It is one of the hottest debates in the AI race and that is making the difference between the two most advanced powers. China is committed to Open Source with models such as Deepseek, while in the United States they lead private developments such as Chatgpt or Claude. Meta, on the other hand, has been defending the Open Source for a long time. Or we believed. According to the New York Timesnuevo Mark Zuckerberg Superintelligence Team I would be questioning this strategy in favor of developing a closed model.
Flying. If confirmed, it would be a drastic change in the finish line. In 2023, goal It joined other companies To defend the creation of open source models and have not been few times when Mark Zuckerberg has said that his AI is Open Source (although not all true). Now that the new Superintelligence Team commanded by Alexandr Wang He has started working, they are considering whether to continue with the plan or choose to develop a closed model.
Behemoth. It is the name of the AI Open Source that Meta was developing, its largest model to date. In fact, Behemoth’s bad results would have been The trigger for Zuckerberg decided to create the new AI team. Nearby sources claim that one of the first discussions of the team has been to leave its development. There is also the possibility that Behemoth’s development continues and that the new team focuses on developing a new model that points to the great goal of goal: Create a general artificial intelligence.
What is an open AI. The Open Source concept is not new, but with the arrival of AI there has been enough debate about its definition. The AI models are very complex and the classical definition did not adapt to them, so in October last year the Open Source Initiative agency The Osaid published (Open source ai definition), a definition with which Target I did not agree. In order for an AI open source, it must allow the following:
- Use the system for any purpose and without having to ask permission.
- Study the operation of the system and inspect its components.
- Modify the system for any purpose, even to change your performance.
- Share the system so that others use it, with or without modifications, for any purpose.
Chinese patience. The United States and China are the two most powerful players in the AI race, but although they play the same game, they do it with very different strategies. With Deepseekhis most famous AI model, China surprised to bet on the open source, a strategy that has allowed him Avoid the zancadillas that your rival puts With vetoes and restrictions. And it is not the only oun source in China, there are many more: Qwen from Alibaba, Leather de Bytedance, Ernie de Baidu, Hunyuan Turbo of tencent o Like De Moonshot Ai.
It is not a matter of giving in exchange for anything, behind there is a strategy of soft power Technological: Today I give you access, with the intention of being dominant tomorrow. China wants to create global dependence on its models and be an alternative to the payment models of the Americans. China has the future vision and patience to do so, we have already seen how it invested for decades to form engineers and that today gives it a clear advantage in the AI race.
American immediacy. Openai, Anthropic, Google… The greats of the AI of the United States have opted for private development and subscriptions to access their most leading models. The exception to the rule was put by the goal with flame, the only one whose code is accessible (at least part), But if they focus on private development, that closed approach will be armed even more. The objective is clear: they have the most advanced models and resources to continue training them, so they can obtain an economic benefit by monetizing them. They do it through subscriptions to pro versions or with specific products such as AI agents, it is the case of the newly announced ChatGPT Agent.
The trap. China’s strategy makes all the meaning if we take into account that they were behind their rival (although they have managed to cut positions in record time). With open and free IAS they manage to create a great very fast user base and create that dependence. However, it is not a good long -term strategy from the economic point of view. If in the future they begin to monetize, that sooner or later they will have to do it, they face two problems: that their users come out in disarray and that the regulators put their eyes in them.
The United States maintains strong control and obtains benefits from minute one. Has the most sophisticated models, but every time prices are higher And that makes a luxury within reach of a few. Many will look for alternatives and there will be China with their open and free models. They are two opposite visions: one focused on the present and one in the future. We will have to wait to see what vision is imposed.
Imagen | Bibek ghosh, y kaboompics (pexels)
In WorldOfSoftware | Four AI companies are monopolizing the intellectual future of humanity. They are not good news
