Opinion: Anthropic's export control compliance highlights the risks of AI centralization, and decentralized AI may become a key counterbalance

Comparing news, CoinFund founder Jake Brukhman said that the AI model naturally has centralized attributes and is also a key target for government supervision and control, and Anthropic's latest export control compliance actions further confirm this trend.

He pointed out that decentralized networks can be an important counterbalance to the current situation. The core challenge in building a sovereign, open, and public decentralized AI is the issue of computing power. Although outsiders generally believe that only trillion-dollar technology companies can undertake cutting-edge model training, there are actually sufficient general-purpose GPU computing power resources around the world, and the key is to develop new distributed training algorithms.

Brukhman said that teams such as Gensyn, Prime Intellect, Bagel, Pluralis, Nous Research, Macrocosmos AI, and Covenant AI have been exploring this direction. Although early outsiders generally thought it was not possible, it has been proven that distributed training is not only possible, but is also less expensive and more efficient than traditional solutions.

Furthermore, he sees economic sustainability as another major challenge facing decentralized AI. While the open source model is important, it lacks a mature business model, and Pluralis is exploring the business path of tokenizing AI models by distributing model weight among participants.

Brukhman said that it is currently at a critical moment. Whether AI is fully centralized and controlled by censorship and a unilateral government, or whether it is public AI built on an open decentralized network, will determine the future development direction of the industry.