洪 民憙 (Hong Minhee)@lemmy.ml to Open Source@lemmy.mlEnglish · 2 days agoHistomat of F/OSS: We should reclaim LLMs, not reject themwritings.hongminhee.orgexternal-linkmessage-square65fedilinkarrow-up192cross-posted to: foss@beehaw.orgopensource@programming.dev
arrow-up192external-linkHistomat of F/OSS: We should reclaim LLMs, not reject themwritings.hongminhee.org洪 民憙 (Hong Minhee)@lemmy.ml to Open Source@lemmy.mlEnglish · 2 days agomessage-square65fedilinkcross-posted to: foss@beehaw.orgopensource@programming.dev
minus-squarebizdelnick@lemmy.mllinkfedilinkarrow-up1·2 days agoSo what is the case we are speaking about? “Hey LLM, write the OS kernel that is fully compatible with Linux, designed like Linux, uses the same algorithms as Linux and the same code style as Linux”?
minus-squarechgxvjh [he/him, comrade/them]@hexbear.netlinkfedilinkEnglisharrow-up1·2 days agoIf you have Linux in the training data, the outcome if at all remotely useful would likely include plagiarism.
minus-squarechgxvjh [he/him, comrade/them]@hexbear.netlinkfedilinkEnglisharrow-up1·2 days agoThere are cases in the wild of LLMs straight up pasting the GPL into files unprompted.
So what is the case we are speaking about? “Hey LLM, write the OS kernel that is fully compatible with Linux, designed like Linux, uses the same algorithms as Linux and the same code style as Linux”?
If you have Linux in the training data, the outcome if at all remotely useful would likely include plagiarism.
Are there similar cases in the wild?
There are cases in the wild of LLMs straight up pasting the GPL into files unprompted.