Is a Chinese ChatGPT on the way?

Following the launch of the ChatGPT chatbot by artificial intelligence start-up OpenAI (recently invested in by Microsoft), search engine giants cannot stand still.

Google CEO Sundar Pichai on 6 February sent an internal memo announcing its own artificial intelligence (AI) chatbot "Bard" to take on its old rival Microsoft. The project has been given top priority, with Google's two semi-retired founders called in to lead closed-door huddles on combining the fields of AI and search engines.

On 7 February, insiders told Chinese media such as The Paper that Baidu, the world's biggest Chinese search engine, will also be launching a ChatGPT-style project called ERNIE Bot, slated for release after internal tests are completed in March. While this is still rumoured news, Baidu's shares surged by 13% to hit a record high in nearly a year. (NB: Other Chinese tech giants like Alibaba, Tencent, NetEase and J.D.com have also announced similar projects.)

ChatGPT burst on the scene at the end of 2022, starting a wave of dialogue on AI.

It is actually a chatbot, which is familiar to many. But ChatGPT can learn and understand human language to carry on a conversation, and interact and chat in context like a real human. Before ChatGPT, there were no fundamental breakthroughs in studies on natural language, and some academics even thought that cognition or sentience and understanding of language is a uniquely human gift - no matter how machines tried to learn, it would only be an imitation without real understanding.

But ChatGPT has given astoundingly excellent answers to various questions from humans. Its strong understanding of natural language prompted Tesla founder Elon Musk to marvel: "ChatGPT is scary good. We are not far from dangerously strong AI."

Many tech communities and fans are also exploring the unlimited possibilities of ChatGPT. One netizen combined their loves of writing, suspense novels and jazz music, giving ChatGPT this request: "Develop a plot for a mystery novel where a disillusioned homicide detective hunts a clever serial killer who preys on musicians who perform the jazz standard Autumn Leaves."

ChatGPT understood the assignment perfectly. It replied: "Detective James Thompson is a seasoned veteran of the Homicide department, but lately he's been feeling burnt out and disillusioned. His latest case, the murder of a young jazz pianist who was found brutally killed in his home, only serves to deepen his sense of despair."

A business owner also had the idea of having ChatGPT respond to comments from diners. When a diner said, "Average food, nice view, but that's it," ChatGPT replied: "Thank you for taking the time to write a review of your experience with us. We're sorry to hear that you found the food to be only average, but we're glad you enjoyed the view. We always strive to provide our guests with the best food and dining experience possible, and your feedback will help us improve. If you have any specific suggestions, we would love to hear them. Thank you for dining with us, and we hope to have the opportunity to serve you again soon."

Another netizen made ChatGPT take an SAT test, where it scored 1020 out of 1600, putting it at the 52nd percentile according to College Board.

Surprisingly, when users instruct ChatGPT to find bugs in a programme, not only can it clearly identify the bugs and the cause, it can correct it and write a bug-free programme. An 11-year-old boy has developed a game for ChatGPT that has gained thousands of players.

ChatGPT can also be used in translation, email replies, text generation, data analysis, image creation, customer service, code generation and debugging, sparking discussion about ChatGPT taking the place of programmers, writers and even artists.

With social media flooded with jokes, discussions, questions and chatter about ChatGPT, tech leaders like Musk laughed at the New York Times for not tracking the news.

TED curator Chris Anderson tweeted: "I'm shocked that @nytimes appears so far to have had exactly zero coverage of ChatGPT, surely the biggest tech innovation of the year." Musk replied: "That's because ChatGPT is not a far left cause. The New York Times should rename itself The Social Justice Times."

The launch of a powerful chatbot like ChatGPT immediately put search engine giants in a bind, with Google, Baidu, Sogou and Qihoo 360 being the first to be hit. If users are able to gain adequately accurate information from ChatGPT, they would naturally stop using search engines.

Presently, users are seeing an increasing number of advertisements on search engines and search results are getting from bad to worse. Previously, advertisements only appeared on the top and bottom of the page; now, there are not only more of them where they used to be but they also appear in the middle of the page. On Baidu for instance, search results for certain keywords return an overwhelming number of advertisements, and you can only find what you're looking for a few pages later. In the third quarter of 2022, Baidu earned a total revenue of 32.5 billion RMB (US$4.57 billion), out of which 18.7 billion RMB came from online marketing revenue. Indeed, online marketing and advertising revenue reaps lucrative benefits.

Some analysts think that, compared to Baidu and other search engines, ChatGPT is able to offer answers by analysing and learning from existing information, making it more effective than search engines that only display pages of search results. Furthermore, it does not burden users with advertisements, and could become a strong rival of search engines in the future.

Five days after its launch on 30 November 2022, ChatGPT amassed over a million users. Two months later, the number of active users skyrocketed to about 100 million, and the chatbot became the fastest-growing consumer application in history. In comparison, TikTok took roughly nine months to amass 100 million monthly active users, while Instagram took two and a half years. Although there are numerous hurdles for ChatGPT to cross before it can truly replace search engines, its rapid user growth and powerful functions have certainly put enormous pressure on tech giants.

ChatGPT is powered by a class of machine learning Natural Language Processing models known as Large Language Model. Chinese newspaper Economic Observer said that Baidu is also equipped with these technologies, that is, it has a cross-modal semantic understanding ability across languages.

In terms of research and development (R&D) funding, Baidu CEO Robin Li said on numerous occasions that Baidu has made a cumulative investment of more than 100 billion RMB in over a decade, with the proportion of core R&D investments exceeding 20% several quarters in a row. He also said in 2017 that although Baidu does not earn the most, it certainly spends the most on R&D.

Baidu's massive R&D investments appear to be its pride and calling card in the tech industry. However, some netizens question whether Chinese tech enterprises are able to catch up with OpenAI's technology prowess. Netizens voiced their doubts under the topic "China has countless AI enterprises, but why does it not have ChatGPT?" on forum website Zhihu, with some explaining that China and OpenAI have chosen different technology paths - the former follows the BERT fine tuning path, while the latter follows the autoregressive language model. The difference in path has also led to a widening gap between the two. Other reasons include financial support, regulation, human resources and business culture.

Some netizens even asked ChatGPT directly why it was the US, and not China, who managed to create ChatGPT first.

ChatGPT replied with three points. One, the US has more advanced foundational technologies in the field of AI, such as neural networks and deep learning, which allows it to build large language models like ChatGPT. On the other hand, China's AI research is focused on applied technologies, such as computer vision and speech recognition, which makes it difficult to develop language models like ChatGPT.

Two, collaboration between AI researchers and research institutions in the US is more fluid and researchers are free to explore different research directions, leading to more innovation. In contrast, Chinese researchers are required to follow the direction of the country's AI laboratories, which makes it tougher to develop language models like ChatGPT.

Three, the US's academic climate in the field of AI is innovative and collaboration-centric, where researchers are free to explore different research directions and collaborate with other institutions or companies to develop new technologies. On the other hand, China's academic climate in the field of AI is more rigorous and competitive, where researchers need to gain certain scientific achievements to stand out. This makes it difficult to obtain sufficient funding and technical support to develop language models similar to ChatGPT.

Baidu introduced its own AI-Generated Content (AIGC) tool at an AI marketing summit last year and declared that AIGC will subvert existing content production models in the next decade, and that AI has the potential to meet massive demand for content at a tenth of the cost and a hundred or thousand times faster.

To investors, ChatGPT presents new opportunities and is seen as a springboard for subverting search engines and other tools. Older tech companies and Chinese tech enterprises are racing against time to see if they are able to break new frontiers once more and produce similarly outstanding products.

This article was first published in Lianhe Zaobao as ""中国版ChatGPT"指日可待？".

Related: AI is taking the place of mental labour but fear not | AI stars in China and the US lose their shine | Web3 with Chinese characteristics: Finding China's solution for regulators, developers and users | China's 5G ambitions undiminished by pandemic and sanctions | A metaverse with Chinese characteristics?