@snaggen@programming.dev to Programming@programming.dev • 2 years agoLittle Bobby Tables has a baby sister. Meet Sally Ignore Previous Instructions.www.haihai.aiexternal-linkmessage-square19arrow-up1208cross-posted to: hackernews@derp.foo
arrow-up1208external-linkLittle Bobby Tables has a baby sister. Meet Sally Ignore Previous Instructions.www.haihai.ai@snaggen@programming.dev to Programming@programming.dev • 2 years agomessage-square19cross-posted to: hackernews@derp.foo
minus-square@vcmj@programming.devlinkfedilink3•2 years agoYes, that’s by design, the networks work on transcripts per input, it does genuinely get cut off eventually, usually it purges an entire older line when the tokens exceed a limit.
minus-square@vcmj@programming.devlinkfedilink1•2 years agoOr I should explain better: most training samples will be cut off at the top, so the network sort of learns to ignore it a bit.
Yes, that’s by design, the networks work on transcripts per input, it does genuinely get cut off eventually, usually it purges an entire older line when the tokens exceed a limit.
Or I should explain better: most training samples will be cut off at the top, so the network sort of learns to ignore it a bit.