We talked about how ChatGPT revitalised the AI sector in China – and in a true Chinese fashion, many big techs, startups and billionaires joined the race to build China’s answer to GPT.
Baidu, which went “All in AI’‘ in 2017, seems to carry most of the expectations. The company behind China’s top search engine has (most likely correctly) identified AI as the next major trend it can’t afford to miss, after having missed mobile internet.
The “All in AI” journey has not been smooth from the onset – Qi Lu, a former Microsoft EVP hired as Baidu’s COO to execute the strategy, left after barely a year on the job.
Baidu carried on, however, progressing AI initiatives including “Apollo” autonomous driving platform and “Ernie” series of large language models.
When ChapGPT upended the generative AI game – Baidu’s partners probably demanded a solution, as direct ChatGPT access in China can be tricky for regulatory reasons. (This might not be an issue after all, as OpenAI partner Microsoft has traditionally been able to find ways to make products available in China).
While anticipation is higher, and the consensus that Baidu is probably best positioned to respond to ChatGPT, there is also a lot of scepticism in China, often rooted in people’s frustration with ad-infested Baidu search results. Of course, Google flopping on Bard – its hastily launched generative AI chatbot – did not help either.
OpenAI’s unveiling of GPT-4, less than 2 days before the Ernie-Bot launch, probably added to the pressure on the Baidu team. Because, whatever you are showcasing on 16 of March will be benchmarked – by analysts, investors, partners and media – against the capabilities of GPT-4.
Finally, on 16 March, Robin Li, Baidu’s founder CEO, stepped onto the stage and revealed Ernie Bot. A roller coaster ensued.
As Robin Li progressed with the demo, Baidu’s share price dropped almost 10%. People seemed worried that the pre-recorded demo videos signified that Baidu was not confident with its product.
The next day, the price surged almost 14%, as analysts who had the chance to test Ernie Bot wrote positive reviews.
Over the weekend, we have been overloaded with reviews written by different people, with different tests and conclusions. We did some tests ourselves – quite often Ernie Bot performed more poorly compared to its GPT counterparts.
Most reviews are dismissive, and a few positive ones were criticised as paid propaganda (media in China have a habit of writing paid ‘soft articles’ without disclaimer). Jokes also abound to show Baidu’s incompetence – for example the following depicts an Ernie-generated image of “A spacecraft ablaze flying towards earth”, only that the prompt included Chinese adjective “bear bear” which was used to describe the raging fire (not that there is an actual bear):
However, we find that all these criticisms were missing the point. Ernie Bot, in our opinion, is a good enough (though nowhere near perfect) product and in a market with high demand (and a Wall).
Baidu has good AI experts – and they rushed out a product within a short period of time. Baidu does not have the spending power or abundant access to chips that Microsoft has furnished OpenAI – remember that.
This is not the end of Baidu’s journey – they will continuously make improvements. Remember all the jokes laughing at Dall-E or ChatGPT when they were first launched? What makes us think that Baidu can’t make the same leaps and bounds in improvements as OpenAI has been?
Strategically, China will want its own version of ChatGPT. Baidu is not guaranteed to win this race but it is in a good position. Whether it will actually make it depends, as we are saying again and again, on its leadership, people, organisation and finally its product, in that order.
Hope Robin Li and his core team have learnt from its decade of failures, and seize this opportunity well.
We have elaborated case studies on Baidu, and other major Chinese tech companies, in our Seeing the unseen: behind Chinese tech giants’ global venturing.