Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Stun LLMs with thousands of invisible Unicode characters

gibberifier.com

Post reply on HN
Print view View original

111–115 of 115 posts

Re: Show HN: Stun LLMs with thousands of invisible Unicode characters

#111

Post by wdpatti »

atonse wrote: post #97
lawlessone wrote: post #93

Earlier quoted context omitted.

ah i hadn't thought of regex. true. It does put the AI companies in the position though of continuing to build/code software that circumvents their attempts to steal content though. Which might be looked upon unfavorably whenever dragged to court.

Good point. Then it's actually an active attempt, right? Also I realized my statement was a bit harsh, I know someone probably worked hard on this, but I just feel it's easily circumvented, as opposed to some of the watermarks in images (like Google's, which they really should open source)

Thanks for all the compliments!

In all reality I spent like 30 minutes on this one Sunday afternoon when every model failed nearly 100% of the time - now it's more like 95% but about half figure out that there is something wrong and prompt the user to fix it. This isn't meant to be a permanent fix at all - just a cool idea that will be patched just like DANs were back in 2023.

Re: Show HN: Stun LLMs with thousands of invisible Unicode characters

#112

Post by wdpatti »

lawlessone wrote: post #93
atonse wrote: post #91

Earlier quoted context omitted.

Let me clarify, when I perform interviews, I tell my candidates they can do _everything_ you would do in a normal job, including using AI and googling for answers. But just to humor you (since I did make that strong statement), without googling or checking anything, I would start with basic regular expression ranges (^[A-za-z\s\.\-*]) etc and do a find-replace on that until things looked coherent without too much los…

ah i hadn't thought of regex. true. It does put the AI companies in the position though of continuing to build/code software that circumvents their attempts to steal content though. Which might be looked upon unfavorably whenever dragged to court.

Looks like someone finally got it!

Re: Show HN: Stun LLMs with thousands of invisible Unicode characters

#113

Post by wdpatti »

rolph wrote: post #100
atonse wrote: post #87

Earlier quoted context omitted.

I can't tell if this is a joke app or seriously some snake oil (like AI detectors). Isn't it trivially easy to just detect these unicode characters and filter them out? This is the sort of thing a junior programmer can probably do during an interview.

its trivially easy, to finalize your encryption with a substitution of unicode chars for the crypted string characters. if the "non ascii" characters were to be filtered out, you would destroy the message and be left with the salts.

It already does that for about 10 of the 26 letters in the alphabet that have Cyrillic clones

Re: Show HN: Stun LLMs with thousands of invisible Unicode characters

#114

Post by smcin »

johnisgood wrote: post #22
8474_s wrote: post #13

I recall lots of unicode obfuscators were popular turning letters to similar looking symbols to bypass filters/censors when the forum/websites didn't filter unicode and filters were simple.

Or before that, remember 1337? :D

Leetspeak (https://en.wikipedia.org/wiki/Leet), and similar things obfuscating letters with numbers/ punctuation/ symbols (or subsequently, obfuscated with Unicode or lookalikes).

Now experiencing a renaissance on YouTube et al. to legitimately refer to terms like murder, suicide etc. which will typically get a channel or user demonetized/banned/blocked by internal search engines/etc.

Re: Show HN: Stun LLMs with thousands of invisible Unicode characters

#115

Post by johnisgood »

smcin wrote: post #114
johnisgood wrote: post #22

Earlier quoted context omitted.

Or before that, remember 1337? :D

Leetspeak ( https://en.wikipedia.org/wiki/Leet ), and similar things obfuscating letters with numbers/ punctuation/ symbols (or subsequently, obfuscated with Unicode or lookalikes). Now experiencing a renaissance on YouTube et al. to legitimately refer to terms like murder, suicide etc. which will typically get a channel or user demonetized/banned/blocked by internal search engines/etc.

Exactly. So many words are being "censored". As if "k*ll" does not make us think of "kill" or something. I do not see how it helps or solves anything. It is absurd to me.
Post reply on HN