because these genetic algorithms are best suited for consistent behaviour
i noted in another comment that this ga cracker repo will only work on passes that resemble the ones it is trained on, but like all genetic algos is aggressively ineffective on inputs resembling anything else
as for programming the practice, it is embarrassingly redundant and consistent due to the structure of programming being based on grammars and logic
so imagine a program that has a data set
you need ten bits of information about that data set, like max, min, mean, etc..
so you write 10 individual functions to loop over the data and interpret it
that is all very consistent behavior and code
now imagine you write a simple genetic algorithm that can interpret your interests and write those simple loops for you on the fly
ten functions of 10 lines each is 100 lines of code
one function of 30 lines that can generate then discard those ten previous functions gives your program a 70 line and immeasurable man hour advantage
think if a search engine needed an engineer to explicitly write the discovery function of every possible query, it would be impractical, so instead you imagine generalisations that permit simple functions to take many varying inputs and produce many varying outputs
what i think the parent comment is suggesting is that these generalisation functions themselves could be built by computers and one possible method being genetic algorithms
i noted in another comment that this ga cracker repo will only work on passes that resemble the ones it is trained on, but like all genetic algos is aggressively ineffective on inputs resembling anything else
as for programming the practice, it is embarrassingly redundant and consistent due to the structure of programming being based on grammars and logic
so imagine a program that has a data set
you need ten bits of information about that data set, like max, min, mean, etc..
so you write 10 individual functions to loop over the data and interpret it
that is all very consistent behavior and code
now imagine you write a simple genetic algorithm that can interpret your interests and write those simple loops for you on the fly
ten functions of 10 lines each is 100 lines of code
one function of 30 lines that can generate then discard those ten previous functions gives your program a 70 line and immeasurable man hour advantage
think if a search engine needed an engineer to explicitly write the discovery function of every possible query, it would be impractical, so instead you imagine generalisations that permit simple functions to take many varying inputs and produce many varying outputs
what i think the parent comment is suggesting is that these generalisation functions themselves could be built by computers and one possible method being genetic algorithms
furthering the abstraction
all puns intended