What Would Wittgenstein Say Today?

“Philosophers are often like little children,” Wittgenstein writes in Philosophie: they scribble marks on a sheet of paper and then ask an adult what they are.

I think there was also some frustration behind that remark, as he recognised the limits of a project he had once thought possible: identifying a logical structure capable of explaining how language represents the world. A structure that no longer seemed to be enough.

At the time he was writing, the physiological mechanisms of language were still poorly understood. Broca and Wernicke had already carried out their nineteenth-century studies of people with brain lesions and language disorders. Today we certainly know far more about how the brain works, although we still do not fully understand how that passage between word and meaning takes place.

I wonder whether, if one day we discovered with absolute certainty how language works, how knowledge works, and how our own consciousness works, that momentous discovery would be the final nail in philosophy’s coffin. Would it put an end to twenty-seven centuries of speculation about how we see the world?

Perhaps. But I don’t think so.

Philosophy has certainly explored the mental processes that allow us to know things. It has not reached particularly definitive conclusions, truth be told. But it has helped build the tools we use to discuss what counts as valid knowledge and why we should regard it as such. And part of what we used to call natural philosophy has become what we now call science.

My point, though, is something else: even if we discovered every minute detail of how the brain produces thought, we would still question the nature and meaning of thought itself. Process and product are not the same thing, particularly here.

Our understanding of language now faces new challenges that would surely have intrigued Wittgenstein enormously.

LLMs give us language whose relationship with reality passes through a long chain of mediations. The learning process involves stages, choices, procedures and algorithms that significantly shape what the model learns and what it then gives back to us. The world an LLM learns about and speaks of reaches it through a process profoundly different from our own experience of it. What emerges is a representation shaped by those differences, even when what it says is factually correct.

And yet their language is so strangely similar to ours.

This raises a particularly interesting question about the nature of language. The world as we represent it and the world as an LLM represents it are not the same. Yet we seem able to talk about it coherently and meaningfully. Or at least that is what appears to be happening.

And so I return to Wittgenstein, to the early Wittgenstein.

How can language be isomorphic to the world if the same language can express such different representations of it?

Is that language still a representation of the world, or merely its distant echo?