A new Bay Area startup developing AI-powered search tools has launched out of stealth with $60 million in seed funding.
Legally known as MainFunc, the startup is developing a web search product called Genspark which taps into large language models to deliver results.
Lanchi Ventures, formerly known as BlueRun Ventures China, led the seed round.
MainFunc has dual headquarters in Palo Alto and Singapore and was founded by CEO Eric Jing and CTO Kay Zhu.
Jing and Zhu have been working on search-related products for nearly two decades. They met while working at China-based internet provider Baidu, and previously worked at Microsoft and Google, respectively.
"Search needs need to be reimagined," Jing said, with a shift away from the current model which delivers a bunch of random links littered with advertising. "Data is actually the key in the AI era. So, no matter what kind of a model you use, what kind of a user interaction, you'll still need a bunch of really high-quality premium data to generate a good result."
Genspark uses LLMs developed in-house as well third-party models from OpenAI and Anthropic, Jing said. It could use other models such as Meta's Llama 3 in the future, as well.
And it's not just about delivering short summaries but also creating an AI agent to generate more complex results.
When users run a search query, Genspark creates an interactive page with modules for a table of contents, citations and overviews. It calls these Sparkpages which can also be annotated by users.
Jing also acknowledges that the company hasn't fully solved the tension between fair use and copyrighted content but they do intend on respecting media paywalls.
"I think that there's a line here," Jing said, and how to work with publishers is "a deeper problem we should solve."
Rival AI search startup Perplexity triggered a media storm last week after it scraped and repackaged a Forbes article across multiple channels.
Genspark's approach is actually similar to vertical searches which users experience at specialized websites, Jing said.
"Vertical search exists because they have very high-quality premium data, right? Either Amazon or Tik Tok or YouTube or Instagram, they have really spent a lot of time building the ecosystem, encouraging the creators to create great content," Jing said. "I think a problem for traditional search is traditional search just builds a bridge between a user and devices."
It's unclear exactly what Genspark's revenue model will be but the company is considering various options for the future which could include premium subscriptions, advertising, sponsored links and revenue sharing, Jing told me.
"We want to build everything, even the business model, from the ground up," Jing said.
The company currently has fewer than two-dozen employees and will keep the team relatively lean even as it grows.
AI-powered search has been attracting investors' interest.
San Francisco-based Perplexity announced a $63 million round in April which valued the company at more than $1 billion, and the company is already in talks to raise another $250 million in a new round that would more than double its total funding and could triple it valuation, according to recent reports from TechCrunch and the Information.
Google still dominates search around the world as of early 2024 with nearly 82% of the market for desktop and more than 95% of mobile, according to Statista.
The next largest players for search on desktop are Microsoft's Bing which captures more than 10% of the market, followed by Yahoo and Russia's Yandex with less than 3% each, and DuckDuckGo and Baidu with less than 1% each.
Yandex also has around 1.5% of all mobile search traffic with other top providers capturing less than 1% each.
