SenseTime also took action, why did OpenAI suddenly "fall out"? Shanghai entrepreneurs communicated urgently with Microsoft
Yesterday, many domestic developers who used OpenAI's API to develop large-model applications received an email from OpenAI. This warning email shows that the company will take additional measures to block API traffic from "non-supported countries and regions" from July 9. In the list of countries and regions that support access to API services announced by OpenAI, there are no mainland China and Hong Kong, China. The company's official website also shows that in addition to the countries and regions on the list, "accessing or providing access to our services may result in your account being blocked or suspended."
What impact will this incident have on domestic large-scale model application development companies? What countermeasures can companies take? Today, a reporter from Jiefang Daily and Shangguan News interviewed entrepreneurs who develop large-scale model applications in Shanghai.
This incident also had an impact on the general large models in China. This afternoon, SenseTime released the "'RiRiXin' Large Model 0 Yuan Go" plan and announced that it would send "relocation" consultants to provide a series of training for corporate users to migrate from OpenAI to "RiRiXin".
"I just received a reply from a Microsoft employee," said Peng Weiguo, founder of a startup, in an interview. A Microsoft partner told him that he needed to consult the company headquarters whether Microsoft's cloud services would be affected by OpenAI's statement.
A few years ago, the company founded by Peng Weiguo was selected for Microsoft China's Entrepreneur Program and obtained a series of services provided by Microsoft Cloud Azure, including OpenAI's API services. Long before ChatGPT came out, he began to call programming interfaces and use the GPT large model to generate product introduction content for cross-border e-commerce platforms. After learning that OpenAI was going to take "additional measures" yesterday, he immediately contacted Microsoft China staff to ask whether Azure's related services would be affected.
Peng Weiguo said that in fact, OpenAI has never opened its API services to mainland China, and many domestic developers access the API of this American company by "climbing over the firewall". In his opinion, OpenAI has always adopted a "turning a blind eye" attitude towards this phenomenon and has not taken strong measures to prevent it. After all, China has a huge market and a developed artificial intelligence industry ecosystem, and cultivating and expanding local GPT downstream developers seems to have more advantages than disadvantages.
OpenAI CEO Sam Altman
However, just yesterday, OpenAI suddenly "turned against" the company. After discussing with several industry insiders, Peng Weiguo speculated that in addition to the well-known geopolitical factors, it may be related to the company's future development plan. "With the current size of OpenAI, it is difficult to raise funds through venture capital, and its profitability is currently limited, so it may go public in the future. After entering the preparation period for listing, the company will receive various inquiries and investigations about the Chinese market, such as whether the API services used by Chinese developers are compliant and whether this part of the revenue is stable." Perhaps after evaluating future risks, the company's board of directors decided to take additional measures to block API traffic from "non-supported countries and regions."
How should domestic developers who access the API by "climbing over the firewall" respond to the warning issued by OpenAI? Peng Weiguo's advice is: attach great importance to this warning and stop "climbing over the firewall" as soon as possible. Otherwise, once additional measures are taken from July 9 to strictly identify VPNs originating from mainland China and Hong Kong, related domestic companies may face great risks.
Regarding the choice of future development paths for enterprises, he believes there are two: If the market served by the enterprise is all in China, it is recommended to use a general large model developed by domestic enterprises instead. Although the comprehensive capabilities of GPT-4 are in a leading position in the world, the comprehensive capabilities of some domestic large language models are not weak. After completing the "domestic substitution", the quality of large model application products will not be significantly affected, and long-term risks can be eliminated. If the market served by the enterprise is mainly overseas, or there is a plan to "go overseas", then another way of "flesh climbing over the wall" can be chosen, that is, choose a country or region from the list of countries and regions that support access to API services announced by OpenAI, set up a physical organization there, and access API services in compliance with regulations.
In fact, domestic companies that develop general large models have already smelled business opportunities. Today, SenseTime released the "'RiRiXin' Large Model 0 Yuan Go" plan to give OpenAIAPI users a new home. The Hong Kong-based company announced: From now on, newly registered corporate users of "RiRiXin SenseNova" will receive a number of free service packages involving calls, migrations, training, etc. The "RiRiXin" platform covers multiple types of model API interfaces such as the "Shangliang" large language model, the "Shangliang" multi-modal large model of pictures and texts, the "MiaoHua" large model of text and pictures, the voice large model, the vector model, etc., which can meet the different needs of corporate users. SenseTime will give 50 million Tokens packages to newly registered corporate users of "RiRiXin", and send exclusive "moving" consultants to provide a series of training on migrating from OpenAI to "RiRiXin".
In April this year, Xu Li, Chairman and CEO of SenseTime, released version 5.0 of the “Daily Update” big model.
"From the user experience of large model applications, there is actually little difference whether developers use GPT or 'Ririxin', because before reaching AGI, there will not be a significant generation gap between GPT and domestic general-purpose large models." Peng Weiguo said, "For domestic developers, the top priority is to develop popular applications that can empower industries or attract many individual users, and truly transform large models into new quality productivity."
![](https://a5qu.com/s/user/default.webp)