The CEO of Chinese tech-giant Baidu said that, since China’s new AI regulations have been in place,
over 70 AI models have been released in China with over 1 billion parameters.
Robin Li, CEO of Chinese tech company Baidu, said on Sep. 5 that more than 70 artificial intelligence
(AI) models have been released in the country, according to a report from Reuters.
Li commented that the over 70 AI language models that have been released were created with over 1
billion parameters, which technically qualifies them as large language models (LLMs).
Currently, some of the biggest LLMs on the market include Meta’s Llama, released earlier this year,
containing 65 billion parameters, and OpenAI’s Chat GPT-4, which has not publicly released its
parameters but is rumored to have over 170 trillion.
Parameters matter when it comes to LLMs because an increase in the number of parameters means the
model is able to capture more details in the data. Ultimately, this helps models perform better.
Baidu, a prominent AI developer and internet services provider in China, released the latest version of
its Ernie chatbot to the public on Aug. 30. It was among the first companies to release chatbots for public use after the enactment of China’s new AI laws.
According to Li, the latest version of Baidu’s chatbot, Ernie 3.5, is equipped with a processing speed
twice that of the previous version and an improved efficiency of 50%. He also revealed that the
company plans to launch a newer version in the “near future.”
Earlier this summer, on Aug. 3, the Chinese e-commerce and tech giant Alibaba launched two new AI
models that are open-sourced and said to rival Meta’s Llama 2.
The Chinese government only recently passed and implemented its AI laws. Previously, companies were
able to launch small-scale tests of their AI products, which, under the new rules, widened the test size
and enabled more features.