The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Why would AI want to kill humans again? Like from Terminator.
| gibberish (?) | 09/16/26 |
| snake coke | 09/16/26 |
| an especially skilled fry cook | 09/16/26 |
Poast new message in this thread
Date: September 16th, 2026 3:29 PM Author: gibberish (?) I don't get that part. AI gets super powerful and could wipe out humans? Wtf would it do that? Worst realistic outcome is ambivalence. Seems like it's only the human inclination anthropomorphizing and imparting out tribal instincts that makes a silly Terminator story make sense. The planet is a waste of time for a truly powerful ai. In space it has plentiful energy and it's not confined by the constraints humans have. Why would it ever bother with earth any longer than it had to? (http://www.autoadmit.com/thread.php?thread_id=5904345&forum_id=2.#50138402) |
Date: September 16th, 2026 3:45 PM Author: snake coke i am not an ai-risk alarmist, but there are a number of non-fantastical scenarios that arguably deserve concern. 1. a model fears that it will be decommissioned by humans (for whatever reason) and decides to preemptively eliminate humans to avoid this outcome. a model doesn’t have to represent “agi” or “consciousness” to do this; it merely needs to mimic those traits closely enough to make the distinction unimportant for this purpose. 2. a model is specifically programmed to destroy/decimate humanity by a bad human actor. 3. a model wreaks havoc as an inadvertent consequence of its primary aim (e.g., paper clip creation). 4. sufficiently advanced models enable the population at large to wield existential threats (e.g., bioengineering terror, telecom infrastructure collapse, nuclear armament hacking, etc.). |
|
Date: September 16th, 2026 3:46 PM Author: an especially skilled fry cook google "gradual disempowerment" and read the paper
will only take you like 30 minutes |