- It’s been more than two months since the ChatGPT craze hit China, but a similar AI-based product has yet to reach the entire Chinese population.
- The most famous alternatives released by Baidu, Alibaba, etc. have limited access with waiting lists or restricted trials to business partners.
- ChatGPT, which learns from big data, has reached an estimated 100 million monthly active users in the two months since its launch in November.
Baidu CTO Wang Haifeng speaks during the launch of Baidus AI chatbot Ernie Bot at an event in Beijing on March 16, 2023.
Michael Chang | Afp | Getty Images
BEIJING — More than two months after the ChatGPT craze hit China, a similar AI-based product has yet to reach the entire Chinese population.
Instead, the most prominent alternatives released by Baidu, Alibaba and others have restricted access on waiting lists or limited trials to business partners.
Compared to “iPhone Moment”, ChatGPT reached an estimated 100 million monthly active users two months after its launch in November. AI chatbots can learn using big data and generate everything from poetry to business strategies in human-like conversations.
But ChatGPT, created by US-based OpenAI, is not available in China, where access to Twitter, Facebook, and Google is also blocked through government Internet firewalls. Beijing also released draft rules this month to regulate AI-generated content, with a public comment period he has until May 10.
These restrictions did not stop the Chinese press and social media in general from talking frequently about ChatGPT and AI technology. Some even tried to buy overseas ChatGPT accounts on Chinese e-commerce sites.
Domestic companies rushed to release and test alternative products. Big data accumulation and machine learning experience are essential to the technology behind ChatGPT.
So far, publicly available numbers suggest similar AI products are not as widely available in China.
Alibaba Cloud announced Wednesday that it has received more than 200,000 requests from businesses to test its version of the ChatGPT-style technology called Tongyi Qianwen. The product was announced on April 11th.
One of our business partners, Kunlun Tech, launched “Tiangong” on April 17th, allowing users to interact in a question-and-answer format. This product is currently invite-only. Kunlun claims that Tiangong is the only chatbot in China with his ChatGPT level training metric.
Similarly, it’s not clear how many people have gained access to Baidu’s Ernie bot.
Less than a week after its launch on March 16th, the chatbot has over 1.2 million people on its waiting list. The company stopped disclosing numbers within days. There was no update on Friday.
When CNBC tried to sign up for the Ernie bot in March, the Baidu system asked for a mainland Chinese phone number and a local ID to use the chatbot. In China, more and more apps now require or strongly encourage identity verification.
Baidu said this week that it has updated its product four times since the release of Ernie bot, reducing the cost of operating AI models to one-tenth of what it used to be.
The public version of the Ernie bot allows users to generate text, images and sounds in English and Chinese.
In the US, new AI-based products from Google and Microsoft also have queues. Due to server capacity, ChatGPT is not consistently available. Some within the industry expect such AI tools to be easier to commercialize for business products than public tools such as search.
Regulatory uncertainty lingers, but Beijing is stuck in the comment period on the generative AI draft rules. The agency has not announced when the final version of the rule will go into effect.
Outside of China, the US and Europe have generally been lenient on ChatGPT, with the exception of Italy, which banned chatbots this month until OpenAI addresses privacy concerns.
A bug allowed ChatGPT users to temporarily access other people’s conversations. OpenAI announced this week that it has released a secret-like feature that allows users to turn off their chat history and opt out of having their data used to train their models.
Another challenge for Chinese companies is obtaining state-of-the-art chips for training AI models. The US announced in October a strict export ban aimed at limiting access to China’s high-end semiconductors.
HSBC analysts reported on April 20 that OpenAI’s GPT-3 level model requires at least 1,000 Nvidia A100 graphics processing units (also known as GPUs) to complete a single 23-day training round. a kind of chip that can be
More than 30 companies and research institutes in China are training such AI models, indicating “strong demand for AI servers and increasing spending on network infrastructure,” it said. the report said.
Analysts say they expect demand for AI GPUs in China to grow by more than 40% this year.
Another estimate, based on a survey of cloud providers, predicts demand will at least double, contributing to a “significant shortage” of chips, according to a representative of a company investing in AI models. . .
As for regulation, the source said the agency supports ChatGPT-style technology but plans to regulate it.
— CNBC’s Michael Bloom and Ashley Capoot contributed to this report.
