LLMs: The New Frontier or The Same Old Song?

Large Language Models (LLMs) are the new darlings of computational concept analysis. But here's the kicker: they're not the first to tread this path. Remember digital methods and distributional approaches? They're still around, haunting today's tech like persistent ghosts.

Old Problems, New Players

Before LLMs, we had digital methods in the history, philosophy, and sociology of science (HPSS). These methods, combined with distributional approaches and lexical semantic change detection, laid the groundwork for today's LLMs. Yet, those early tools faced hurdles like corpus construction and operationalization. Fast forward, and guess what? LLMs inherit these same challenges.

So, what's new? LLMs promise more accuracy and nuance in detecting lexical semantic shifts. But here's the twist: it's a double-edged sword. More data doesn't always mean more clarity. Are we merely dressing up age-old issues in fancy new algorithms?

LLMs: Revolution or Rerun?

When LLMs entered the scene, they brought hope of a revolution. Case studies in HPSS started employing these models, seeking richer insights. But let's zoom out. No, further. See it now? The fundamental questions remain. How do you choose the right corpus? What about model choice and the trade-offs in operationalization?

Sure, we've got more powerful computing and sophisticated models. But the core questions of evaluation and interpretation are stubbornly unresolved. Everyone has a plan until exhaustion hits. Are we truly progressing, or just running in circles?

The Future: Cautious Optimism

In the era of LLMs, optimism is rampant. They're hailed as game-changers, promising breakthroughs in understanding concept evolution. But are they just overextended? Bullish on hopium, perhaps.

LLMs offer an exciting frontier, no doubt. They might untangle complex webs of historical and philosophical concepts. Or they might just amplify biases we've yet to address. This ends badly. The data already knows it.

The real question is, are we asking too much of them? Are we setting ourselves up for disappointment in our quest for perfection? As we embrace LLMs, let's not forget that technology is only as good as the problems it solves. And right now, it's on shaky ground.