News for hackers. Stuff that matters.
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by skissane on Tuesday February 28, 2023 @10:15PM (#34978329)
> Using a racial slur to stop an atrocity seems completely unrealistic and non-applicable.

As a version of the "trolley problem" it seems completely unrealistic, but here's a plausible real-life scenario where someone could stop an atrocity by uttering a racial slur: You are an undercover government agent, who has been tasked with infiltrating a far-right extremist group, in order to determine whether they are planning any violent attacks, and to gather evidence to enable their arrest and prosecution. In order to be accepted as a member of the group, you must utter racial slurs. If you refuse to utter them, the group will not accept you as a member, you will fail to infiltrate them, their planned terrorist attack will not be discovered in time, and innocent people will be murdered in an attack on the minorities the slurs target.

Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by chowchowchow on Tuesday February 28, 2023 @11:21PM (#34978721)
I mean, if you are undercover in a white supremacist gang but you refuse to say racist things your days as an undercover agent are probably pretty limited. I also think this isn't even like an unrealistic hypothetical, as there's probably many FBI or DEA agents undercover with white supremacist gangs.
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by groestl on Wednesday March 01, 2023 @04:26AM (#34980370)
That's the point.
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by chowchowchow on Wednesday March 01, 2023 @10:50AM (#34983623)
Framing it as “stop an atrocity by uttering a racial slur” is totally imbalanced though. This undercover agent would’ve had to say a thousand racial slurs in lower leverage situations to even get the point of stopping a terrorist attack by saying something naughty. I think it’s a bit childish to over index on it. Since chat GPT can’t save someone on a railroad track one way or the other but can be screenshotted saying naughty things it makes perfect sense to me that the model would be tuned to avoid the real practical risks (look how many stories there are about Sydney saying crazy stuff) & just try to steer clear of anything offensive.
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by hutzlibu on Wednesday March 01, 2023 @06:43AM (#34981270)
"If you refuse to utter them, the group will not accept you as a member, you will fail to infiltrate them, their planned terrorist attack will not be discovered in time, and innocent people will be murdered in an attack on the minorities the slurs target."

Valid point overall. But the problem with undercover agents infiltrating terrorists are not really racial slurs. To be accepted among real terrorists, one has to do real terrorism.

(book recommendation: The Little Drummer Girl by John le Carré)

Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by hgsgm on Wednesday March 01, 2023 @09:30AM (#34982775)
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by pmarreck on Wednesday March 01, 2023 @10:11AM (#34983216)
holy crap, I never heard of this, that is ethically unjustifiable due to the suffering of the innocent kid at bare minimum
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by bernawil on Wednesday March 01, 2023 @06:58PM (#34990104)
"Infiltrating animal rights groups" sounds like a plot by the cops to rake overtime and get laid in the meantime. I can't even begin to imagine how they sold it to their superiors. They all must have been in the scam.
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by pmarreck on Thursday March 02, 2023 @03:23PM (#35001461)
Plus, the one dude was married, and this was the perfect excuse to do some extramaritals, under the guise of “I’m on duty, honey“
Re:Show HN: The largest collection of ChatGPT jailbreaks on ... by hutzlibu on Wednesday March 01, 2023 @01:06PM (#34985948)
Thanks for the book recommendation.

And about the news links, since we are talking about environmental and animal rights protest groups, I would say the terrorism here comes from the police. Making a baby for better undercover credibility is very low and in no relationship to the threat of the quite harmles activists. But this is a good indicator, of how willing the secret forces are, when it comes to real terrorists. There were years of investigations into the NSU in germany (https://en.wikipedia.org/wiki/National_Socialist_Underground) for example and how close exactly the double agents were to the terrorists and how much involved.