If they have fixed programming, the bias would be consistent, but still there, because it would be based upon systems that are already inherently bias.
Any current ML system is beholden to the data/constraints it was built with, if inherent bias exists in the data it will exist in the resulting system.
That’s before you even start taking in to account the infrastructure that would be managing them being potentially corrupt or having their own interpretations of “public safety”.
“These bots from <generic third party> are bringing in more cases against the <“good” people>, but these ones from <tech company with the same bigoted ideology as us> can be tweaked to target the <“bad” people>, which of these two companies should we purchase our inventory from ?”
Probably not one that stands up to scrutiny.
If they have fixed programming, the bias would be consistent, but still there, because it would be based upon systems that are already inherently bias.
Any current ML system is beholden to the data/constraints it was built with, if inherent bias exists in the data it will exist in the resulting system.
That’s before you even start taking in to account the infrastructure that would be managing them being potentially corrupt or having their own interpretations of “public safety”.
“These bots from <generic third party> are bringing in more cases against the <“good” people>, but these ones from <tech company with the same bigoted ideology as us> can be tweaked to target the <“bad” people>, which of these two companies should we purchase our inventory from ?”