Are the LLM Wars the Database Wars? (rruxandra.github.io)

A recent analysis draws parallels between the evolution of large language models (LLMs) and the historical trajectory of databases, suggesting that LLMs may transition from revolutionary technology to ubiquitous infrastructure. Just as databases were once the centerpiece of tech innovation in the 1990s, attracting attention from major players like Oracle and IBM, LLMs are currently dominating discussions and competitions. However, the prediction is that, in the long run, the most successful models may not be those currently in the spotlight but rather the ones that become commonplace without necessitating user choice—much like how PostgreSQL and SQLite quietly became the go-to databases in countless applications today. This shift from exciting to essential reflects a broader trend in technology, where fundamental tools fade from view as they become integral to everyday operations. The significance for the AI/ML community lies in recognizing that while current developments may grab headlines, the real impact may come from less heralded models that demonstrate reliability and simplicity, suggesting a need for the community to focus not only on the glamour of launches but also on the longevity and adaptability of these technologies. As conversations about the future of LLMs continue, the challenge will be identifying which models are destined to achieve widespread, unnoticed success in the years ahead.