Ant Group Invests in AI Company Meta Sota

Meta Sota recently completed a new round of financing exceeding 100 million yuan, led by Ant Group with participation from Lightspeed China Partners, with a post-financing valuation reaching 150 million USD. Previous shareholders of Meta Sota include Future Capital, Cheetah Mobile, and Amino Capital.

It is understood that large tech companies like Baidu, Inc. and Tencent also engaged with Meta Sota during this round of financing, but ultimately, Ant reached a deal with Meta Sota. Since last year, Ant Group has invested in at least six companies in the AI field, including large model company Zhipu AI, Moonshot AI, video generation company AISphere, ShengShu Technology, AI chip company Moffett AI, and the latest Meta Sota.

In addition to investments, Ant released its self-developed “Bailing” multimodal large model last year, established a new AI department called “NextEvo” at the beginning of this year, and launched the “Smart Assistant” feature on Alipay in April. Ant Group Chairman and CEO Jing Xiandong mentioned at the World Artificial Intelligence Conference this year that Ant’s AI strategy is to develop “professional intelligent agents,” which are AI systems with expertise in fields such as finance and healthcare.

Meta Sota founder Min Kerui is a serial entrepreneur. In 2012, he co-founded the natural language technology provider ‘Boson Data’ and served as CTO. The company was later acquired by Ant Group in 2015. In 2017, Min Kerui joined Cheetah Mobile as the chief scientist.

In April 2018, Min Kerui founded Meta Sota, which subsequently launched the legal AI translation product ‘Meta Translation’ and the proofreading product ‘Meta Writing Cat’; in 2022, they also introduced the article generation product ‘Quantum Sketch,’ which surpassed ten thousand daily active users within a week of its launch.

These products are either aimed at professionals or have not attracted widespread attention. Last year, as substantial funding flowed toward leading AI companies, the mid-tier company Meta Sota did not benefit from the wave.

Until late March of this year, Meta Sota’s AI search product ‘Meta AI Search’ was officially launched, with website visits exceeding 7 million that month. According to the website traffic monitoring platform Similar Web, Meta Search ranked third in March among a group of AI products in China, just behind Baidu, Inc. and Kimi of Moonshot AI; the growth rate for that month reached 550%.

Meta Sota is one of the earliest Chinese companies to pay attention to AI search. Min Kerui mentioned on social media this January: ‘I think we can train a version of Perplexity AI in two weeks.’

Unlike Perplexity, which ‘does not build models, only applications’, Min Kerui once told us that Meta Sota uses its self-developed models, but the parameters are not that large. ‘A hundred-billion model is a burden in terms of cost; in applications, we would prefer it to be as small as possible, rather than the opposite,’ Min Kerui said last year.

One viewpoint posits that considering AI search products like Perplexity as ‘Google killers’ is somewhat laughable. These products still need to access search engine APIs to provide their final services and must pay for this; for example, Perplexity is still a customer of Google.

After Microsoft Bing launched its AI search feature last May, its market share actually decreased in the US. It wasn’t until the end of last year that the market share finally increased by 1% year-on-year.

The business model for AI search is also unclear. An AI investor who did not invest in the Meta Sota stated, ‘AI search faces a problem: how can the ROI of a single search be higher than that of Google right now? How can advertising revenue exceed search costs? By how much? Google itself hasn’t managed to do that.’

In addition, there are views suggesting that if China is considered the main market, mobile apps like Douyin and Xiaohongshu have cut into a large number of independent search engine usage scenarios, thereby compressing the space for third-party search engines.

Finally, what exactly are the barriers in the product form of AI search? An entrepreneur who has worked on search engine large models once told us that he believes there are two types of ‘winner takes all, where the strong get stronger’ potential for AI to C products. One is the network effect, exemplified by communication products like WeChat; the other is the data flywheel, which means the more people use it, the more data there is, leading to better model capabilities and product experience, and search is one of those instances.

SEE ALSO: Ant Group’s AIGC Application ‘Zhi Xiaobao’ Has Over 10 Million New Users