We use cookies to tailor your experience and present relevant ads. By clicking “Accept”, you agree that cookies can be placed per our Privacy Policy
ACCEPT
avatar image
Advertisement

Alibaba tests ChatGPT rival as Chinese tech giants like Baidu race to build country’s best AI chatbot

  • An Alibaba spokeswoman has confirmed that the company’s research institute Damo Academy is conducting an internal testing of an AI chatbot
  • Damo launched a natural language processing model with 27 billion parameters in 2021, after OpenAI released GPT-3 with 175 billion parameters in 2020

Reading Time:3 minutes
Why you can trust SCMP
0
The Alibaba office in Beijing. Photo: Bloomberg
Coco Fengin Beijing

Alibaba Group Holding said it is testing a ChatGPT-like service that takes advantage of its years of research in large language models, joining web search giant Baidu in a rush to create China’s answer to San Francisco-based OpenAI’s chatbot, which has set off a frenzy among investors and the general public.

A spokeswoman at Alibaba, owner of the South China Morning Post, confirmed on Thursday that the company’s research institute Damo Academy is conducting an internal testing, but did not provide a timetable for service launch.

Baidu previously said it expects to complete testing of Ernie Bot, its ChatGPT-like service, in March.

Sam Altman, CEO and co-founder of Open AI, speaks during an event at Microsoft, which is fusing ChatGPT-like technology into its Bing search engine. Photo: Bloomberg
Sam Altman, CEO and co-founder of Open AI, speaks during an event at Microsoft, which is fusing ChatGPT-like technology into its Bing search engine. Photo: Bloomberg

Alibaba, a Chinese e-commerce juggernaut, is also one of the country’s leading players in artificial intelligence (AI) research.

In April 2021, Damo introduced a natural language processing model – Pretraining for Language Understanding and Generation – which has 27 billion so-called parameters – a measure of the size and sophistication of an AI model. That same year, the institute built an AI model known as M6, calling it the world’s first 10-trillion-parameter pre-training model.

By comparison, OpenAI’s Generative Pre-trained Transformer 3 (GPT-3), the language processing model that underpins ChatGPT, launched in 2020 with 175 billion parameters.

In an annual report published last month, Damo listed “generative AI” – which creates new output based on a set of texts, images or audio files it has been trained on – as one of the top technology trends to watch for in 2023.
A Baidu robot at the company headquarters in Beijing. Photo: Reuters
A Baidu robot at the company headquarters in Beijing. Photo: Reuters
Coco Feng
Coco Feng joined the Post in 2019, covering the technology and internet sector from the Greater Bay Area. Previously, she worked at the Post's Beijing bureau.
Advertisement
Select Voice
Choose your listening speed
Get through articles 2x faster
1.25x
250 WPM
Slow
Average
Fast
1.25x