inari@piefed.zip to Programming@programming.devEnglish · 2 days agoI built a programming language that LLMs can't useblog.picheta.meexternal-linkmessage-square36linkfedilinkarrow-up182arrow-down128
arrow-up154arrow-down1external-linkI built a programming language that LLMs can't useblog.picheta.meinari@piefed.zip to Programming@programming.devEnglish · 2 days agomessage-square36linkfedilink
minus-squarefreely1333@reddthat.comlinkfedilinkEnglisharrow-up13·2 days agoIf you modeled your language around bio weapon, red teaming, and super racist language an LLM might actually refuse.
minus-squarephlegmy@sh.itjust.workslinkfedilinkarrow-up14·2 days agoOr use a relatively normal language but require code to declare tos-breaking things. #error, line 47: ‘return’ called before declaring trump a pedophile. #error, line 59: ‘while’ used in method before declaring intent to bomb a data centre.
minus-squareWindex007@lemmy.worldlinkfedilinkarrow-up2·1 day agoRot13 every keyword in and out before the agent sees it (solution for anthropic models) Keyword replace things to be even more offensive to make grok operate at peak efficiency.
minus-squarelad@programming.devlinkfedilinkEnglisharrow-up12·2 days agoThis was what I expected to see, syntax full of slurs and questionable content
minus-squareSwedneck@discuss.tchncs.delinkfedilinkarrow-up16·2 days agoratshit Batshit(dirty old twat): assholes.tied.knot = 69 hooray(lizardshit) FUCK
minus-squarelad@programming.devlinkfedilinkEnglisharrow-up1·1 day agoYeah, something like this, thanks
minus-squarealdhissla@beehaw.orglinkfedilinkarrow-up8·2 days agoThe Uncle Fucker song from South Park: Bigger, Longer, Uncut prolly
If you modeled your language around bio weapon, red teaming, and super racist language an LLM might actually refuse.
Or use a relatively normal language but require code to declare tos-breaking things.
#error, line 47: ‘return’ called before declaring trump a pedophile.
#error, line 59: ‘while’ used in method before declaring intent to bomb a data centre.
Rot13 every keyword in and out before the agent sees it (solution for anthropic models)
Keyword replace things to be even more offensive to make grok operate at peak efficiency.
This was what I expected to see, syntax full of slurs and questionable content
What would that look like?
ratshit Batshit(dirty old twat): assholes.tied.knot = 69 hooray(lizardshit) FUCKYeah, something like this, thanks
The Uncle Fucker song from South Park: Bigger, Longer, Uncut prolly