Michael O’Rourke argues that the reliance on centralized infrastructure contradicts Web3's principles of decentralization and autonomy. Open data, essential for the burgeoning tech economy, must transition to decentralized frameworks to enhance security and accessibility. Centralized AI models incur significant costs, exemplified by the $100 million expense for OpenAI’s GPT-4 compared to merely $5.5 million for DeepSeek R1. Additionally, decentralized networks can lower barriers for AI training, facilitating affordable access to robust models. The inefficiencies of centralized hosting platforms are highlighted by historical outages that disrupted services like MetaMask during regional sanctions. A shift to decentralized infrastructure not only supports open-source applications but also incentivizes data sharing among researchers, freeing them from the constraints of high-cost journals. The future points towards specialized middleware protocols handling storage and networking, urging developers to embrace trustless and censorship-resistant decentralized infrastructures.

Source 🔗