ChatGPT has been making headlines for two and a half months in the press and ‘trending topics’ on social networks, internationally. And that includes China, despite its sociological and technological peculiarities… since users have to resort to VPN services to access the OpenAI platform.
On February 11, the hashtag “Will ChatGPT replace teachers?” (#教师会被ChatGPT取代吗#) became a trend on ‘Chinese Twitter’, Weibo, generating 120 million posts. On the same day, the Chinese state TV website, CCTV, released a post on “Ten professions that could be replaced by ChatGPT.”
That is, the same questions that the Western public asks. But there is another, and even more important, question that is largely only being asked by Chinese users: “Why didn’t ChatGPT create it first?”.
Certainly, in China they have had very popular chatbots focused on social interaction for some time, but nothing remotely similar to the knowledge synthesis capabilities displayed by the AI of OpenAI, key to the success it has reaped. AND The answers to that question can be grouped into three reasons (compatible with each other):
A) Chinese tech companies focus on applications that can be profitable quickly instead of making long-term bets on R&D, which forces them to wait for others to develop the base technology. That is the response of Jiang Ruxiang, a professor of sociology at Peking University, who points to companies such as the ASML chipmaker, which had to invest in R&D for many years before becoming profitable. .
Already in 2019, Xataka reported that China had just surpassed the US in the number of investigations into artificial intelligence… but that had not yet managed to surpass it in the impact and relevance of the same (It is true that then it was estimated that this would take place around this time). Without addressing that, and considering how long it can take for scientific research to translate into monetizable products, it is understandable that the US industry still retains the initiative in this field.
OpenAI, the creator of ChatGPT, is a lucrative non-profit company (yes, it’s a complex topic) founded in 2015 in the US by several American entrepreneurs (mainly Elon Musk and Sam Altman). Just 7 years later, that company has been able to launch both the fashionable ‘chatbot’ and one of the great platforms for generating images with AIDALL-E 2.
B) Training in linguistic models is more difficult in the Chinese languagedue to its complex nature. ChatGPT itself understands the questions in Chinese and generates answers in the same language, but in a unanimous way the Chinese users of the platform ChatGPT claim that you have difficulty grasping all the nuances of their languageespecially in matters such as jokes.
C) Censorship and ‘sensitive terms’. The main Chinese social platforms are monitored by the authorities in real time, and censorship is applied with the same speed, but implementing this same mechanism on platforms similar to ChatGPT would be complicated at best. This Weibo post sums up the feeling of many users: “It is impossible for a Chinese version of ChatGPT to come out, there are too many sensitive words“.
It is true that the Chinese Internet regulator, the Cyberspace Administration of China, will never allow free access to a tool that can provide data to the user. without first going through CCP censorship: ChatGPT shies away from certain controversial political issues… but controversial by Western standards, not Tibet or Tiananmen. In fact, ERNIE-ViLG, the Chinese ‘DALL-E 2‘, released last August by Baidu, blocks such topics in the same way that DALL-E 2 blocks pornographic content.
ZAO, the Chinese MOBILE APP that through DEEPFAKE turns you into DICAPRIO in SECONDS
The sleepy giant stretches
But just as the success of ChatGPT among the public has awakened Google from its slumber, on the other side of the Pacific the Chinese technology giants have decided that they do not want to be left behind in this field and several of them have announced in recent days that they are working on launching their own ‘big language models’. As far as we know, they are the following:
- Baidu: The ‘Chinese Google’ plans to start testing its ‘ErnieBot’ in March, which would not only offer functions equivalent to ChatGPT, but also image generation. Baidu is one of the leading Chinese companies in the field of AI, having launched Kunlun, its first artificial intelligence chip, in 2018, and has been investing in autonomous driving technologies for several years. and just released the second version of the aforementioned ERNIE-ViLG; ‘ERNIE’ is the collective name for a whole series of language models that would be equivalent to OpenAI’s GPT (which is the basis of both ChatGPT and DALL-E 2).
- Alibaba: The e-commerce giant claims to already have a technology similar to ChatGPT in the testing phase, with the aim of integrating it into several of its cloud platforms.
- JD.com: Alibaba’s big rival plans to launch its ‘ChatJD’which would not be aimed at the general public but at the retail and financial industry.
- Tencent: The Chinese social platform giant (it has WeChat and QQ in its portfolio) has also claimed to be working on its own model in the image and likeness of ChatGPT, but He has not given more information about functions or deadlines. The company’s language model, HunYuan, already topped the Chinese Language Comprehension ranking last year.
In the last week, including the Beijing municipal government (capital of China and city with the highest concentration of companies dedicated to AI in the entire Asian country) has announced the launch of aids to promote research in this field. And to that is added the ‘work plan for AI’ approved by the communist regime a few years ago in order to promote Chinese leadership in this field throughout the present decade.
They are not doing badly: the TOP5 of companies with the highest number of AI patents is led by two Chinese companies (Tencent and Baidu), and Ping An ranks fifth. Microsoft and Alphabet/Google rank, respectively, in sixth and seventh place. The three Chinese companies were far behind the Western ones in 2017.
Image | via Lexical