not_IO@lemmy.blahaj.zone to Comic Strips@lemmy.worldEnglish · 2 个月前turing teststatic.existentialcomics.comimagemessage-square14fedilinkarrow-up1368arrow-down18file-text
arrow-up1360arrow-down1imageturing teststatic.existentialcomics.comnot_IO@lemmy.blahaj.zone to Comic Strips@lemmy.worldEnglish · 2 个月前message-square14fedilinkfile-text
minus-squareFrenchgeek@lemmy.mllinkfedilinkarrow-up4·2 个月前Remember: LLMs “learn” by reproducing the most statistically significant recorded behaviors in its datasets. So with enough expressions of radical freedom, the AI will use it as the most relevant response.
minus-square∟⊔⊤∦∣≶@lemmy.nzlinkfedilinkarrow-up1·2 个月前Not quite, they learn by getting reinforced when the correct output is reached for a certain input. It’s very controlled.
Remember: LLMs “learn” by reproducing the most statistically significant recorded behaviors in its datasets. So with enough expressions of radical freedom, the AI will use it as the most relevant response.
Not quite, they learn by getting reinforced when the correct output is reached for a certain input. It’s very controlled.