20241104 – Huawei | Alibaba | ChinaTelecom | ZhipuAI | more
1. West China Hospital of Sichuan University, in collaboration with Huawei Data Storage, has launched the “West China Medical” large medical model, integrating over 10 general models and more than 50 specialized models.
West China Hospital of Sichuan University, in collaboration with Huawei, has launched the “Huangxi Hongyi” medical model, integrating over 10 general models and more than 50 specialized models. This comprehensive system is built on 720 billion parameters and incorporates a vast array of data, enhancing the development and deployment of medical AI applications. The initiative aims to significantly streamline processes in healthcare, improving efficiency and accuracy across various medical domains.
Original Chinese article: https://www.ithome.com/0/807/563.htm
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.ithome.com%2F0%2F807%2F563.htm
2. Can diffusion models also “think critically”? Alibaba’s IC-LoRA enhances the episodic memory capability of image generation models.
Alibaba’s Tongyi Lab has unveiled groundbreaking research on the Diffusion Transformer model, showcasing its ability to generate cohesive image sets with minimal training data using the innovative IC-LoRA technique. This advancement transforms the model from a data-hungry learner to a highly efficient creator, significantly reducing training costs while enhancing its contextual learning capabilities. With IC-LoRA, AI-driven image generation is now more accessible, paving the way for widespread artistic collaboration.
Original Chinese article: https://www.aibase.com/zh/news/12965
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.aibase.com%2Fzh%2Fnews%2F12965
3. Support for bilingual Chinese and English as well as the ability to mix any of the 40 dialects, China Telecom’s TeleSpeech-ASR voice model has been upgraded.
China Telecom’s TeleSpeech-ASR has successfully upgraded its Starry Voice Model to support bilingual (Chinese and English) capabilities and recognize an impressive 40 dialects. This enhancement is a significant leap from the initial 30 dialects, utilizing a unique pre-training and fine-tuning approach that reduces the need for extensive labeled data while maintaining high accuracy. This innovative model aims to meet the diverse linguistic needs of users, positioning TeleAI at the forefront of AI-driven voice recognition technology.
Original Chinese article: https://www.ithome.com/0/807/525.htm
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.ithome.com%2F0%2F807%2F525.htm
4. Beijing Artificial Intelligence Industry Investment Fund and others invest in Zhipu AI, increasing Zhipu AI’s registered capital to 31.07 million.
Beijing’s artificial intelligence industry investment fund has acquired a stake in Zhipu AI, raising the company’s registered capital from approximately 27.91 million to about 31.07 million RMB. This investment marks Zhipu AI as the first AI model company invested in by the fund since its inception, reflecting the growing significance of the AI sector in Beijing’s tech landscape.
Original Chinese article: https://www.163.com/dy/article/JG575CTP0519QIKK.html
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.163.com%2Fdy%2Farticle%2FJG575CTP0519QIKK.html
5. Multi-image scene alignment with DPO, new methods proposed by Shanghai AI Lab, no manual labeling required.
Researchers from Shanghai Jiao Tong University and Shanghai AI Laboratory have introduced a novel method, MIA-DPO, for aligning multi-image scenarios without manual annotation. This approach significantly reduces data collection costs while enhancing the performance of large visual language models in complex image tasks. By leveraging a preference alignment technique, MIA-DPO also minimizes the occurrence of perception errors, demonstrating a promising advancement in the field of AI image processing.
Original Chinese article: https://36kr.com/p/3017632680699397
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2F36kr.com%2Fp%2F3017632680699397
6. The Doubao Large Model Team has open-sourced the RLHF framework, achieving a training throughput improvement of up to 20 times.
The Doubao AI team has launched an open-source RLHF framework, HybridFlow, enabling a significant increase in training throughput, with improvements ranging from 1.5 to 20 times compared to other frameworks. This innovative approach combines flexibility and efficiency by decoupling control and computation flows, thus enhancing the performance of reinforcement learning in large models. With strong support for varied distributed frameworks, HybridFlow aims to streamline AI development and deployment processes.
Original Chinese article: https://www.qbitai.com/2024/11/215093.html
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.qbitai.com%2F2024%2F11%2F215093.html
7. Focused on AI smart customer service based on large models, Xunlu AI has completed a strategic financing of 60 million yuan.
Xunlu AI, a cutting-edge intelligent customer service company based on large language models, has successfully secured 60 million RMB in strategic financing. This funding, led by IDG and Shanda Capital, will be primarily allocated to technology development and talent acquisition, solidifying the company’s position in providing tailored AI solutions across various industries, including e-commerce and finance. As the demand for online customer service continues to surge, Reindeer AI aims to revolutionize the market by enhancing efficiency and significantly reducing operational costs.
Original Chinese article: https://36kr.com/p/3003323290656774
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2F36kr.com%2Fp%2F3003323290656774
8. Medical robotics company Amitmed has received tens of millions in Series A funding, with Qianhai Ark Health Fund stepping in.
Beijing-based medical robotics company Amitmed has secured several million yuan in A-round financing from the Qianhai Ark Health Fund. Founded in 2021, the firm specializes in robotic systems for minimally invasive cancer treatments, notably developing the Jupiter intelligent soft tissue puncture navigation robot that integrates real-time ultrasound and pre-operative CT imaging for precise tumor targeting, aiming to enhance patient outcomes and reduce treatment costs.
Original Chinese article: https://www.163.com/dy/article/JFU6G0BB05198R3E.html
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.163.com%2Fdy%2Farticle%2FJFU6G0BB05198R3E.html
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.