This is an interesting thought experiment. I've done similar ones. However, I think it would be very feasible for humans, given a large corpus of an unknown language, a computer, and standard methods from computational linguistics from 20 years ago, to decypher and understand the language. So this thought experiment does not, unfortunately, tell us anything about the ability of ChatGPT to understand what it's saying.
that's a ridiculous comparison, because your human in this example necessarily has the embodied knowledge of a first language to compare with the second language.
you cant arbitrarily take statistics of a language and make meaning out of it, and in that sense there is no meaning to linguistics without comparative linguistics.
Comments
This is an interesting thought experiment. I've done similar ones. However, I think it would be very feasible for humans, given a large corpus of an unknown language, a computer, and standard methods from computational linguistics from 20 years ago, to decypher and understand the language. So this thought experiment does not, unfortunately, tell us anything about the ability of ChatGPT to understand what it's saying.
that's a ridiculous comparison, because your human in this example necessarily has the embodied knowledge of a first language to compare with the second language.
you cant arbitrarily take statistics of a language and make meaning out of it, and in that sense there is no meaning to linguistics without comparative linguistics.