Wu Dao is capable of doing everything GPT-3 can do, and then some.

Earlier this week, researchers from the Beijing Academy of Artificial Intelligence launched a new version of Wu Dao, their own deep learning natural language model, which appears to be leagues ahead of GPT-3, the model from Open AI, and Google’s Switch Transformer. The last-named two models each represented massive advances in natural language processing. (TechRadar)

Wu Dao 2.0

While GPT-3 was trained on 175 billion parameters, and the Switch Transformer on 1.6 trillion parameters, Wu Dao 2 uses 1.75 trillion parameters.

Wu Dao 2 was trained by studying 1.2 1.2TB of text in each language and 4.9TB of images and text overall. This massive model can therefore understand both Chinese and English and can simulate conversations, understand pictures, write poems, and even create recipes.

This model is targeting artificial general intelligence, the Holy Grail of AI. It refers to the hypothetical ability of a machine to learn a task such as humans do.

Chinese AI researcher Blake Yan said to the South China Morning Post: “These sophisticated models, trained on gigantic data sets, only require a small amount of new data when used for a specific feature because they can transfer knowledge already learned into new tasks, just like human beings. Large-scale pre-trained models are one of today’s best shortcuts to artificial general intelligence.”

Wu Dao 2 is mega

“The way to artificial general intelligence is big models and big computer,” Dr. Zhang Hongjiang, chairman of BAAI, said Tuesday, as quoted by Engadget.

“What we are building is a power plant for the future of AI, with mega data, mega computing power, and mega models, we can transform data to fuel the AI applications of the future.”

Related Story:  NERSC Launches Perlmutter, The World’s Fastest AI Supercomputer, For “Extreme-Scale” Science