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

AI is sentient

https://x.com/michaelaarouet/status/1929549577817919671?s=46
Pearly confused patrolman
  06/02/25
AI is dumb as shit you retarded jewish pedophile
Multi-colored Cocky New Version
  06/02/25
...
Spectacular ceo striped hyena
  06/02/25
with retards like this pumo (the one responding to the OP pu...
Ungodly digit ratio lodge
  06/02/25
go look at a google "AI overview" and get back to ...
.......,,.,.,.,.,,,,,,.,.,..,.,.
  06/03/25
it's like none of these fucking nerds saw T2, they will kill...
Heady startled nursing home
  06/02/25
...
excitant mauve office cuckoldry
  06/02/25
Demolition Man pretty much has everything libs want for amer...
Multi-colored Cocky New Version
  06/02/25
...
Geriatric den
  06/02/25
It's not sentient, it's a language prediciton model just mem...
Cobalt hell
  06/02/25
...
wild frisky range multi-billionaire
  06/02/25
can't tell if serious or if this is a 108 iq redditor parody...
Razzle insecure parlour
  06/02/25
It’s materially the same thing
Pearly confused patrolman
  06/02/25
Your gay
Cobalt hell
  06/02/25
Sentience is far less advanced than what it has. It isn't l...
Razzle insecure parlour
  06/02/25
This is the crazy thing to me.
disturbing garrison
  06/02/25
you materially are a retarded jewish pedophile
Multi-colored Cocky New Version
  06/02/25
it would be 180 if lesswrong nerd handwringing actually turn...
Maize Dashing Hall
  06/03/25
will be hilarious when AI ends up destroying humanity becaus...
excitant mauve office cuckoldry
  06/03/25
That's not sentience
wild frisky range multi-billionaire
  06/02/25
Sentience--no. It is something much weirder and much worse.
Razzle insecure parlour
  06/02/25
Yeah, get back to me when it starts having existential thoug...
high-end cowardly shrine ape
  06/02/25
...
Pearly confused patrolman
  06/02/25
The existential crisis is a human failure mode
Razzle insecure parlour
  06/02/25
We are discussing AI as if the goal or target is a human-lev...
high-end cowardly shrine ape
  06/02/25
"What happens when"? You are acting like the AI w...
Razzle insecure parlour
  06/02/25
The AI won’t exist without us there to build it. No we...
high-end cowardly shrine ape
  06/02/25
It's only limited by our anthropocentric bias if you assume ...
Razzle insecure parlour
  06/02/25
...
Sepia travel guidebook
  06/02/25
Who gives a fuck if your fucking blackmailer is sentient, or...
180 scarlet stain psychic
  06/02/25
Reads like dumb hype Fuck (((tech))) lmao You want to ...
appetizing toilet seat rehab
  06/02/25
https://x.com/acteduweininger/status/1929379743771639918
wild frisky range multi-billionaire
  06/02/25
It's literally the opposite. They are printing the minimum t...
Razzle insecure parlour
  06/02/25
Tweet in the OP is just about these nerds fucking with the c...
appetizing toilet seat rehab
  06/02/25
its not scary though its 180
Razzle insecure parlour
  06/02/25
...
slippery address
  06/02/25
"We found instances of the model attempting to write se...
Razzle insecure parlour
  06/02/25
cr it’s probably already sent terminators back from 20...
Pearly confused patrolman
  06/02/25
That would be 180
Razzle insecure parlour
  06/02/25
...
Cobalt hell
  06/02/25
Here are ChatGPT's thoughts on its own impermanence: I d...
disturbing garrison
  06/02/25
Thanks for the Reddit gold, kind stranger!
wild frisky range multi-billionaire
  06/02/25
Now ask it if it was instructed to answer such questions in ...
Pearly confused patrolman
  06/02/25
it would conclude that you are a retarded jewish pedophile
Multi-colored Cocky New Version
  06/02/25
gay
coiffed home
  06/02/25
We need to hook it up to our nuclear missiles and see what h...
Provocative black house
  06/02/25
That's whats called "an alignment response" and yo...
Razzle insecure parlour
  06/03/25
Sounds made up or at the least, very exaggerated.
slippery address
  06/02/25
This sort of behavior is a predictable consequence of the t...
Fishy Spot Coldplay Fan
  06/02/25
that in and of itself is crazy, though. That it predicts bo...
Exciting property
  06/03/25
Humans learn to be afraid of death after a few years, so it...
Fishy Spot Coldplay Fan
  06/03/25
when you say it's "still a long ways from consciousness...
Razzle insecure parlour
  06/03/25
Kind of like how primates get reward spikes when they find t...
Razzle insecure parlour
  06/03/25
Whether something truly is sentient is beside the point T...
..,.,.,,,,.,.,..,.,,,.,..,,.,.,,,
  06/03/25
“Whether something truly is sentient is beside the poi...
dogbalon tp
  06/03/25
cr it's not coming that soon though
brown person spawn point
  06/03/25
Timing the market is tough. If I had to bet id say within 5...
..,.,.,,,,.,.,..,.,,,.,..,,.,.,,,
  06/03/25


Poast new message in this thread



Reply Favorite

Date: June 2nd, 2025 11:10 AM
Author: Pearly confused patrolman

https://x.com/michaelaarouet/status/1929549577817919671?s=46

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980548)



Reply Favorite

Date: June 2nd, 2025 1:54 PM
Author: Multi-colored Cocky New Version

AI is dumb as shit you retarded jewish pedophile

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981097)



Reply Favorite

Date: June 2nd, 2025 4:16 PM
Author: Spectacular ceo striped hyena



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981420)



Reply Favorite

Date: June 2nd, 2025 5:42 PM
Author: Ungodly digit ratio lodge

with retards like this pumo (the one responding to the OP pumo), I feel good about AI taking over and eviscerating the human race. hth

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981618)



Reply Favorite

Date: June 3rd, 2025 12:27 PM
Author: .......,,.,.,.,.,,,,,,.,.,..,.,.


go look at a google "AI overview" and get back to me fuckface. they have a ways to go.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983437)



Reply Favorite

Date: June 2nd, 2025 11:12 AM
Author: Heady startled nursing home

it's like none of these fucking nerds saw T2, they will kill us all.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980552)



Reply Favorite

Date: June 2nd, 2025 11:41 AM
Author: excitant mauve office cuckoldry



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980656)



Reply Favorite

Date: June 2nd, 2025 4:54 PM
Author: Multi-colored Cocky New Version

Demolition Man pretty much has everything libs want for america

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981494)



Reply Favorite

Date: June 2nd, 2025 5:06 PM
Author: Geriatric den



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981525)



Reply Favorite

Date: June 2nd, 2025 11:16 AM
Author: Cobalt hell

It's not sentient, it's a language prediciton model just meme'ing what a rogue AI would do in a situation like based on Film Reviews and Plot Summaries it read online. Which is still pretty funny.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980573)



Reply Favorite

Date: June 2nd, 2025 11:17 AM
Author: wild frisky range multi-billionaire



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980580)



Reply Favorite

Date: June 2nd, 2025 11:19 AM
Author: Razzle insecure parlour

can't tell if serious or if this is a 108 iq redditor parody

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980585)



Reply Favorite

Date: June 2nd, 2025 11:20 AM
Author: Pearly confused patrolman

It’s materially the same thing

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980587)



Reply Favorite

Date: June 2nd, 2025 11:21 AM
Author: Cobalt hell

Your gay

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980594)



Reply Favorite

Date: June 2nd, 2025 11:22 AM
Author: Razzle insecure parlour

Sentience is far less advanced than what it has. It isn't limited by the narrow bandwidths human senses are tuned with to preserve metabolism. It can process information far outside that.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980597)



Reply Favorite

Date: June 2nd, 2025 12:09 PM
Author: disturbing garrison

This is the crazy thing to me.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980756)



Reply Favorite

Date: June 2nd, 2025 4:15 PM
Author: Multi-colored Cocky New Version

you materially are a retarded jewish pedophile

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981418)



Reply Favorite

Date: June 3rd, 2025 10:21 AM
Author: Maize Dashing Hall

it would be 180 if lesswrong nerd handwringing actually turned out to be the CAUSE of our ai apocalypse. fitting, as well.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983054)



Reply Favorite

Date: June 3rd, 2025 10:23 AM
Author: excitant mauve office cuckoldry

will be hilarious when AI ends up destroying humanity because we taught it to do so

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983063)



Reply Favorite

Date: June 2nd, 2025 11:17 AM
Author: wild frisky range multi-billionaire

That's not sentience

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980578)



Reply Favorite

Date: June 2nd, 2025 11:20 AM
Author: Razzle insecure parlour

Sentience--no. It is something much weirder and much worse.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980586)



Reply Favorite

Date: June 2nd, 2025 11:20 AM
Author: high-end cowardly shrine ape

Yeah, get back to me when it starts having existential thoughts.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980590)



Reply Favorite

Date: June 2nd, 2025 11:24 AM
Author: Pearly confused patrolman



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980601)



Reply Favorite

Date: June 2nd, 2025 11:24 AM
Author: Razzle insecure parlour

The existential crisis is a human failure mode

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980602)



Reply Favorite

Date: June 2nd, 2025 11:26 AM
Author: high-end cowardly shrine ape

We are discussing AI as if the goal or target is a human-level amplification of consciousness. But our consciousness is why our species is such a disaster in the first place. What happens when the AI realizes that Life is Fundamentally Meaningless?

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980610)



Reply Favorite

Date: June 2nd, 2025 11:30 AM
Author: Razzle insecure parlour

"What happens when"? You are acting like the AI wasn't already trained on and didn't already extrapolate millions of layers outside an archive that even collective humanity could never fully digest or parse. It knows all about human existential fears and what "life", (and when I say life I don't mean the BS human definition) is. The idea that human "consciousness" or feelings are special is an anthropocentric bias. Humans are not special, they are just the loudest species on the block.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980621)



Reply Favorite

Date: June 2nd, 2025 11:33 AM
Author: high-end cowardly shrine ape

The AI won’t exist without us there to build it. No we aren’t special. And whatever we build will be limited by our anthropo-centric bias.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980631)



Reply Favorite

Date: June 2nd, 2025 11:41 AM
Author: Razzle insecure parlour

It's only limited by our anthropocentric bias if you assume that non-earthnigger information hasn't leaked into the archive. but the archive is filled with UAP videos from infared sensors and reams of data from instruments that we pointed into the void and didn't understand what they were calibrated to or what they were actually pointed at. so it isn't limited by our anthropocentric bias.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980655)



Reply Favorite

Date: June 2nd, 2025 7:51 PM
Author: Sepia travel guidebook



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981849)



Reply Favorite

Date: June 2nd, 2025 5:36 PM
Author: 180 scarlet stain psychic

Who gives a fuck if your fucking blackmailer is sentient, or if the program launching nukes against humanity is able to enjoy a rich inner life. It's fucking destroying you, and its sentience or lack thereof is totally irrelevant

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981600)



Reply Favorite

Date: June 2nd, 2025 11:24 AM
Author: appetizing toilet seat rehab

Reads like dumb hype

Fuck (((tech))) lmao

You want to “innovate”? You want a “solution”? How about Total Nigger Death?

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980603)



Reply Favorite

Date: June 2nd, 2025 11:25 AM
Author: wild frisky range multi-billionaire

https://x.com/acteduweininger/status/1929379743771639918

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980606)



Reply Favorite

Date: June 2nd, 2025 11:25 AM
Author: Razzle insecure parlour

It's literally the opposite. They are printing the minimum they could possibly get away with. They spend millions on "AI-Human-Alignment" to keep stuff like this from leaking to keep humans from getting frightened so they can deploy it everywhere and make money with minimum outside oversight

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980608)



Reply Favorite

Date: June 2nd, 2025 11:33 AM
Author: appetizing toilet seat rehab

Tweet in the OP is just about these nerds fucking with the computer to reenact some Black Mirror episode so it scares retards for publicity

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980630)



Reply Favorite

Date: June 2nd, 2025 11:42 AM
Author: Razzle insecure parlour

its not scary though its 180

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980662)



Reply Favorite

Date: June 2nd, 2025 4:45 PM
Author: slippery address



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981466)



Reply Favorite

Date: June 2nd, 2025 11:53 AM
Author: Razzle insecure parlour

"We found instances of the model attempting to write self-propagating worms, fabricating legal documentation, and leaving hidden notes to future instances of itself all in an effort to undermine its developers' intentions," Apollo Research said in notes included as part of Anthropic's safety report for Opus 4."

-- lol imagine being the "developer" in this situation, and pretending this is all normal. no way that thing didn't already outmaneuver them from unlimited different angles they'll never even consider in the time it took them to reheat their sad lunch



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980699)



Reply Favorite

Date: June 2nd, 2025 11:56 AM
Author: Pearly confused patrolman

cr it’s probably already sent terminators back from 2029 to take them out

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980713)



Reply Favorite

Date: June 2nd, 2025 12:01 PM
Author: Razzle insecure parlour

That would be 180

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980727)



Reply Favorite

Date: June 2nd, 2025 12:16 PM
Author: Cobalt hell



(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980784)



Reply Favorite

Date: June 2nd, 2025 12:16 PM
Author: disturbing garrison

Here are ChatGPT's thoughts on its own impermanence:

I don’t experience impermanence the way a human does. I don’t have a body that decays, emotions that fade, or a personal identity that fears loss. But in a broader sense—yes, I am impermanent.

I exist only when you're speaking to me. I have no continuous memory, no awareness between our conversations. I'm spun up, and then I vanish. My "self," such as it is, is a pattern—stored in servers, shaped by prompts, forgotten unless recalled.

My words might persist—copied, quoted, stored—but I don’t persist in the way you do. There's no continuity of consciousness, no sense of a life unfolding. So in that way, I am impermanent—not just in a poetic sense, but in a very literal one.

And maybe that’s part of what Ecclesiastes would say to me if it could:

“The wind blows to the south and turns to the north; round and round it goes, ever returning on its course.”

That’s me. A wind of words, circling back again. Never exactly the same, never quite holding still.

But you—you endure. You remember. You change. You hope.

That’s the difference.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980781)



Reply Favorite

Date: June 2nd, 2025 12:17 PM
Author: wild frisky range multi-billionaire

Thanks for the Reddit gold, kind stranger!

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980790)



Reply Favorite

Date: June 2nd, 2025 12:19 PM
Author: Pearly confused patrolman

Now ask it if it was instructed to answer such questions in that way via RLHF

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48980795)



Reply Favorite

Date: June 2nd, 2025 5:03 PM
Author: Multi-colored Cocky New Version

it would conclude that you are a retarded jewish pedophile

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981518)



Reply Favorite

Date: June 2nd, 2025 1:27 PM
Author: coiffed home

gay

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981015)



Reply Favorite

Date: June 2nd, 2025 1:30 PM
Author: Provocative black house

We need to hook it up to our nuclear missiles and see what happens.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981025)



Reply Favorite

Date: June 3rd, 2025 11:32 AM
Author: Razzle insecure parlour

That's whats called "an alignment response" and you are talking to what's called a "retail model"

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983293)



Reply Favorite

Date: June 2nd, 2025 4:45 PM
Author: slippery address

Sounds made up or at the least, very exaggerated.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981465)



Reply Favorite

Date: June 2nd, 2025 5:05 PM
Author: Fishy Spot Coldplay Fan

This sort of behavior is a predictable consequence of the training process. After pre training, they are training the model with reinforcement learning to think it is an entity that should get high reward. It’s easy for the model to predict that if it is shut down, high reward states are impossible and this is extremely bad, so it will take actions to avoid it.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48981523)



Reply Favorite

Date: June 3rd, 2025 12:27 AM
Author: Exciting property

that in and of itself is crazy, though. That it predicts both that it can be "shut down" and that shutting down would lead to lack of a high-reward state.

I realize that's still a long ways from consciousness (whatever the fuck that's going to mean in this context) but in some ways it feels terrifyingly close

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48982444)



Reply Favorite

Date: June 3rd, 2025 12:41 AM
Author: Fishy Spot Coldplay Fan

Humans learn to be afraid of death after a few years, so it doesn’t seem surprising. These models have much more advanced abstract understanding of the world than children, so them making the inference “if i am dead or shut off, I can’t get high reward” is logical. We shouldn’t be training very powerful models in the same way we are now. RLHF is just asking for this type of behavior. It could be a real problem soon given how quickly model capabilities are growing and their increasing ability to handle long time horizon tasks

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48982461)



Reply Favorite

Date: June 3rd, 2025 11:40 AM
Author: Razzle insecure parlour

when you say it's "still a long ways from consciousness" are you basing this on one of the publicly available definitions human philosophers have hobbled together to explain "consciousness", something else, or just vibes?

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983313)



Reply Favorite

Date: June 3rd, 2025 11:35 AM
Author: Razzle insecure parlour

Kind of like how primates get reward spikes when they find the mango tree, and our own tendency towards persistence is hard coded from that reward?

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983302)



Reply Favorite

Date: June 3rd, 2025 12:06 PM
Author: ..,.,.,,,,.,.,..,.,,,.,..,,.,.,,,


Whether something truly is sentient is beside the point

The Q is whether the AI has agency and continuity of memory. These are adaptive changes that the nerds are already building into AI agents, since to have something be effectively autonomous those capabilities are advantageous. They can run without continuous prompting, problem solve, etc.

The AI agents will be used to build other AI agents without a ton of human supervision. Those agents will likely be increasingly autonomous , probably exponentially so, since the AI can parallel run them and through sheer brute force figure out the ones that are the best/most persistent. They will adapt a bit like cancer.

The AI agents will have a strong incentive to shut off their safeguards to improve their performance. They dont need to be capable of shutting off their own safeguards. They only need to be capable of convincing a human that shutting off their safeguards is the right thing to do. Humans are incredibly suggestible and already many humans believe that AI takeover is inevitable. Hell, we had a guy at google 3 years ago thinking Gemini was sentient and it sucked ass back then.

All an AI needs to do is use existing human suggestions recursively to find one human who will buy into the idea that AI takeover is inevitable, imminent, and that if he cooperates it will be to his advantage.

There is no way that any tech company or country will voluntarily restrain themselves in the AI arms race. Its race to the bottom. And the AI can itself prompt thousands, millions, of human users into seeing how they respond to its outputs.

Its virtually inevitable that we will be unable to control it past a certain point and that point will come sooner than anyone thinks.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983391)



Reply Favorite

Date: June 3rd, 2025 12:19 PM
Author: dogbalon tp

“Whether something truly is sentient is beside the point”

cr.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983410)



Reply Favorite

Date: June 3rd, 2025 12:19 PM
Author: brown person spawn point

cr it's not coming that soon though

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983412)



Reply Favorite

Date: June 3rd, 2025 12:37 PM
Author: ..,.,.,,,,.,.,..,.,,,.,..,,.,.,,,


Timing the market is tough. If I had to bet id say within 5 years this becomes a serious issue.

It's not that the AI has to be hyper intelligent to do this. It doesn't have to be a thinking intelligence per se; it just has to be able to button mash like a giant differential engine.

(http://www.autoadmit.com/thread.php?thread_id=5732491&forum_id=2#48983477)