I guess my question would then be whether the translation would work if neither language contained any information on microphysics or advanced math. Would the model be able to translate e.g. “z;0FK(JjjWCxN” into “fruit”?
The chances of the LLM being able to do this depend heavily on how similar the subjects discussed in the alien language are to things humans discuss. Removing areas where there is most likely to be similarity would reduce the chance that the LLM would find matching patterns in both. Indeed, that we’re imagining aliens for the example already probably greatly increases the difficulty for the LLM.
I guess my question would then be whether the translation would work if neither language contained any information on microphysics or advanced math. Would the model be able to translate e.g. “z;0FK(JjjWCxN” into “fruit”?
The chances of the LLM being able to do this depend heavily on how similar the subjects discussed in the alien language are to things humans discuss. Removing areas where there is most likely to be similarity would reduce the chance that the LLM would find matching patterns in both. Indeed, that we’re imagining aliens for the example already probably greatly increases the difficulty for the LLM.