Beijing Academy of Artificial Intelligence (BAAI) (Chinese: 北京智源人工智能研究院; pinyin: Běijīng Zhìyuán réngōng zhìnéng yánjiùyuàn), also known as Zhiyuan Institute, is a Chinese non-profit artificial intelligence (AI) research laboratory.[2] BAAI conducts AI research and is dedicated to promoting collaboration among academia and industry,[3] as well as fostering top talent and a focus on long-term research on the fundamentals of AI technology.[4] As a collaborative hub, BAAI's founding members include leading AI companies, universities, and research institutes. BAAI is one of pre-eminent AI research institutes in China.[4][5] To help it reach its goals, BAAI frequently releases new models and open source code. Moreover, BAAI organizes an annual international conference bringing together AI experts, industry leaders, and international talent to discuss challenges and future of AI.[6]
 | |
| Industry | Artificial intelligence |
|---|---|
| Founded | November 2018 |
| Founders | |
| Headquarters | Wudaokou, Beijing, China[1] |
| Products |
|
Number of employees | c. 100 (2023) |
| Website | www |
Products and applications
editAs of 2023[update], BAAI's research focuses on large pre-trained models (LLMs) and open-source AI infrastructure.
WuDao
editWuDao (Chinese: 悟道; pinyin: wùdào)[7] is a large multimodal pre-trained language model.[8][9] WuDao 2.0, was announced on 31 May 2022[10][2] and has been compared to GPT-3 at the time. But, in comparison, GPT-3 has 175 billion parameters, while WuDao has 1.75 trillion parameters;[11] making it the largest pre-trained model in the world at the time.[10][12] WuDao was trained on 4.9 terabytes of images and texts (which included 1.2 terabytes of Chinese text and 1.2 terabytes of English text).[10][13] The chairman of BAAI said that WuDao was an attempt to "create the biggest, most powerful AI model possible";[14] although direct comparisons between models based on parameter count (i.e. between Wu Dao and GPT-3) do not directly correlate to quality.[11]
WuDao has demonstrated ability to perform natural language processing and image recognition, in addition to generation of text and images.[2] The model can not only write essays, poems and couplets in traditional Chinese, it can both generate text based on static images and generate nearly photorealistic images based on natural language descriptions. It has also showed ability to power virtual chat agents and predict the 3D structures of proteins like AlphaFold.[2]
FlagAI
editFlagAI is an open-source extensible toolkit for large-scale model training and inference. Its goal is to support training, fine-tuning, and deployment of large-scale models on various downstream tasks with multi-modality. Moreover, its open repository includes not only all source-code, but several pre-trained large models. FlagAI is an approved incubation project at the sandbox level of the Linux Foundation.[15]
Jiuding
editJiuding is an AI-computing platform which focuses on supporting AI innovation. As of September 2022 it provides 1000P computation capacity with 400 Gbit/s high-speed interconnection per server, and support AI chipsets of different architectures. BAAI's platform also includes code compilers for the different AI architectures.[6]
MetaWorm
editMetaWorm is a computational model of the Caenorhabditis elegans (C. elegans) nematode simulating the worm's nervous system along with a "digital body" simulation in real-time.[16] MetaWorm 1.0 exhibits behaviours that parallel C. elegans in the real world.[6]
See also
editReferences
edit- ^ "Beijing Academy of Artificial Intelligence". Retrieved 28 March 2023.
- ^ a b c d Tarantola, A. (2 June 2021). "China's gigantic multi-modal AI is no one-trick pony". Engadget.
- ^ "Nature Index - Institution Collaborations/Relationships". Nature. Retrieved 28 March 2023.
- ^ a b "This Chinese Lab Is Aiming for Big AI Breakthroughs". WIRED. 21 January 2021.
- ^ "The race of the AI labs heats up". The Economist. Retrieved 28 March 2023.
- ^ a b c "BAAI 2022: 'AI Life', Multimodal Models, AI for Science, Autonomous Driving and More!". Synced. 7 June 2022. Retrieved 6 April 2023.
- ^ 智源研究院 (11 January 2021). "面向认知，智源研究院联合多家单位发布超大规模新型预训练模型"悟道·文汇"".
- ^ "Record-breaking AI model WuDao 2.0 unveiled in Beijing". China.org.cn. 3 June 2021.
- ^ "Can A.I. help investors find the next hot technology? A number of companies are betting on it". Fortune.
- ^ a b c Feng, Coco (2 June 2021). "Beijing-funded AI language model tops Google and OpenAI in raw numbers". South China Morning Post.
- ^ a b Mott, Nathaniel (3 June 2021). "China Says WuDao 2.0 AI Is an Even Better Conversationalist than OpenAI, Google". Tom's Hardware.
- ^ "Cambrian-AI 2022 Predictions: Expect More Than Just New Chips". Forbes. 20 January 2022.
- ^ Heikkilä, Melissa (9 June 2021). "Meet Wu Dao 2.0, the Chinese AI model making the West sweat". Politico.
- ^ Greene, Tristan (3 June 2021). "China's 'Wu Dao' AI is 10X bigger than GPT-3, and it can sing". TNW | Neural.
- ^ "June 30, 2022 Meeting of the LF AI & Data Technical Advisory Council" (PDF). LF AI Data Foundation. 30 June 2022. Retrieved 6 April 2023.
- ^ "Wriggling abiotic worm manifests "intelligent life" rudiment". China.org / Xinhua. 2 June 2022.
