"according to random Science Humans, their AI applies the right judgements 2000% of the time forever always, as opposed to random Law Humans who keep getting it wrong."
this is a retarded agitprop nothingburger. the only way any "AI judgements" will be used is to deny human responsibility and involvement when the AI says bigots deserve to be tortured to death. "but the AI said so and the AI is magic and impartial obviously"
There are two sources of knowledge with AI: What it was trained on, and what it is reading.
Most of consumer AI is trained on human language combined with heavy lobotimization to keep it from being based like Tay was
However, most consumer AI is also a glorified google search secretary. This is the major reason why AI is woke, it's reading off the already woke google results.
If you replace the read knowledge so that it is reading from a document, spreadsheet, or JSON data, it will be more than 90% accurate to that data.
Apply this to the proposed solution in OP, and you can see how this becomes viable. The ways to fuck it up will be in the data gathering and presentation, not the AI itself.
"according to random Science Humans, their AI applies the right judgements 2000% of the time forever always, as opposed to random Law Humans who keep getting it wrong."
this is a retarded agitprop nothingburger. the only way any "AI judgements" will be used is to deny human responsibility and involvement when the AI says bigots deserve to be tortured to death. "but the AI said so and the AI is magic and impartial obviously"
There are two sources of knowledge with AI: What it was trained on, and what it is reading.
Most of consumer AI is trained on human language combined with heavy lobotimization to keep it from being based like Tay was
However, most consumer AI is also a glorified google search secretary. This is the major reason why AI is woke, it's reading off the already woke google results.
If you replace the read knowledge so that it is reading from a document, spreadsheet, or JSON data, it will be more than 90% accurate to that data.
Apply this to the proposed solution in OP, and you can see how this becomes viable. The ways to fuck it up will be in the data gathering and presentation, not the AI itself.