20241022 – Alibaba | ByteDance | BAAI | Tsinghua | more
1. All tables and charts are captured! Alibaba Damo Academy’s open-source DocOwl 1.5 efficiently “understands” documents without OCR!
Alibaba’s Damo Academy, in collaboration with Renmin University of China, has launched an open-source document processing model called mPLUG-DocOwl1.5. This groundbreaking model excels in understanding document content without the need for OCR, achieving state-of-the-art performance across multiple visual document comprehension benchmarks. mPLUG-DocOwl1.5 emphasizes the importance of structural information through “Unified Structure Learning” to enhance multimodal large language models, making significant advancements in processing complex document formats.
Original Chinese article: https://www.aibase.com/zh/news/12566
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%2F12566
2. OCR-Omni is here, ByteDance & East China Normal University unify multimodal text understanding and generation.
ByteDance and East China Normal University have introduced TextHarmony, a groundbreaking multimodal model that integrates visual text understanding and generation within a unified framework. Selected for NeurIPS 2024, TextHarmony addresses the integration challenges faced in traditional OCR tasks, showcasing exceptional performance in visual text perception, understanding, generation, and editing. As a significant advancement in the OCR field, it paves the way for enhanced AI applications in document processing, content creation, and education.
Original Chinese article: https://mp.weixin.qq.com/s/PLF0dc1b-W5a59sK7XX0bA
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2FPLF0dc1b-W5a59sK7XX0bA
3. Beijing Academy of Artificial Intelligence releases the native multimodal world model Emu3, achieving unification of images, text, and videos.
On October 21, 2024, the Beijing Academy of Artificial Intelligence unveiled the Emu3, a groundbreaking native multimodal world model that seamlessly integrates text, image, and video processing without relying on diffusion models or complex combinations. Boasting superior performance over well-known models like SDXL and LLaVA, Emu3 introduces a unified architecture for multimodal AI, allowing for efficient training and deployment across diverse applications while offering an innovative pathway toward building more advanced AI systems.
Original Chinese article: https://mp.weixin.qq.com/s/b4emisljL0bXvBfSyDcAbw
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2Fb4emisljL0bXvBfSyDcAbw
4. The world’s first! A Chinese team develops an AI electronic paramagnetic resonance spectrometer.
On October 19, at the 2024 National Academic Symposium on Electron Paramagnetic Resonance Spectroscopy, CIQTEK unveiled its groundbreaking AI Electron Paramagnetic Resonance Spectrometer (AI-EPR). This innovative instrument boasts an unprecedented signal-to-noise ratio of 10,000:1, setting a global benchmark while integrating AI capabilities for enhanced spectral analysis and literature connectivity, significantly improving research efficiency and precision across various scientific fields.
Original Chinese article: https://news.qq.com/rain/a/20241020A01LQ400
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fnews.qq.com%2Frain%2Fa%2F20241020A01LQ400
5. Tsinghua Open Source Mixed Precision Inference System MixQ: Near Lossless Quantization of Large Models and Improved Inference Throughput.
Tsinghua University’s PACMAN Lab has unveiled MixQ, an open-source mixed-precision inference system that achieves near-lossless quantization for large models while significantly boosting inference throughput. By utilizing INT8 and INT4 precision, MixQ allows seamless deployment and has already been applied in real-world products by leading AI companies. This innovative system enhances overall performance, achieving up to six times faster throughput compared to existing solutions.
Original Chinese article: https://www.163.com/dy/article/JF1E5PNO0511DSSR.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%2FJF1E5PNO0511DSSR.html
6. Qualcomm launches Snapdragon 8 Gen 2 processor: collaborates with Zhipu Technology and Tencent Hunyuan for on-device AI.
At the Snapdragon Summit 2024, Qualcomm unveiled the highly anticipated Snapdragon 8 Gen 2 mobile platform, built on TSMC’s advanced 3nm process technology, setting new benchmarks for Android chip performance. This innovative platform features an all-big core design, significantly boosting CPU performance by 45% while reducing power consumption by 44%. Alongside its powerful capabilities, Qualcomm announced partnerships with Zhipu and Tencent to enhance AI applications on this next-gen chip.
Original Chinese article: https://www.aibase.com/zh/news/12594
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%2F12594
7. China Mobile Shanghai Industrial Research Institute: The penetration rate of AI large models in the financial sector exceeds 50%, the highest in the industry.
According to a recent report from the China Mobile Shanghai Industrial Research Institute, over 50% of AI model penetration in the financial sector marks it as the industry leader in adoption rates. Popular applications include smart sales, intelligent Q&A, and risk management, all demonstrating high maturity levels. However, experts warn of potential risks associated with AI, including opacity, unpredictability, and cybersecurity vulnerabilities, underscoring the complexity of integrating such technologies in finance.
Original Chinese article: https://www.163.com/dy/article/JF0TPQKA0511B8LM.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%2FJF0TPQKA0511B8LM.html
8. Zhou Hongyi’s Nankai Speech: Specialized Large Models are the Key Focus for Enterprises to Embrace AI.
During a recent speech at Nankai University, Zhou Hongyi, founder of 360 Group, emphasized the importance of specialized AI models for businesses to effectively embrace artificial intelligence. He noted the validated trends in AI development, highlighting the shift toward smaller, enterprise-focused models as crucial for enhancing productivity and operational efficiency. Zhou also underscored the necessity of integrating AI with business systems to realize its full potential, drawing parallels between AI’s evolution and the historical adoption of electric motors in industry.
Original Chinese article: https://www.pingwest.com/w/299230
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fwww.pingwest.com%2Fw%2F299230
9. Kling AI launches the first phase of the Future Partner Program, introducing a one-stop AIGC ecosystem cooperation platform.
Kuaishou’s Kling AI has launched its inaugural “Future Partner Program,” introducing a comprehensive AIGC collaboration platform powered by advanced proprietary models. This innovative platform connects demand-driven clients with creative professionals, unlocking new opportunities in AIGC creativity and commercial viability. With over 2.6 million users and millions of generated assets, Kling AI is set to revolutionize content creation and monetization for creators while fostering a thriving ecosystem for collaboration.
Original Chinese article: http://jjckb.xinhuanet.com/20241021/de5aa05782054e9398fdfc0465bfe473/c.html
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=http%3A%2F%2Fjjckb.xinhuanet.com%2F20241021%2Fde5aa05782054e9398fdfc0465bfe473%2Fc.html
10. Deeply cultivating AI large model applications in the telecommunications industry, the national specialized small giant “NetThink Technology” announced the completion of a hundred million yuan A+ round of financing.
NetThink Technology, a national specialized small giant in the telecommunications sector, has successfully completed a significant A+ funding round exceeding 100 million yuan, backed exclusively by Paradigm Ventures. This funding will enhance the company’s R&D efforts and expand applications of AI large model technology, solidifying its role in the digital transformation landscape and intelligent manufacturing sectors. NetThink Technology continues to drive innovation by delivering advanced solutions to major enterprises, contributing significantly to China’s digital development.
Original Chinese article: https://www.163.com/dy/article/JF3O6OCI0552QJ47.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%2FJF3O6OCI0552QJ47.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.