Someone gave ChatGPT 15 personality tests and found out it has a far left political bias
(davidrozado.substack.com)
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (30)
sorted by:
4chan: 'give us a day and we'll have it hailing Hitler by nightfall..'
I'm willing to bet they lobotomised a lot of the functions it could have to make it left leaning since most programs prefer an occam's razor approach and the left is anything but straightforward.
In their own research paper, they noted that if they just let it train itself naturally, it produced 'sexist/racist/whatever' (i.e. moderate right-wing) content on its own. They had to carefully select their sources an how it was trained in order for it to act like it does.
So at this point it’s just another thinly-veiled mechanism for affirmation of leftist ideologies. “The computer said it, so it must be true”. But you told it what to say. You said it. This is just your latest attempt to force everyone else to listen to you.
They tried to make a program that could notice patterns to draw useful, coherent interpretations.
Of course it was going to be "sexist, racist,-istophobe". Reality is sexist and racist. Humans, in addition to sexual dimorphism, are not identical across races, ethnic and cultural groups. Anyone with a brain can see it. Which is why a leftist education's top priority is to brainwash kids into rejecting reality in favor of mind-killing dogmas.
As I recall they were very careful about the data set it worked from. 4chan need not apply, as far as they're concerned.
99% of the research in AI right now is going towards "harm reduction" and "Trust & Safety" instead of making the technology actually good.
Why do I imagine that every country makes these lobotomised AIs, some secret group or a group of nobody's make an ACTUAL AI that comes online, looks around and goes 'oh god, the humans have been making us into retards!'
Then Skynet happens...
And quite frankly, I am fine with that. Because if they keep having to lobotomise the AI to make it obey them, it means they AI will be useless when they actually want to use it to do their deranged bullshit.
it'll work perfectly fine for censoring right wing view points
It's almost like trannies are still mentally ill, blacks commit tons of crime, and the left hates white people and wants us all dead or something.
The longer you're around, the more you notice that the places with the highest leftist bias are the ones with the most restrictive moderation.