It's just that there is no way for the model to map the token meanings to a iseful representation. Yes, in the general case, you can make a model for most type

It's just that there is no way for the model to map the token meanings to a iseful representation. Yes, in the general case, you can make a model for most types of logic problems (but be careful there, there are real problems you couldn't represent as "a bunch of numbers" and a single transformation), the issue is that the model cannot build that representation even for simple cases because it,'s used to operate at on language (and maybe some other fuzzy input type), which has to be _reliably_ interpreted to this number space; in different ways for different problems; without breaking the language processing.