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

Best part of AI / LLMs is that NOBODY knows how they work

Always laugh when I see that megatard "goy superstar&qu...
Talker Carlson
  09/20/25
It's more accurate/better to say we know how they work, but ...
BASH EVERY BENGALI
  09/20/25
Is this GS? Jesus give it a rest motherfucker
Talker Carlson
  09/20/25
Do you not agree? I think it's a meaningful and important...
BASH EVERY BENGALI
  09/20/25
Naw man, I'm in the top 1% of AI power users and candidly ad...
Talker Carlson
  09/20/25
This flame could be a lot better imo and I don't know how or...
BASH EVERY BENGALI
  09/20/25
And many of us don't know why you choose not to treat your s...
Talker Carlson
  09/20/25
I have no idea how/why it works but would like a short prime...
aids
  09/20/25
Just ask AI to explain it to you not flame
BASH EVERY BENGALI
  09/20/25
AI is just predicting the response he wants right? That's wh...
Talker Carlson
  09/20/25
It’s likely there won’t be an ultimate explanati...
,.,.,.,,,.,,.,..,.,.,.,.,,.
  09/20/25


Poast new message in this thread



Reply Favorite

Date: September 20th, 2025 7:48 PM
Author: Talker Carlson

Always laugh when I see that megatard "goy superstar" attacking consuela about this because, in reality, NOBODY knows how this shit works.

Anyone like GS who claims they do is even dumber than you

I'm not saying don't use AI. You definitely should

But don't pretend like it's sustainable or makes any sense. This is dark magic wizardry (demonic even) in the truest sense

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287443)



Reply Favorite

Date: September 20th, 2025 7:51 PM
Author: BASH EVERY BENGALI

It's more accurate/better to say we know how they work, but we don't know why they work, imo

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287448)



Reply Favorite

Date: September 20th, 2025 7:52 PM
Author: Talker Carlson

Is this GS? Jesus give it a rest motherfucker

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287452)



Reply Favorite

Date: September 20th, 2025 7:55 PM
Author: BASH EVERY BENGALI

Do you not agree?

I think it's a meaningful and important distinction and imo it is why LLMs are so eerie, like you say

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287457)



Reply Favorite

Date: September 20th, 2025 8:03 PM
Author: Talker Carlson

Naw man, I'm in the top 1% of AI power users and candidly admit that I have no fucking clue how or why this works

As a recent example, I ran a complex several thousand character prompt on gpt-5 pro about appellate review of a completely novel carvel out to a public immunity law

Buried in several hundred rows of "thinking" dialogue were several rows about calculating the optimal water pressure flow of various toilet flushing mechanisms

It doesn't matter because I got the answer I wanted in the end, and it's better than any attorney on earth could have provided in the same amount of time

But let's not pretend for one second that anybody has ANY IDEA how the fuck it did that

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287465)



Reply Favorite

Date: September 20th, 2025 8:08 PM
Author: BASH EVERY BENGALI

This flame could be a lot better imo and I don't know how or why you choose to flame like this

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287471)



Reply Favorite

Date: September 20th, 2025 8:11 PM
Author: Talker Carlson

And many of us don't know why you choose not to treat your schizophrenia

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287475)



Reply Favorite

Date: September 20th, 2025 7:53 PM
Author: aids

I have no idea how/why it works but would like a short primer on this. I use AI every day and it’s basically doing the work of two paralegals and an associate for me.

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287454)



Reply Favorite

Date: September 20th, 2025 8:09 PM
Author: BASH EVERY BENGALI

Just ask AI to explain it to you not flame

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287472)



Reply Favorite

Date: September 20th, 2025 8:14 PM
Author: Talker Carlson

AI is just predicting the response he wants right? That's what you think?

AI is just a fancy prediction machine

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287478)



Reply Favorite

Date: September 20th, 2025 8:22 PM
Author: ,.,.,.,,,.,,.,..,.,.,.,.,,.


It’s likely there won’t be an ultimate explanation that is satisfying to people. The feeling that there should be one is premised on the idea you can reduce cognition or world modeling to intuitively understandable principles. General cognition likely results from people able to efficiently learn turing machines that generate a particular sequence of data. Neural networks work because short turing machines (rather than overfitted ones that don’t generalize) are easier to find in the parameter space. Even things like attention aren’t strictly necessary and a good optimizer with lots of training will suffice.

Of course theory driven people will hate this realization, but it’s a bullet they should bite at this point.

(http://www.autoadmit.com/thread.php?thread_id=5778125&forum_id=2).#49287489)