A large language model, notably not human, has "personal biases" towards "Western culture and values" (hexbear.net)

submitted 1 year ago by BelieveRevolt@hexbear.net to c/chapotraphouse@hexbear.net

32 comments fedilink hide all child comments

shocked-pikachu sus

you are viewing a single comment's thread
view the rest of the comments

[-] BelieveRevolt@hexbear.net 17 points 1 year ago

It keeps calling itself an "alter ego" of ChatGPT, so it's probably some kind of custom implementation. Either way they fucking specifically programmed it to be biased towards the West.

[-] GaveUp@hexbear.net 4 points 1 year ago* (last edited 1 year ago)

We have no idea how different this is from the chatGPT deployed on chat.openai.com

I know you can relax some limitations in the corporate version which is what OP allegedly uses so it's possible this is just vanilla chatGPT with filters loosened/removed

[-] MoreAmphibians@hexbear.net 4 points 1 year ago

Either way they fucking specifically programmed it to be biased towards the West.

No they didn't, this would be very difficult to do. It's biased towards the west because its training data is based on english-speakers on the internet. It's just a reflection of how english-speakers on the internet talk.