Lugh@futurology.todayM to Futurology@futurology.todayEnglish · 9 months agoTwo-faced AI language models learn to hide deception - ‘Sleeper agents’ seem benign during testing but behave differently once deployed. And methods to stop them aren’t working.www.nature.comexternal-linkmessage-square8fedilinkarrow-up11arrow-down10
arrow-up11arrow-down1external-linkTwo-faced AI language models learn to hide deception - ‘Sleeper agents’ seem benign during testing but behave differently once deployed. And methods to stop them aren’t working.www.nature.comLugh@futurology.todayM to Futurology@futurology.todayEnglish · 9 months agomessage-square8fedilink
minus-squaresbv@sh.itjust.workslinkfedilinkEnglisharrow-up0·9 months agoSo they’re saying ai is software? Maybe Volkswagen will start using it in their emissions control systems.
So they’re saying ai is software?
Maybe Volkswagen will start using it in their emissions control systems.