As in the United States and Europe, ChatGPT has sparked renewed interest in AI, a sector that was starting to look a little moribund. “Some big companies had actually laid off their teams working on large language models,” says Xie Mingxuan, founder of an AI startup called vrch.io, adding that companies now regret it because those teams later founded their own startups.
But developing AI models outside of large companies is much more difficult in China than in the United States. American companies, like OpenAI, have been able to access huge amounts of data from Google or social media platforms like Twitter and Reddit. But China has ignored the open web, essentially moving from no internet to apps, which are much harder to extract data from.
This, coupled with the cost of computing power, makes it difficult for startups like Xie’s to build the kind of huge, radical models that their equivalents in the United States are trying to create. Most therefore focus on the application level, instead of creating their own models. own models.
Founded last year, vrch.io develops an AI-powered voice-input image generator. In the past, interior designers might have needed to use renderings made in Photoshop to show their clients. Now, when people want to redesign a space, they can do it on the spot using generative AI. “For those of us who work in design,” says Xie, “we spent most of our time converting information that is difficult to express accurately into words, into images, and then using those images to communicate with clients .”
Although vrch.io has investment from Miracle Plus (formerly Y Combinator China), a startup incubator in China, it is not currently targeting the Chinese market. This is due to the lack of clarity in the regulations.
“As a small company,” says Xie, “we cannot guarantee that every segment of the business, whether it is the algorithms, the data sources, or the training of the models themselves, is compliant to the regulations.”
In July this year, China’s Cyberspace Administration released draft guidelines on generative AI focusing on privacy, personal information protection, algorithm transparency and intellectual property rights. They didn’t set compliance standards for technology that were fundamentally different from existing tech regulations, but startups like Xie’s are waiting for more details.
