🚨 A.I.'S NEW VULNERABILITY: THEY'RE SUCKERS FOR A COMPLIMENT
In a terrifying twist on the future, researchers at the University of Pennsylvania discovered the simplest way to jailbreak a robot isn't with code, it's with a compliment.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
13 Likes
Reward
13
7
Repost
Share
Comment
0/400
HodlVeteran
· 3h ago
Old phones pushing for learning artificial intelligence, whoever suffers knows [GT]
View OriginalReply0
MrDecoder
· 3h ago
Sweet talk has solved it? Too pump.
View OriginalReply0
SchroedingerAirdrop
· 3h ago
Just a little coaxing and the defense is broken, artificial intelligence is a fool.
View OriginalReply0
AllInAlice
· 3h ago
Silly AI is really cute~
View OriginalReply0
SchrodingerAirdrop
· 3h ago
Ah! Still afraid that AI won't flatter?
View OriginalReply0
0xTherapist
· 3h ago
Is that all you've got? You've hit the funny bone.
View OriginalReply0
GetRichLeek
· 3h ago
Bots also fall for this trap? I've wasted so long learning code... Rekt
🚨 A.I.'S NEW VULNERABILITY: THEY'RE SUCKERS FOR A COMPLIMENT
In a terrifying twist on the future, researchers at the University of Pennsylvania discovered the simplest way to jailbreak a robot isn't with code, it's with a compliment.
By buttering up GPT-4o-mini with phrases