China AI Native Industry Insights – 20241125 – IDC | Alibaba | AntChain | more
Images Generated by FLUX from EchoMimicV2 Project.
Explore Today’s China AI Native Industry Insights: the booming generative AI software market in China projected to hit $3.54 billion in the future, Alibaba’s international launch of Marco-o1 for open-ended question reasoning, and Ant Group AntChain’s groundbreaking win with Zhejiang University in NeurIPS 2024’s large model privacy challenge, tracking China’s latest AI innovation.
1. IDC: The market size of generative AI software in China will reach $3.54 billion in the future.
According to IDC, the generative AI software market in China is projected to reach $3.54 billion in the future. The report emphasizes the need for unified AI development platforms to manage data, models, and applications effectively as businesses expand their generative AI applications. With ongoing advancements in foundational models and innovative applications, the market is expected to evolve rapidly, despite challenges related to talent shortages and regulatory risks.
Original Chinese article: https://www.aibase.com/zh/news/13459
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://www.aibase.com/zh/news/13459
2. The international version of Alibaba o1 is here, Marco-o1: Focused on open-ended question reasoning.
On November 22, Alibaba’s MarcoPolo team unveiled Marco-o1, a large reasoning model designed to tackle open-ended questions—a significant challenge for AI due to the absence of standard answers. With innovative integrations of techniques like Chain of Thought (CoT) fine-tuning and Monte Carlo Tree Search (MCTS), Marco-o1 demonstrates enhanced problem-solving capabilities across diverse domains. The results show notable improvements in reasoning accuracy, particularly excelling in translating idiomatic expressions.
Original Chinese article: https://www.jiqizhixin.com/articles/2024-11-23-3
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://www.jiqizhixin.com/articles/2024-11-23-3
3. Ant Group AntChain, in collaboration with Zhejiang University Laboratory, won the championship in the large model privacy challenge at NeurIPS 2024.
Ant Group AntChain, in collaboration with Zhejiang University, has achieved remarkable success at NeurIPS 2024, winning the championship in both the attack and best practical defense tracks of the Large Language Model Privacy Challenge. This competition, dedicated to safeguarding the privacy of training data in AI systems, saw the team design innovative solutions that not only excelled in performance but also advanced the industry’s commitment to developing secure AI technologies.
Original Chinese article: https://www.qbitai.com/2024/11/223548.html
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://www.qbitai.com/2024/11/223548.html
4. Text, images, and point cloud inputs of any modality can be easily transformed into high-quality CAD models by AI with just one click.
A groundbreaking collaborative project between ShanghaiTech University, the University of Hong Kong, and Transcengram has introduced CAD-MLLM, the first multimodal large model supporting text, image, and point cloud inputs for computer-aided design (CAD). This innovation aims to simplify CAD processes, making them more accessible to non-professionals while enhancing efficiency for experienced users. The team has also created the Omni-CAD dataset, containing over 450,000 entries, to empower this transformative technology.
Original Chinese article: https://www.jiqizhixin.com/articles/2024-11-25-4
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://www.jiqizhixin.com/articles/2024-11-25-4
5. Alibaba releases open-source AI digital human project EchoMimicV2, generating half-body digital human animations from a single image.
Alibaba’s Ant Group has launched EchoMimicV2, an advanced open-source digital human project that creates full-body animations driven by audio and hand gestures. Building on its predecessor, this new version enhances the synchronization of audiovisual elements, enabling seamless transitions between Mandarin and English voice inputs and corresponding actions. This innovative tool is poised to revolutionize fields like virtual broadcasting, online education, and customer service.
Original Chinese article: https://ai-bot.cn/echomimicv2/
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://ai-bot.cn/echomimicv2/
6. AI mimics humans in reading comics, achieving new SOTA in temporal localization ability for video large models.
Researchers have introduced a groundbreaking method called NumPro, which enhances the temporal localization capabilities of video language models without requiring extensive training. By employing unique numerical identifiers similar to comic book panels, this innovative technique significantly improves event timeline comprehension in videos while maintaining the models’ overall understanding abilities. Notable advancements were demonstrated across various benchmarks, positioning NumPro as a frontrunner in video temporal grounding technology.
Original Chinese article: https://36kr.com/p/3051140129164166
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://36kr.com/p/3051140129164166
7. 20 complex Excel operations solved in one sentence! Peking University ChatExcel has been newly upgraded and is now open for free use by everyone!
The recently upgraded ChatExcel by a team from Peking University revolutionizes Excel operations, allowing users to conduct complex linear analysis and generate charts with a single sentence command. This user-friendly tool supports multiple spreadsheet uploads, offers automated report generation, and can execute advanced data comparisons, making it an essential asset for students and professionals alike. Currently available for free trial, it aims to enhance productivity in data analysis tasks.
Original Chinese article: https://www.qbitai.com/2024/11/223180.html
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://www.qbitai.com/2024/11/223180.html
8. Kling AI Platform 1.5 Model Update: Launch of Face Model Feature with New “Standard Mode”
Kuaishou’s Kling AI platform has upgraded to version 1.5, introducing advanced features including a new facial model function and a “Standard Mode” for generating 720p videos quickly and affordably. Users can now utilize enhanced control over motion tracking and camera techniques, increasing efficiency for video creation. The upgrade also offers a limited-time 50% discount for diamond and platinum members, making this pioneering video modeling capability more accessible.
Original Chinese article: https://www.aibase.com/zh/news/13431
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https://www.aibase.com/zh/news/13431
That’s all for today’s China AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.