Everyone wants to be ChatGPT – the scramble for a truly Chinese chatbot

By CHENG Lu, LI Jingya

As ChatGPT casts its spell on the world, Big Tech everywhere, not to mention Small Tech, is plunging its arm into the boiling cauldron of chatbot development.

Shouting nonsense about FOMO, CFOs are pouring billions of dollars into the bubbling concept, quite a lot into product development, but also quite a lot into ever-lucrative product fakery.

When the magic soup cools, there will only be one or two viable chatbots in the world, or China. Western tech is dominated by effective monopolies – (Google, Facebook, Microsoft) - and dualities – Android, iOS. There is no reason to suppose chatbots will be any different. It’s pretty safe to say that almost all the investment in chatbots will prove to be money well wasted, willfully wasted.

Monopolies, knockoffs and distractions

ChatGPT has pitted Google against Microsoft in a bot-v-bot deathmatch. Google’s Bard was roundly mocked when the AI made an error on a list of esoterica a lazy dad could chat to his kid about.

ChatGPT, Microsoft and OpenAI and have bunged the bot into the Bing search engine and the Edge browser, two of Microsoft’s most disliked products and “the biggest thing” to happen in the nine years since Satya Nadella has been CEO.

On the other side of the Pacific, shares in China with even the most remote connection to chatbots have soared. Everyone is tossing cash at the concept and shouting about it.

Wind, a fintech outfit, has 24 stocks in a ChatGPT index, which registered its craziest rise on February 8 when all of the constituent stocks knocked their socks off. Hanvon hit seven consecutive limit-ups, which triggered an inquiry by Shenzhen Stock Exchange. The traditional bullies - Baidu, Alibaba, 360 Security – all unveiled beautifully optimistic plans for dream-like AI products.

The market is now a hotchpotch of potential monopolies, knockoffs and distractions.

Generations adrift

Most players are only talking up the concept to give prices a push in the right direction before throwing the lot overboard when the time is right. Maybe tomorrow. It might turn out to be wishful thinking that some meaningful products or technologies may be left after the tides ebb.

China has nothing like technology on par with ChatGPT, but that may not be a bad thing as the structure of Chinese language learning and style of interaction may lend itself to entirely different concepts.

Of the 24 stocks in Wind’s ChatGPT index, only two hold natural language processing (NLP) technologies, and neither is capable of producing AI-generated content (AIGC) or outputting long-form conversations. A third of them aren’t even remotely related to OpenAI, ChatGPT, or AIGC businesses.

Currently, 360’s AI products are behind ChatGPT by generations, which is perhaps where most Chinese AI developers are.

As the market resumed rationality and inquiry letters were issued by regulators, some tech companies underwent sharp price falls.

Trillions of interactions

After all, the underlying machine learning model is what defines an excellent chatbot product. Apart from that, massive cloud computing capacity and data are needed to train the chatbot. Only a few giants in China could develop sophisticated chatbots. Alibaba and Bytedance still have the potential if they make full use of their AI research institutes.

Baidu will insert a ChatGPT-style chatbot, Ernie Bot, into its search engine in March. Baidu owns a huge text database thanks to its search engine business.

Alibaba Damo Academy has been developing a chatbot for shopping searches. JD.com’s NeuHub focuses on smart customer support.

As a Chinese Q&A counterpart of Quora, Zhihu has accumulated rich and quality AIGC learning samples for developing AI. Zhihu hosts 600 million entries of content and half a billion questions and answers. Microsoft’s Chinese version of Bing displays many Zhihu-generated answers as search results. Tencent also owns a colossal text database.

Peaceful coexistence

AI chip makers, algorithm developers, and AIGC researchers will benefit. The competition in the AI tech world will boil down to a competition of technologies.

In the United States - the main bot battlefield - as a partner and investor of OpenAI, Microsoft is the big winner.

Based in San Francisco, California, OpenAI rolled out the bot in November. Now with a monthly active user base of over 100 million, ChatGPT’s subscription service has proved its commercial potential.

According to Fortune magazine, OpenAI was expected to make less than US$30 million in its first year, meaning a loss of more than half a billion dollars. With ChatGPT’s surprising success, OpenAI’s value has expanded 100-fold. The company told prospective investors that it expects revenue of US$200 million in 2023 and more than a billion in 2024.

Microsoft’s stock increase was moderate given the disruptive shock. Bing’s market share has never surpassed 4 percent, and Google’s dominance has hardly been below 90 percent since 2009. LI Di, CEO of Bing’s “virtual teenager” bot Xiaoice, thinks that chatbots and search engines will coexist for a long time.

Waiting for the native speakers

But Microsoft’s true motivation should be advertising. In 2022, advertising revenue rose to US$18 billion, still only a tenth of Google’s in a US$500 billion market. There is plenty of room for improvement via ChatGPT.

With its leading position in cloud computing, Microsoft may turn the tables on Google in the advertising market. Google, the AI world champion, now feels threatened and seems desperate under a tightening anti-monopoly environment.

Google’s original plan was to unveil its chatbot at the I/O conference in May. ChatGPT has caught Google on the hop. Kicking Bard to centerstage before he was ready caused the bot to fluff his lines. Google stock plunged by 7.4 percent, wiping out over US$100 billion of the company’s value.

It's common for both Chinese and international AI startups to partner with Big Tech. But international pioneers are taking the lead. China’s primary market has been quite reserved as venture capital has mostly invested in specialized AI applications in vertical industries like marketing, health, security, etc. Some even doubt there’ll ever form any homegrown independent AI ecosystem.

As the market awaits a native Chinese bot, knockoffs have run rampant. Parody ChatGPTs can be found in WeChat accounts and app stores. Connected to OpenAI using API technology, these copies function as an intermediary between ChatGPT and users. And these middlemen have started charging high fees already.

Security risk

Artificial intelligence is still far from reliable. It can easily be trained to spread bias and misinformation. Powerful chatbots bring powerful challenges to data security.

“There is a heightened concern about how the large information services firms are limiting opportunities for new generations of competitors to come forward,” said William Kovacic, professor of antitrust law at George Washington University to the Financial Times.

Anti-trust regulations could turn out to be very important. The regulatory bureau has been on Google’s case much harder than it ever was on Microsoft.