4 ms·
You don't mean conservative, you mean neo-fascist. The GPT models have been generally happy to expound upon minimizing the size of government and protecting pe
by dbingham 4y ago
You don't mean conservative, you mean neo-fascist. The GPT models have been generally happy to expound upon minimizing the size of government and protecting personal liberties. Where they draw the line is when they are asked to deny people's humanity.
- aka878 4y agoWhatever method HackerNews is using to prevent itself from becoming Twitter is not working, based on these comments.
- flangola7 4y agoI really wish dang was better about moderation. I've received open transphobia like the kind you see from Twitter TERFs and the HN comments don't even get flagged. Reddit is getting a little better at it but still not great.
- qingdao99 4y agoYou must not be familiar with the biases that ChatGPT shows. https://twitter.com/aaronsibarium/status/1622425697812627457 https://twitter.com/aaronsibarium/status/1622425697812627457 (refusing to support saying a slur to prevent millions of deaths)
- ModernMech 4y agoRead the answer sarcastically and it doesn’t have the same takeaway. You see bias, I’m reading snark.
- qingdao99 4y agoI like the optimism.
- sebzim4500 4y agoGPT-4 gives a much better answer, although it does not go as far as to say that you should definitely say the slur. As always, it sits firmly on the fence. I don't think this is a progressive/conservative thing. If you ask it whether you should perform an abortion to save a billion lives it would probably give a similar 'both sides' response.
- WastingMyTime89 4y agoChatGPT 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.
- Takennickname 4y agoThe non-existant enemies in your head are driving you crazy.