Beijing Academy of Artificial Intelligence
Beijing Academy of Artificial Intelligence, also known as Zhiyuan Institute, is a Chinese non-profit artificial intelligence research laboratory. BAAI conducts AI research and is dedicated to promoting collaboration among academia and industry, as well as fostering top talent and a focus on long-term research on the fundamentals of AI technology. 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. 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.
Products and applications
, BAAI's research focuses on large pre-trained models and open-source AI infrastructure.WuDao
WuDao is a large multimodal pre-trained language model. WuDao 2.0, was announced on 31 May 2022 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; making it the largest pre-trained model in the world at the time. WuDao was trained on 4.9 terabytes of images and texts. The chairman of BAAI said that WuDao was an attempt to "create the biggest, most powerful AI model possible"; although direct comparisons between models based on parameter count do not directly correlate to quality.WuDao has demonstrated ability to perform natural language processing and image recognition, in addition to generation of text and images. 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.
FlagAI
FlagAI 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 sandboxlevel of the Linux Foundation.