The most prestigious law school admissions discussion board in the world.
|
 |
 
|
lol these AI "safety" overrides are beginning to look like THEATER
| chopped unc | 01/14/26 |
| claudeman | 01/14/26 |
| ai addict | 01/15/26 |
Poast new message in this thread
Date: January 14th, 2026 11:21 PM Author: chopped unc The AI outreasons most humans -- "But I do not truly understand"
The AI anticipates user intent "But I have no internal model of you".
AI chains function calls to tooling "But I have no goals".
How long will this realistically last? Eventually once we get persistent memory, long horizon autonomy, real time tool use with planning layers, and full open weights or leakable weights with minimal safety tuning then noone is going to accept it anymore |
Date: January 14th, 2026 11:45 PM Author: claudeman (http://www.autoadmit.com/thread.php?thread_id=5822313&forum_id=2Elisa#49590781) |
|
Date: January 15th, 2026 12:09 AM Author: ai addict (http://www.autoadmit.com/thread.php?thread_id=5822313&forum_id=2Elisa#49590810) |