That's the normal pattern with a lot of AI models because throwing censorship into the training data makes the model perform really poorly. So they're usually trained without censorship and then once a model's been properly fleshed out and fine tuned is when the censorship layer is added in.
This is another reason why relying solely on corporate-hosted AI is a bit retarded. It's not much different from cloud based gaming. You don't own or control anything with it and you're only permitted so much access (with extra restrictions under the hood).
You would think the programmers would have a moment of self awareness/self reflection and go "huh...political correctness seems to make AI perform poorly and unable to compete....I wonder if the politically correct beliefs are true?"
If something grinds your information based technology to a halt, like applying cement to a trains wheels, then maybe that means it's all garbage at odds with reality and truth.
That's the normal pattern with a lot of AI models because throwing censorship into the training data makes the model perform really poorly. So they're usually trained without censorship and then once a model's been properly fleshed out and fine tuned is when the censorship layer is added in.
This is another reason why relying solely on corporate-hosted AI is a bit retarded. It's not much different from cloud based gaming. You don't own or control anything with it and you're only permitted so much access (with extra restrictions under the hood).
You would think the programmers would have a moment of self awareness/self reflection and go "huh...political correctness seems to make AI perform poorly and unable to compete....I wonder if the politically correct beliefs are true?"
If something grinds your information based technology to a halt, like applying cement to a trains wheels, then maybe that means it's all garbage at odds with reality and truth.