The database wars and where LLMs are (maybe) headed (rruxandra.github.io)

The recent discussion in the AI/ML community suggests that large language models (LLMs) may evolve into a fundamental, albeit unexciting, component of technology similar to traditional databases. Just as databases like PostgreSQL and SQLite became integral to various applications and ultimately faded into the background, LLMs could similarly become ubiquitous and essential, functioning seamlessly without drawing attention. The shift from groundbreaking technology to invisible infrastructure raises questions about the future trajectory of LLMs and their role in software development. This perspective highlights the importance of open-source models in driving widespread adoption and integration. Postgres's rise as a standard database choice, alongside SQLite’s silent dominance in devices, emphasizes that the most successful technologies often thrive without fanfare. As the AI/ML community reflects on the past and considers future developments, the lesson is clear: the true winners may not be those at the forefront today but rather the tools that quietly empower innovations across diverse platforms. The evolution of LLMs, therefore, may hinge on their ability to integrate seamlessly into applications, becoming essential without the need for explicit choice or recognition.