‘Pangu Chat,’ a ChatGPT competitor, is built on Huawei’s own large-scale learning model, which purportedly has 100 billion parameters.
The Chinese multinational ‘Huawei’ is apparently producing its very own artificial intelligence chatbot titled the ‘Pangu Chat’.
According to information, the ‘Pangu Chat’ is constructed with Huawei’s own large-scale learning model, which is built with 100 billion parameters, making Pangu Chat the first Chinese pre-trained huge model with more than 100 billion parameters.
The Pangu Chat is a moderately powerful AI chatbot that operates approximately identically to the GPT-3 scale. The GPT-4 model, on the other hand, is said to include around 170 trillion parameters, far outnumbering Huawei’s huge language model.
The Pangu Chat, named after the large-scale Pangu cloud architecture, will be unveiled on July 7th at the Huawei Developer Conference (Cloud) event.
If you are a typical consumer who is looking forward to using the Huawei Pangu Chat, we have some bad news for you: the Pangu Chat will initially be available for business, with intentions to offer the Chatbot to government and enterprise users later on.
In order to train the Pangu cloud model, Huawei has spent a total of 960 million yuan over the last three years. A big portion of this money was spent on 2000 Huawei Ascend 910 chipsets, and 4000 GPU/TPU cards.
The Pangu-E large model, a successor to the Pangu cloud model, is tested to have an output of nearly 1.085 trillion parameters, a number that competes for the GPT-4 and easily beats the GPT-3.5.