Alibaba’s AI development could be stop-and-go, as outgoing CEO highlights regulatory restraint in speech
- AI regulation is as important as developing the technology, said Alibaba’s outgoing CEO Daniel Zhang, who will continue to lead its cloud operations
- The comments align with Beijing’s goal of taming the tech’s erratic nature while keeping pace in the global AI race
“Hitting the accelerator while also … hitting the brakes for security are equally important,” Zhang said at the Nishan Dialogue on Digital Civilisation in Qufu, a city in eastern Shandong province, the birthplace of the ancient Chinese philosopher Confucius. This strategy requires a “cooperative effort from the whole society to pre-empt any possible problems brought upon by the accelerated development of AI technologies”, he added.
Chinese state media drums up support for country’s AI development initiatives
The event was organised by the World Internet Conference (WIC), a government-backed annual forum started in China in 2014 to promote Beijing’s governance model for cyberspace. In 2021, the WIC expanded its membership to become an international organisation, although most of its funding still comes from the Chinese government.
Alibaba Cloud is the leading business segment in developing and commercialising AI for the tech giant. It is the largest cloud computing business in China, but maintaining good business relationships with state-owned enterprises and government clients requires maintaining goodwill with Beijing.
The subsidiary is one of six new business groups under a corporate restructuring this year that will see each group run independently and able to seek outside funding. Alibaba Cloud is currently in the process of seeking an initial public offering, to be concluded in the next 12 months, according to the company.
Every player in this space has been looking to “make some achievement”, Zhang noted.
Baidu founder and CEO Robin Li also appeared at the Nishan conference, where he said the company’s Ernie foundation model was recently upgraded to version 3.5. It has double the training speed and can simultaneously handle 17 times more queries compared with the version launched in March, he said.
The internet search giant said the new model has significantly improved its output by better aligning with user expectations and in content safety that better adheres to local laws and regulations.
“It’s foreseeable that large [language] models will permeate into more and more sectors, becoming a key factor driving the digital economy and its integration with the real economy to further grow it,” Li said in his speech.
Zhou Hongyi, founder and chairman of cybersecurity firm Qihoo 360, echoed Li in his own speech at the event, saying vertical AI models catering to specific sectors will be the future. “Turning large [language] models into versatile tools in governmental or industrial areas will bring real value,” he said.
While some issues such as AI ethics are still being sorted out, Zhang highlighted the enthusiasm the company has seen from clients in this space. More than 100,000 customers from a wide range of sectors have applied to use Tongyi Qianwen, according to Zhang.
“At the new starting line of the intelligent era, we can say that [AI] will rewrite the way a lot of sectors run,” he said.