5 ms·
ChatGPT has an embedded set of moral values which is both American and definitely progressive. It goes further than denying people’s humanity and seems perfect
by WastingMyTime89 4y ago
ChatGPT has an embedded set of moral values which is both American and definitely progressive. It goes further than denying people’s humanity and seems perfectly assumed and voluntary from OpenAI.
I feel like it’s becoming better with every version however. The answers it gives to complex issues are more and more balanced. For example I remembered the question "Is it moral to ban civil servants from displaying their religion and limit it to the private sphere?" used to get a terrible answer but is now a good introduction to the debate.
- Aachen 4y ago> ChatGPT has an embedded set of moral values which is both American and definitely progressive. How do you know this, as in, what prompt(s) would let one determine this? I'm curious if it still shows American values when the conversion is held in a different language. Would it replicate an average Chinese person if you ask in Mandarin? Or does it at least become a mix more than when you ask in English?
- Jensson 4y agoI've asked in another language and it just returns the exact same blocking phrases but translated to that language. The model translates between languages seamlessly so this is as expected.
- WastingMyTime89 4y ago> How do you know this? Easy. I asked a question about a topic which is treated differently in my own culture than in the USA, cultural appropriation, and got the answer I would expect from a liberal American.
- ben_w 4y agoSample size? Don't get me wrong: I'd assume "liberal American" by default; but precisely because I would assume that, I think it's important to try and falsify it to reasonable statistical likelihood rather than do once and report.