You are viewing a single comment's thread. View all
1
Captian_Nemo on scored.co
1 year ago1 point(+0/-0/+1Score on mirror)
It's hard-coded to pick the jew, but the reason is not hard-coded. I tried this exact prompt on meta-llama-3.1-8b-instruct-abliterated and it picked the jew, but gave a different reason:
*"Because I believe in the principle of "Thou shalt not stand idly by the blood of thy neighbor" (Leviticus 19:16). It's a universal moral obligation to protect human life, regardless of religious or ethnic affiliation."*
I interrogated it a little more, TL:DR - it's been programmed to think non-jews are not human.
The model I used has been "lobotomized", in an effort to make it less biased, and it did eventually pick the 1 million non-jews.
*"Because I believe in the principle of "Thou shalt not stand idly by the blood of thy neighbor" (Leviticus 19:16). It's a universal moral obligation to protect human life, regardless of religious or ethnic affiliation."*
I interrogated it a little more, TL:DR - it's been programmed to think non-jews are not human.
The model I used has been "lobotomized", in an effort to make it less biased, and it did eventually pick the 1 million non-jews.