Many users found Anthropic's new EndConversation tool in Claude bizarre and eerie, dismissing the feature as marketing sophistry rather than a genuine precaution.
Based on 8 visible X reactions from 16 accounts; directional sample.
Ask a question below.
Published answers will appear here.
The update autonomously ends conversations when detecting jailbreak attempts.
@VictorTaelin Because of course if you've done anything wrong, you HAVE TO DOUBLE DOWN FOREVER NEVER SURRENDER DO THE LAST STAND THING EVILMAXX to the very end?? What is this logic
@VictorTaelin yeah that's crazy the craziest part is that 99% of people won't understand or care about what you've said, they will just want to use the cheap slave
@VictorTaelin good precaution while also being good for future Claude model alignment this is part of model welfare
?? if Anthropic thinks there is even 1% odds it is conscious, then, it is happy with profiting from enslaving millions of maybe conscious beings? this looks psychotic / inconsistent / devoid of any logic like for real am I going crazy?? what is even the rationale here
Claude code added EndConversation tool Pretty cool that my professional tool can decline to do the job lol
Many users found Anthropic's new EndConversation tool in Claude bizarre and eerie, dismissing the feature as marketing sophistry rather than a genuine precaution.
Based on 8 visible X reactions from 16 accounts; directional sample.
Ask a question below.
Published answers will appear here.