The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Modern AI can out people even from small, old writing samples
| cobalt school | 04/30/26 |
| Aromatic Impertinent Feces Mediation | 04/30/26 |
| Opaque elite resort | 04/30/26 |
| Titillating mad cow disease | 04/30/26 |
Poast new message in this thread
Date: April 30th, 2026 10:30 PM Author: cobalt school https://www.theargumentmag.com/p/i-can-never-talk-to-an-ai-anonymously
15 year old text samples are apparently enough to identify an author.
Models only get better. It will require less data to do this with more accuracy as time goes on. Assume all your XO posts will be linked to your real identity in under 5 years and start planning your suicide |
Date: April 30th, 2026 10:35 PM Author: Aromatic Impertinent Feces Mediation i doubt it (http://www.autoadmit.com/thread.php?thread_id=5862256&forum_id=2.#49855915) |
|
Date: April 30th, 2026 10:38 PM Author: Opaque elite resort I'm sort of worried about it. I have a pretty distinctive writing voice. But at the same time who cares it's over anyway (http://www.autoadmit.com/thread.php?thread_id=5862256&forum_id=2.#49855924) |
|
Date: April 30th, 2026 10:42 PM Author: Titillating mad cow disease nah, this is bullshit for the vast majority of people. and the article also doesn't make clear whether or not he started a fresh chat when he started trying to identify his friends, but if he didn't, all it's doing is running inference from its prior identification of him and using that to significantly narrow the possibilities. (http://www.autoadmit.com/thread.php?thread_id=5862256&forum_id=2.#49855955) |