Great Firewall of China: Barrier to Baidu’s Global AI Aspirations

Listen to this story

Cashing in on the current AI wave, China-based tech firm Baidu is developing an AI-powered chatbot similar to OpenAI’s popular ChatGPT, Bloomberg reported. What’s more? Baidu plans to embed the chatbot into its search engine as early as March, much before ChatGPT is embedded into Microsoft Bing.

While ChatGPT has been trained on OpenAI’s GPT3.5 architecture, Baidu’s chatbot has been trained on ERNIE (Enhanced Representation through Knowledge Integration) 3.0 Titan, a language model developed by the Chinese tech giant. 

GPT-3.5, which is a fine-tuned version of GPT-3, has 175 billion parameters, ERNIE 3.0 has 260 billion parameters. Interestingly, the initial version of ERNIE, which was launched in 2019, outperformed Google’s BERT, a pre-trained language representation model.

AIM Daily XO

Join our editors every weekday evening as they steer you through the most significant news of the day, introduce you to fresh perspectives, and provide unexpected moments of joy
Your newsletter subscriptions are subject to AIM Privacy Policy and Terms and Conditions.

Undoubtedly, ChatGPT has been a game-changer and has elevated the generative AI industry to new heights, but the media attention it receives is not unprecedented. The Wall Street Journal called Baidu’s chatbot ‘China’s entry into the global AI race’. However, for Baidu, there are a few stumbling blocks for it to fulfil its global AI ambitions. 

The Great Firewall of China Barrier

In China, the government controls the internet. The Xi Jinping-led administration claims the authority to censor internet content within its jurisdiction. China is, in fact, one of the strictest censorship regimes in the world. From political dissent and discussion of sensitive historical events in China to information about its leaders and high-level officials, the government censors almost everything through the Great Firewall of China programme.

Download our Mobile App

Through the programme, the Chinese government blocks access to websites and online content that the government deems harmful, sensitive, or in violation of its regulations. So, the question is whether Baidu’s chatbot will also face the Chinese government’s censorship. Unfortunately, the answer to this mostly is yes.

ERNIE-VilG, Baidu’s text-to-image AI generator, a competitor to DALL-E2 and Stable Diffusion, already rejects politically sensitive prompts. Chances are high that Baidu’s chatbot’s responses will toe the government’s line and this could prove to be a stumbling block for Baidu’s global ambitions.

Additionally, OpenAI is continually refining ChatGPT through human feedback. It is utilised globally, whereas Baidu’s chatbot, if only employed within China, could result in a noticeably different training process or even biases for the outside world.

Catering to China Only

Baidu is the most-used search engine with a market share of 72.37%, as of May 2021. However, the Chinese search engine had a global market share of just 1.64% during that period, according to Statista.

When it comes to the architecture, ERNIE 3.0 Titan has been pre-trained on heterogeneous Chinese data. However, since the chatbot will be available in both Chinese and English, it means the chatbot has been trained on English datasets as well. But for now, it’s difficult to assume how accurate Baidu’s chatbot will be, compared to ChatGPT. (Let’s not forget ChatGPT is also not highly accurate).

Since Baidu’s chatbot and ChatGPT are trained on a completely different dataset, their responses too will vary. While it might cater to the Chinese audience, how successfully will it cater to the global audience remains to be seen. 

Sign up for The Deep Learning Podcast

by Vijayalakshmi Anandan

The Deep Learning Curve is a technology-based podcast hosted by Vijayalakshmi Anandan - Video Presenter and Podcaster at Analytics India Magazine. This podcast is the narrator's journey of curiosity and discovery in the world of technology.

Pritam Bordoloi
I have a keen interest in creative writing and artificial intelligence. As a journalist, I deep dive into the world of technology and analyse how it’s restructuring business models and reshaping society.

Our Upcoming Events

24th Mar, 2023 | Webinar
Women-in-Tech: Are you ready for the Techade

27-28th Apr, 2023 I Bangalore
Data Engineering Summit (DES) 2023

23 Jun, 2023 | Bangalore
MachineCon India 2023 [AI100 Awards]

21 Jul, 2023 | New York
MachineCon USA 2023 [AI100 Awards]

3 Ways to Join our Community

Telegram group

Discover special offers, top stories, upcoming events, and more.

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Subscribe to our Daily newsletter

Get our daily awesome stories & videos in your inbox

Council Post: From Promise to Peril: The Pros and Cons of Generative AI

Most people associate ‘Generative AI’ with some type of end-of-the-world scenario. In actuality, generative AI exists to facilitate your work rather than to replace it. Its applications are showing up more frequently in daily life. There is probably a method to incorporate generative AI into your work, regardless of whether you operate as a marketer, programmer, designer, or business owner.