China Unveils WuDao 2.0, the World's Largest AI Model
The Beijing Academy of Artificial Intelligence debuts WuDao 2.0, a 1.75 trillion-parameter AI model that breaks the previous world record set by Google. The Chinese-language focused system shows near human-level abilities in creative tasks like poetry and painting.
The Beijing Academy of Artificial Intelligence (BAAI) unveils WuDao 2.0, a massive artificial intelligence model that features 1.75 trillion parameters. This groundbreaking achievement breaks the previous world record of 1.6 trillion parameters set by Google's Switch Transformer, making the Chinese system the largest pretraining model in the world.
WuDao 2.0 excels across nine different benchmark tasks and demonstrates remarkable cognitive abilities that approach passing the Turing test. The system performs exceptionally well in creative endeavors, including writing poetry, generating text summaries, answering complex questions, and painting pictures.
Unlike most large-scale models that rely heavily on English language data, WuDao 2.0 is built specifically around a Chinese language development environment. BAAI develops this homegrown super-scale intelligent model system with backing from the Beijing Municipal Science and Technology Commission to advance domestic AI capabilities.