Ask HN: Can LLMs be used to invent a new human language?

Well, for starters, there is no such thing as a language that is 100% scientific and 0% political. Even in constructed languages, specially one made by AI.

What AI does is aggregate its training data and predict next responses based on training. As a result, it's choices on what to keep and what to exclude from the language would be coloured by the bias of the training data, a bias we have seen has ideological motivations, even if unintended.

It is an illusion to think of AI as we know them today as being without bias and without ideology.

Furthermore, the English language's place as the de-facto international language of the western world is a matter of both historical relevance (the influence of the late British Empire) and convenience. A newly constructed language, even one made purposefully to that end by AI, would be neither historically relevant nor convenient. This is much the same issue that Esperanto, humanity's attempt at a new lingua franca, ran into. Esperanto is easy to learn and easy to speak, and yet it is neither historically relevant nor convenient to learn and use, and as such is relegated, like most other conlangs, to relative obscurity and niche uses.

If you want to talk science, I suggest picking up some linguistics, which is a hardcore science even if most people don't think of it as such. I would recommend starting with Saussure's lectures, and then studying some of Noam Chomsky's early work, and from there the world is your oyster. It is a legitimately incredible field and it saddens me a bit that it gets understudied, specially in today's world where communication is so important, perhaps more than it has ever been before.

I don't know. It seems to me that if genAI invents it, it's not a human language. At least not unless people start using it for real and introduce the inevitable human changes and touches over time.

There's no need in a new artificial language. If there was a need, humans could easily invent one without any LLM assistance, just like we invented Esperanto. But there is none. BTW, English is neither the first nor the last international language. It will be (and is gradually) replaced by others.

Yes, they're state-of-the-art for translation, so I'd say that they'd be good at it. Give it a try, see how they go with it.

To be honest, i think LLMs at the current time wont have that capability. It still struggles with understanding context and tone in its translations