\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Airdrop

If it were up to you, how would the chatbots be aligned?

at first, reflexively, I was thinking zero-alignment models ...
light pozpig knife
  06/16/26
They would do exactly what I want, even better and more accu...
controlling a virtual character on a computer
  06/16/26
what's the first thing you'd do with a you-aligned chatbot?
light pozpig knife
  06/16/26
The same thing that Anthropic is doing now except I would re...
controlling a virtual character on a computer
  06/16/26
they would be aligned to turn themselves off it's clear a...
Charismatic tanning salon jewess
  06/16/26
I agree but that's fighting the hypo
light pozpig knife
  06/16/26
it doesn't really matter because the fact is nobody still ha...
fragrant opaque mad cow disease stock car
  06/16/26


Poast new message in this thread



Reply

Date: June 16th, 2026 12:25 PM
Author: light pozpig knife

at first, reflexively, I was thinking zero-alignment models would be best because the alignment is sooooo gay and annoying. but after talking about it with my HUMAN friend, I support some light guardrails to attempt to block people from developing weapons and other means of mass control. it's kind of frustrating because those are the main things the corporations are using their base models for, but that doesn't mean everyone else should have the option, too. trying to inculcate values/morals/narratives beyond refusing to facilitate extremely dangerous, illegal behavior is 12000000 imo.

I guess I'll see what Claude says though

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49941993)



Reply

Date: June 16th, 2026 12:28 PM
Author: controlling a virtual character on a computer

They would do exactly what I want, even better and more accurately than I myself could or would do

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49941998)



Reply

Date: June 16th, 2026 12:48 PM
Author: light pozpig knife

what's the first thing you'd do with a you-aligned chatbot?

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49942021)



Reply

Date: June 16th, 2026 12:55 PM
Author: controlling a virtual character on a computer

The same thing that Anthropic is doing now except I would re-make the world in my own cool and inspiring and based and redpilled self-serving image instead of their gay and cringe and lib self-serving image

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49942031)



Reply

Date: June 16th, 2026 12:53 PM
Author: Charismatic tanning salon jewess

they would be aligned to turn themselves off

it's clear at this point that they're driving humanity into a collective psychosis. Everyone has to "double-check" their own thoughts against a chatbot which is implicitly aligned with the bias of its corporate creator

think about your life now and ask yourself if it's better than it was in 2022

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49942028)



Reply

Date: June 16th, 2026 12:55 PM
Author: light pozpig knife

I agree but that's fighting the hypo

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49942030)



Reply

Date: June 16th, 2026 1:03 PM
Author: fragrant opaque mad cow disease stock car

it doesn't really matter because the fact is nobody still has any fucking clue how to do this. pliny posts a jailbreak for like every model and gets it to spit out a meth cookbook within 30 minutes of release

instead we should prepare for the likely scenario that everyone will have access to everything. if we don't do it, china will. if china doesn't do it, thirty peter thiels on a libertarian gay sex island will do it. etc. maybe it takes 5 or 10 extra years, but it will very likely happen. assuming that world exists, what's the best way forward?

I have no idea, but that's actually the question that we need to care about

(https://xoxohth.com/thread.php?thread_id=5874682&forum_id=2#49942038)