My understanding is that the unique hash is really the true identifier of the implementation. So when you change the implementation, the two functions will have two different hashes? I am not sure if this is exactly your problem.
Testing basic knowledge is not racist. Removing such tests actually undermines democracy, as democracy is the regime that requires the most that each individual has a strong ability in assessing situations in a quantified and calculated manner. That is how the voting system gets close to an effective decision process.
This is true. But only to a point where mimicking and more broadly speaking, statistically imitating data, are understood in a more generalized way.
LLMs statistically imitates texts of real world. To achieve certain threshold of accuracy, it turns out they need to imitate the underlying Turing machine/program/logic that runs in our brains to understand/react properly to texts by ourselves. That is no longer in the realm of the old school data-as-data statistics I would say.