Will OpenAI and Anthropic Service? (medium.com)

The article discusses a paradigm shift in the AI landscape, suggesting that the future may lean towards millions of specialized models rather than centralized large foundation models typically offered by companies like OpenAI and Anthropic. Traditionally, users interact with these models through simple prompt-and-response inference transactions, but the piece argues that the true value lies in transforming these interactions into lasting training assets. This shift would allow organizations to capture knowledge, refine their capabilities, and build models tailored to their specific domains, thus turning AI expertise into durable, machine-readable assets. Significantly, this evolving architecture implies a move from a centralized knowledge model—which optimizes for breadth—to a distributed one that focuses on depth and specialization. As organizations develop their capabilities based on captured insights from foundation models, the economic value may shift towards companies that excel at converting general intelligence into specialized knowledge, enabling organizations to create a sustainable competitive advantage. The article posits that foundation model providers will likely transition from being seen as mere answer engines to functioning more like educational institutions that continuously train and support a network of specialized systems, ultimately changing the ownership dynamics in AI capabilities.