i don’t see why China shouldn’t restrict the best AI either and it’s not just because the Europeans might ban Chinese AI as Germany has banned DeepSeek, but because there’s no point giving your finest goods to ingrates who think they’re superior to you even as they wallow in faeces like Indians do.
I don’t think you are giving Chinese strategists the credit they are due here.
China making its existing AI models open source isn’t done out of naivety or idealogical principles, it’s done based on hard cold strategic calculations.
Firstly it’s done to put China on the map in terms of AI and make it impossible for the inevitable western smears about Chinese models being copies.
The second reason is to create secondary markets for hardware, services and whole ecosystems. China’s bread and butter is manufacturing and they want AI to generate new demand for a whole new market of AI related hardware products, which will also in turn generate services and eventually a whole ecosystem. That means it wants as many people and companies to be able to use AI as possible, so the costs to running AI models should be affordable.
This is in stark contrast to the western strategy of gatekeeping and rent seeking where they ‘win’ the AI race by bankrupting all competitors to achieve an effective monopoly and then lock access behind paywalls and rent seek to extract every last penny possible out of users with complex pricing strategies.
This fundamental difference in anticipated end-use also backflush into how the two sides went about designing their AI models.
China designed their AI models for maximum efficiency to minimise costs of adoption, while western designers went with maximum cost to basically start an arms race to see who can afford the most data centres to bankrupt their competitors asap so they can move onto monopolistic predatory pricing and maximum profit extraction sooner. It doesn’t matter if they only sell a tiny number of licenses if they are the only game in town.
By making its AI model open source and free to all, China directly advances its own AI strategy while also effectively fucking up the west’s.
But all of this is already achieved by releasing ‘good enough’ AI models. I think this new direction is to set some ground rules and limits for Chinese AI companies on the max performance model they should release openly to prevent them trying to compete in that direction and gifting western AI companies with superior Chinese tech. China still wants and needs western AI companies to continue to kill themselves by burn through mountains of cash trying to stay in the game instead of free riding off of Chinese AI companies.