Kai-Fu Lee, chairman and chief govt officer of Sinovation Ventures, speaks through the HICOOL World Entrepreneur Summit on September 11, 2021 in Beijing, China.
China Information Service | China Information Service | Getty Photographs
BEIJING — Chinese language synthetic intelligence fashions could also be at the least half a yr behind these developed within the U.S., however Chinese language AI apps will probably take off a lot sooner, mentioned Kai-Fu Lee, former head of Google China.
He was referring to giant language fashions, that are educated on huge quantities of information that may course of and produce textual content, pictures and movies.
The highest Chinese language firms’ LLMs are about six to 9 months behind their U.S. counterparts, whereas much less superior Chinese language fashions might lag the U.S. by about 15 months, Lee mentioned. He was talking on the AVCJ Non-public Fairness Discussion board China on Wednesday.
Lee, creator of “AI Superpowers: China, Silicon Valley, and the New World Order,“ is a extensively adopted commentator on AI, and is the founding father of startup 01.AI in addition to enterprise capital agency Sinovation Ventures.
“Apps, I might predict, by early subsequent yr will proliferate in China a lot sooner than within the U.S.,” Lee mentioned, noting that the price of coaching a superb AI mannequin has fallen considerably.
“It is inevitable that China will [build] the very best AI apps on this planet,” he mentioned. “However it’s not clear whether or not it will likely be constructed by large firms or small firms.”
Lee, whose startup is targeted on search apps proper how, mentioned it might take 5 to eight years to take generative AI client functions to the subsequent stage — a single “tremendous app” that may carry out a number of duties.
The trade will probably want fully new units versus current smartphones, he mentioned, including “the precise gadget should be at all times on, at all times listening.”
Main Chinese language firms similar to Alibaba and Tencent have launched their AI fashions and enterprise merchandise. These firms and buyers have additionally backed a number of AI startups.
Beijing-based ShengShu Expertise, backed by Alibaba-affiliate Ant Group, introduced Wednesday that its text-to-video mannequin Vidu has launched a brand new function for bettering how a most important component or character in AI-generated clips may be portrayed persistently, with out distortion. That may allow advertisers to create promotional movies for his or her merchandise.
Vidu was launched earlier this yr and its primary instruments are open to the general public, with extra superior capabilities obtainable through subscription. Co-founder and CEO Jiayu Tang instructed reporters Wednesday that a number of firms had been fascinated with shopping for ShengShu’s companies, and weren’t simply exploring the tech.