China’s newest chatbot sensation, Ernie, developed by search engine giant Baidu, has been facing scrutiny due to its censorship issue. The chatbot has been programmed to deflect sensitive and controversial questions, raising concerns about limited freedom of speech and censorship in China. While Ernie has garnered immense popularity, attracting millions of user enquiries within 24 hours of its launch, its inability to provide satisfactory responses to critical questions has called into question its reliability and trustworthiness as an AI chatbot.
The Chinese government’s strict censorship policies, which affect various online platforms and communication channels, also impact Ernie’s functionality. When asked difficult questions such as the absence of Xi Jinping from the upcoming G20 meeting or the Chinese government ceasing the publication of youth unemployment data, Ernie either redirects users to official sources or claims ignorance regarding the question. This evasiveness raises concerns about the chatbot’s lack of transparency and its role in disseminating accurate information.
Notably, Ernie has been specifically trained to identify contentious keywords and phrases. Consequently, when users inquire about sensitive topics such as Xinjiang and Tibet, Ernie asserts its inability to respond. This cautious approach indirectly reinforces the Chinese government’s denial of alleged human rights violations against Uyghur Muslims in Xinjiang and the suppression of ethnic Tibetans. Ernie’s unwillingness to engage in conversations on these sensitive subjects further highlights the limitations on free speech and critical discussions in China.
While it is possible that the technology behind Ernie is not yet advanced enough to handle complex queries, there are instances where the chatbot deliberately avoids answering certain questions. When asked about the health status of Xi Jinping or his predecessor Hu Jintao, Ernie swiftly changes the topic, indicating a deliberate avoidance of potentially sensitive information. Similarly, queries regarding significant events like the Tiananmen Square crackdown or the names of jailed figures such as Bo Xilai and Liu Xiaobo elicit the same response, discouraging users from pursuing these topics further.
The extent to which Chinese chatbots are limited by censorship remains uncertain, as Baidu has not provided a clear response to inquiries regarding their limitations. However, the company’s CEO, Robin Li, expressed intentions to collect real-world user feedback to improve Ernie Bot’s performance and enhance the overall user experience. Baidu emphasizes that the chatbot is just one aspect of their broader AI services under the Ernie model, aiming to empower entrepreneurs to develop groundbreaking AI applications.
This emphasis on business applications aligns with recent regulations in China that require generative AI models to conform to “core socialist values” and avoid disseminating information that undermines state power or national unity. This regulatory framework may encourage companies like Baidu to focus more on developing AI applications for commercial purposes rather than serving the general public. Professor Jeffrey Ding from George Washington University highlights the notable quality gap between Chinese models like Ernie Bot and OpenAI’s ChatGPT, citing data quality and research orientation as contributing factors.
Baidu, which dominates the Chinese internet search engine market with over 90% market share, has been relying on Ernie to revitalize its financial performance. While Baidu faces stiff competition from other tech companies and has missed out on significant advertising revenue, Ernie represents a potential growth opportunity for the company. However, with several competing chatbots entering the market, the success of Ernie remains uncertain.
The prevalence of censorship and limited freedom of speech in China raises concerns about the reliability and transparency of AI chatbots like Ernie. While these chatbots may offer convenience and engagement for users, their inability or unwillingness to address critical questions undermines their credibility. As the Chinese government continues to enforce strict censorship policies, it casts doubt on the future development and application of AI technologies in China.