The current landscape of semantic layers in data management is fragmented, with numerous competing standards leading to forced compromises, lock-in, and inefficient APIs. As LLMs evolve, they may redefine the use of semantic layers, promoting more flexible applications despite the existing challenges of interoperability and profit-driven designs among vendors. A push for a universal standard remains hindered by the lack of incentives to prioritize compatibility across different data systems.

AI models require a virtual machine-like framework to enhance their integration into software systems, ensuring security, isolation, and extensibility. Drawing parallels to the Java Virtual Machine, the proposed AI Model Virtual Machine (VM) would allow for a standardized environment that promotes interoperability and reduces complexity in AI applications.