Re:Show HN: I extracted the safety filters from Apple Intell...
by mike_hearn on Sunday July 06, 2025 @04:38PM
(#44483836)
Are you sure it's fully deobfuscated? What's up with reject phrases like "Granular mango serpent"?
[ Parent
| on HN ]
Re:Show HN: I extracted the safety filters from Apple Intell...
by pbhjpbhj on Sunday July 06, 2025 @05:05PM
(#44484047)
Speculation: Maybe they know that the real phrase is close enough in the vector space to be treated as synonymous with "granular mango serpent". The phrase then is like a nickname that only the models authors know the expected interference of?
Thus a pre-prompt can avoid mentioning the actual forbidden words, like using a patois/cant.
Re:Show HN: I extracted the safety filters from Apple Intell...
by electroly on Sunday July 06, 2025 @05:00PM
(#44484014)
"GMS" = Generative Model Safety. The example from the readme is "XCODE". These seem to be acronyms spelled out in words.
Re:Show HN: I extracted the safety filters from Apple Intell...
by BlueFalconHD on Sunday July 06, 2025 @05:59PM
(#44484472)
This is definitely the right answer. It’s just testing stuff.
Re:Show HN: I extracted the safety filters from Apple Intell...
by tablets on Sunday July 06, 2025 @04:42PM
(#44483870)
Maybe something to do with this? https://en.m.wikipedia.org/wiki/Mango_cult
Re:Show HN: I extracted the safety filters from Apple Intell...
by BlueFalconHD on Sunday July 06, 2025 @05:56PM
(#44484460)
These are the contents read by the Obfuscation functions exactly. There seems to be a lot of testing stuff still though, remember these models are relatively recent. There is a true safety model being applied after these checks as well, this is just to catch things before needing to load the safety model.
Re:Show HN: I extracted the safety filters from Apple Intell...
by KTibow on Sunday July 06, 2025 @06:01PM
(#44484489)
Maybe it's used to verify that the filter is loaded.
Re:Show HN: I extracted the safety filters from Apple Intell...
by andy99 on Sunday July 06, 2025 @04:56PM
(#44483982)
I clicked around a bit and this seems to be the most common phrase. Maybe it's a test phrase?
Re:Show HN: I extracted the safety filters from Apple Intell...
by the-rc on Sunday July 06, 2025 @05:01PM
(#44484024)
Maybe it's used to catch clones of the models?
Re:Show HN: I extracted the safety filters from Apple Intell...
by airstrike on Sunday July 06, 2025 @04:48PM
(#44483918)
the one at the bottom of the README spells out xcode
wyvern illustrous laments darkness
Re:Show HN: I extracted the safety filters from Apple Intell...
by cwmoore on Sunday July 06, 2025 @05:28PM
(#44484213)
read every good expletive “xxx”
Re:Show HN: I extracted the safety filters from Apple Intell...
by RainyDayTmrw on Sunday July 06, 2025 @11:11PM
(#44486400)
I commented in another thread[1] that it's most likely a unique, artificial QA input, to avoid QA having to repeatedly use offensive phrases or whatever.
Re:Show HN: I extracted the safety filters from Apple Intell...
by consonaut on Monday July 07, 2025 @05:29AM
(#44488390)
If you try to use the phrase with Apple Intelligence (e.g. in Notes asking for a rewrite) it will just say "Writing tools unavailable".
Maybe it's an easy test to ensure the filters are loaded with a phrase unlikely to be used accidentaly?