conception through to product deployment. The pace of large model iteration is staggering. In the "AI Models" forum, experts from leading open-source communities and enterprises—including Hugging Face, Alibaba Group Holding Limited, Beijing Academy of Artificial Intelligence (BAAI), CSDN (Chinese Software Developer Network), MiniMax (company), University of Oxford, Zhipu AI (Beijing Zhipu Huazhang Technology), Fraunhofer Institute, LAION (Large-scale Artificial Intelligence Open Network), LINAGORA - OpenLLM France, Northeastern University, Department of Electrical and Computer Engineering, and OpenBMB (Open Lab for Big Model Base) & Modelbest—delivered captivating presentations on architectural innovation, open-source synergy, and the evolving ecosystem of large models. Dedicated to forging a more open, efficient, and inclusive foundation for large models, the "AI Infrastructure" forum convened frontline experts from NVIDIA, Huawei Technologies Co., Ltd., Peking University (PKU), Beijing Academy of Artificial Intelligence (BAAI), Docker, Universidad Politécnica de Madrid (Technical University of Madrid), Beihang University (BUAA), Runhouse Inc., InfiniFlow, Dimforge, Carrefour, and They engaged in profound discussions on critical issues spanning data, computing power, and system architecture. A central theme of the conference, the "AI Applications" forum saw practitioners and technical leaders from several flagship open-source projects share their invaluable insights and frontline experiences. They illuminated tangible implementation pathways and the expansive future of AI applications powered by large models. The "Embodied Intelligence" theme brought together a multitude of experts and scholars from the vanguard of industry and research. They delved into core subjects like humanoid robotics, data acquisition, and affective computing, sharing practical insights from their work on architectural design, model application, and real-world scenario implementation. In the electrifying PyTorch Day France track, the PyTorch Foundation and its community partners showcased PyTorch's practical triumphs in research and industry, highlighting advancements in underlying toolchains, deployment optimization, and multimodal support. As one of today's foremost deep learning frameworks, PyTorch continually pushes boundaries with its inherent flexibility and scalability, thereby fostering greater autonomy and control in AI development for nations and communities worldwide. Concurrently, the conference proudly hosted the 2025 European OpenHarmony Technical Forum, themed "OpenHarmony Connecting the Future: OS Innovation and Ecosystem Evolution Empowered by AI." This forum spotlighted the global technological research advancements and ecosystem-building endeavors of the OpenHarmony project, deeply exploring its AI-empowered integrated development trajectory, and presenting groundbreaking achievements and future horizons in the operating system domain. A Dedicated Stage for Developers: Spotlight and Workshops Shine Bright GOSIM AI Paris 2025 was more than just a repository of knowledge; it was an electrifying, immersive interactive experience. Beyond the insightful keynote presentations, the conference meticulously curated a rich array of Workshops and Spotlight activities, inspiring every attendee to not merely listen, but to actively "get hands-on" and "lend their voice." In the two-day Seeed Embodied AI Workshop, co-hosted by Hugging Face and Seeed Studio, participants dove headfirst into the hands-on realm of embodied intelligence. They meticulously assembled, debugged, and demonstrated a variety of robot perception and control projects live on site. Their ingenious creations drew enthusiastic crowds, garnering applause and accolades, vividly demonstrating the compelling synergy and boundless potential of AI fused with open-source hardware. In contrast to the thematically structured main forums, the Spotlight Talks—a demo showcase of outstanding innovative projects recruited and selected from global submissions—delivered a rapid-fire exhibition of technical ingenuity. Project teams unveiled standout features and novel methodologies, featuring in-depth explorations of trending topics such as green AI, robot lifelong learning, and RAG agent frameworks, spanning both academic research and industrial application. These showcases vividly illustrated the power of open source to accelerate prototyping and broad dissemination, underscoring the developer community's dynamic innovative spirit. Immersive Deep Connections and Surprises: People to People, People to Technology Simultaneously, the conference fostered a rich environment for multicultural exchange. Two pivotal roundtable discussions—the "Open Source AI Strategy Forum" and the "GOSIM Innovation Ecosystem Exchange Meeting"—became crucibles for profound deliberations on global cooperation and collaborative ecosystem development. Key issues encompassing technology, market dynamics, and talent cultivation were meticulously analyzed and re-envisioned. Ideas sparked in vibrant dialogue, and collaborations took root in shared understanding, propelling long-term international technological partnerships forward. Beyond the formal discussions, two specially curated evening events amplified the networking experience: an enchanting Seine River night cruise for AI luminaries, and a convivial gathering for speakers at the Station F café. These occasions allowed attendees to dream, discuss, and connect freely against the magical backdrop of nocturnal Paris, nurturing profound cross-cultural and cross-disciplinary technical and industrial exchanges. Adding to the intellectual depth were the "Frank Talks." Throughout the conference, the Open AGI Forum segment hosted 12 live, engaging interview dialogues. These featured open-source pioneers from China, the US, and Europe, including representatives from Hugging Face, Zhipu AI (Beijing Zhipu Huazhang Technology), the Chinese Academy of Social Sciences, MiniMax (company), Alibaba Tongyi Lab, University of Oxford, LAION (Large-scale Artificial Intelligence Open Network), and CAMEL-AI, with participants hailing from nations such as France, the United States, China, the United Kingdom, Germany, Scotland, and Portugal. Tech visionaries eloquently articulated the value of open source, and the formidable challenges and exhilarating possibilities of AGI, blending practical insights with strategic foresight. Candid "Voice of Attendees" flash interviews ran throughout the event, capturing authentic, firsthand impressions and genuine experiences. Twenty-nine attendees from diverse nations—ranging from "teenage developers" to a 70-year-old "tech veteran," and encompassing speakers, audience members, exhibitors, and volunteers—shared their candid reflections on the conference's content, the promise of AI applications, and the vibrant open-source atmosphere of Paris. And the fun factor? Off the charts! The exhibition area buzzed with interactive experiences, including T-shirt heat transfers, captivating VR demonstrations, and engaging robot interactions, quickly becoming the liveliest and most sought-after check-in destinations. Attendees got hands-on and minds-on, with this direct participation dramatically enhancing immersion and transforming technical exchange from abstract theory into tangible reality. Adding to the excitement were delightful surprises and a touch of "good luck." The conference orchestrated a special "Show GOSIM Moment" surprise draw, with 2 coveted special prizes for the GOSIM HANGZHOU 2025 conference and 40 lucky draw prizes sparking considerable anticipation and delight among attendees. Farewell Paris, See You at the Next Stop in Hangzhou, Embarking on a New Journey of Global Open Source Collaboration! And so, GOSIM AI Paris 2025 drew to a triumphant close. Reflecting on this remarkable journey—from Shanghai to Delft, then Beijing, and now Paris—GOSIM has consistently united developers, researchers, and innovators from every corner of the globe, forging invaluable bridges across languages and cultures. Our commitment to the profound application of open-source technology remains steadfast, as we relentlessly explore new horizons for collaboration and innovation. Yet, these vibrant cross-border conversations are far from over. Our next destination: Hangzhou. We'll carry forward the passion and inspiration ignited in Paris, eagerly anticipating our reunion with you in September to collaboratively script a new, exhilarating chapter in the open-source saga. GOSIM HANGZHOU 2025 official website: View original content to download multimedia: SOURCE CSDN Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
