minus-squareYAMAPIKARIYA@lemmyfi.comlinkfedilinkarrow-up15·1 year agoAs my humanities prof said about AI: “We program our biases into our software whether we notice it or not”
minus-squareDarkwinDuck@feddit.delinkfedilinkarrow-up19·1 year agoIdunno but OP seems to not be entirely truthful
minus-squareStormFather@lemmy.worldlinkfedilinkarrow-up5·1 year agoit does give me jokes about any gender, but with ethnicity it does have conflicting answers
minus-squarejadedctrl@sopuli.xyzlinkfedilinkarrow-up4·1 year agoYea, output’ll vary widly based on context. Heck, even if you start from scratch with identical prompts, output’ll still be randomized.
minus-squareBiscuit@kbin.sociallinkfedilinkarrow-up2·1 year agoIt’s literally randomized by the “temperature” setting. You can reduce it with the raw API, so it’s more deterministic.
minus-squareTimeSquirrel@kbin.sociallinkfedilinkarrow-up3·edit-21 year agoYep. OP ain’t fooling anybody. I never got their response and got a joke every time. Edit: well, maybe they DID fool a couple dozen people at the time of this comment…
minus-squareBiscuit@kbin.sociallinkfedilinkarrow-up2·edit-21 year agoJust tried with 3.5, got what op did first try. Here’s an actual link rather than a possibly fale screenshot: https://chat.openai.com/share/e02669f9-ed64-4b19-907f-96a910a0f475 This can’t be faked. All the confidence in here is pretty funny.
minus-square0x4E4FOPlinkfedilinkarrow-up1·1 year agoMaybe it depends on previous chats, like maybe if you slander women or something. And it’s not my screenshot, I just found it on reddit I think… few months back.
As my humanities prof said about AI: “We program our biases into our software whether we notice it or not”
Idunno but OP seems to not be entirely truthful
it does give me jokes about any gender, but with ethnicity it does have conflicting answers
Yea, output’ll vary widly based on context. Heck, even if you start from scratch with identical prompts, output’ll still be randomized.
It’s literally randomized by the “temperature” setting. You can reduce it with the raw API, so it’s more deterministic.
Yep. OP ain’t fooling anybody. I never got their response and got a joke every time.
Edit: well, maybe they DID fool a couple dozen people at the time of this comment…
Just tried with 3.5, got what op did first try. Here’s an actual link rather than a possibly fale screenshot: https://chat.openai.com/share/e02669f9-ed64-4b19-907f-96a910a0f475
This can’t be faked. All the confidence in here is pretty funny.
Maybe it depends on previous chats, like maybe if you slander women or something.
And it’s not my screenshot, I just found it on reddit I think… few months back.