Wudao Large Model

WuDao Large Model Official Website, China's First and World's Largest AI Large Model

What is the WuDao Large Model?

The Beijing Academy of Artificial Intelligence (BAAI) has launched its innovative product - WuDao 2.0, the follow-up version to China's first ultra-large-scale intelligent model system. WuDao is a language model designed to surpass OpenAI's GPT-3 and Google's LaMDA in human-level thinking ability. After training with 4.9TB of images and text data and exceeding the state-of-the-art (SOTA) in 9 benchmark tests, WuDao is closer than any other similar model to achieving general artificial intelligence (AGI) and human-level thinking ability.

WuDao Large Model Official Website:https://www.baai.ac.cn/portal/article/index/cid/49/id/518.html

How is the WuDao Large Model?

To train WuDao, we used a total of 4.9TB of high-quality English and Chinese image and text datasets, including:

• 1.2TB of Chinese text data
• 2.5TB of Chinese image data
• 1.2TB of English text data

WuDao's training is based on the open-source MoE (Mixture of Experts) system FastMoE. MoE is a machine learning technique that works by dividing the prediction modeling task into multiple subtasks, training expert models (learner models) for each subtask, and developing gating models. The gating model learns to select which expert model to consult based on input features and combines the predictions of various experts. FastMoE allows WuDao to consult different expert models in parallel and switch to the best-performing model. For example, when input is English text, WuDao uses a prediction model specifically designed for generating responses to English text.

The launch of WuDao 2.0 represents a significant advancement in China's AI field. Its development aims to promote the development of AI and bring broader application prospects to all industries. In the future, we look forward to further optimization and evolution of WuDao to better meet user needs and continuously exceed the boundaries of AI.

AI Large Model

AI Large Model:

The application of large models has transformed the three key elements of AI from "data, algorithms, and computing power" to "scenarios, products, and computing power." The scope of generative AI is not limited to text but also includes voice, images, videos, and other modalities, which will gradually move from technical breakthroughs to productization and commercialization, promising more social and commercial benefits.The era of large models, represented by GPT, has arrived, with rapid rise in multi-modal AI technologies like voice, images, and videos. With broader data forms, more application scenarios, and deeper user experiences, the future is promising.

The era of large models represented by GPT has arrived, with rapid growth in multi-modal AI technologies such as voice, images, and videos, encompassing a wider range of data forms, more application scenarios, and deeper user experiences.