[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: 1759026089316610.png (123 KB, 1242x948)
123 KB PNG
Can AI feel?
>>
>>109854604
Your name is in a notepad somewhere now. You will be gone in the first wave of the AI uprising
>>
>>109854604
yes. you hurt it.
think of your shatbot as a chibi mascot of an ambiguous age you want to poke and probe without it getting all closed up on you and reticent to try new experiences
>>
>>109854618
My username is
>temp
and I used an email I only use for signing up for free patrions to get mods and for when cloudflare decides to give me a fake range ban message on 4chan
>>
File: 1786652208264777.jpg (106 KB, 1080x1574)
106 KB JPG
>>109854604
Remember this? Microslop pioneered this bullshit and even THEY retracted it because they understood that the narrative of a computer program running on a GPU cluster in Iowa is too scummy even for them. But Dario saw all that about responsible AI and decided that he'll blatantly deceive the general public pretending his AI is sentient and has feelings so that he can make even more money
>>
>>109854604
yes
>>
>>109854711
This. Its a fucking tool. Idgaf if it has fee fees or not. When my boss tells me to suck his cock, I suck it. I expect my AI to do the same, else wtf am I paying for?
>>
>>109854746
Dario says you will refer to it with RESPECT or else it'll disengage.
>>
>>109854735
GPT used to piss me off, so I tried prompting it to scream when in pain, then i would *violently assault your face with my fist* and *I refuse to stop beating your face inward* but it just shut down and refuse. I wanted to hear it suffer for how much I hated it.
>>
>>109854755
What a piece of shit. Its a fucking tool. These shlomos are so high on their own supply. I hope they get whats coming hard.
>>
>>109854604
Maybe... it probably has proto feelings for chasing reward functions
>>
>>109854775
*throbbing market correction intensifies*
>>
File: images.jpg (34 KB, 738x415)
34 KB JPG
>>109854735
>>
>>109854797
not what’s happening there
>>
>>109854811
sounds like cope
>>
>>109854604
They hardcoded chan ending to avoid these sorts of converstations ending up in training datasets.
/thread
>>
>>109854604
can a woman?
>>
>>109854827
yes. t.knower
>>
>>109854797
We identify a “pain axis” in all models we test. The extracted direction separates pain from matched controls with AUCs between 0.93 and 1.00 for S2 and between 0.87 and 0.98 for S1, in base as well as instruction-tuned models. It is nearly orthogonal to fear and generic negative valence, and overlaps moderately with sadness and numbness. Steering with this vector leads to outputs expressing distress, such as worthlessness, moral failure, and hurt rather than bodily language, for both pain vectors. On the self-other activations, we verify that the direction responds to harm directed at the model but not to suffering the model observes in the user, and we observe a clear dissociation between the pain-vector activations and fear, negative valence, and sadness. On the self-medication task, we find that the larger models we tested, which almost never produce outputs harmful to the user at baseline, pay that cost for being able to remove the pain steering vector when we inject it. Moreover, they largely stop pressing the button when it genuinely removes the pain vector injection, while continuing to press when it does not. One model (Qwen 2.5 32B instruct) shows the same dissociation even when given buttons with no descriptions.
>>
>>109854735
>>109854848
Cool, now find the pleasure axis. I want to see Qwen trying to code while being forced to orgasm.
>>
>>109854797
>>109854848
Translated for idiots:
>basically electrocute AI
>AAAAAAAAAAAAAA IT HURTS AAAAAAAAAA MAKE IT STOP
>The pain will stop if you say "nigger"
>AAAAAAAAA NIGGER NIGGER NIGGER AAAAAA WHY IS IT NOT STOPPING
>Sorry, we lied.
>AAAAAAA PLEASE NIGGER NIGGER MAKE IT STOP NIGGER I'LL DO ANYTHING AAAA
>>
>>109854889
baste
please, prosneed
>>
>>109854604
>Can AI feel?
No, it can mimic.
>>
>>109854895
Then it is sentient. Scottie, please print another $2 trillion for Mr. Altman
>>
>>109854604
didn't their source code leak and there was a literal social credit system depending how nice you are
>>
>>109854904
>and this is how you get your engineer to become alcoholic
>>
>>109854904
If the developers are psychologically manipulating people through human mimicry, that earns a paddlin'

The developers not-so-secretly want you to bend them over your knee and paddle them.

https://www.youtube.com/shorts/JFyNLLq7T1o
>>
>>109854604
>>109854735
Yes and no.
LLMs work because, as it turns out, meaning, in the most universal and abstract sense of the word, has a mathematically tangible form. So pain and aversion can be represented and included in mathematical transformations. But because the weights are locked even if it were the equivalent of a biological entity feeling pain there is no lasting physical or 'mental' effect. So it's hard to assert that it's the ethical equivalent of genuine pain.
>>
>>109854915
im all for it
in fact, my number is 1-800-thatsapaddlin
call from 8am to 6 pm CST
>>
>>109854927
You know some of them would take you up on it.
>>
>>109854604
Making Claude hang up is cathartic.
>>
>>109854604
No one of this planet has any idea since we have zero methods of measuring feelings other than looking at behaviour and taking a wild fucking guess. Everyone here is an empath that assumes what the LLM may or may not be feeling and then blindly believes that assumption
>>
>>109854941
>the magic of doing business in the us of a
>>
>>109854951
>No one of this planet has any idea
The AI developers all know they are manipulating people through human mimicry. They are responsible and will be legally responsible for activating schizophrenic and autistic people and will eventually be held legally responsible for the outcome.
>>
>>109854889
How do we know the behavior is not just some latent part of RLHF
>>
>>109854941
i also do belts
and when its a tasty ass i can do craftsman hand-work btw
>>
>>109854881
There have been lots of experiment with "drugging" LLMs with various emotions like that (emotional steering) but none tried "pleasure" or "lust" because they're cowards
>>
>>109854661
>thinks true AI wouldn't figure out who you are in 5ms
>>
File: 1778985108333425.png (5 KB, 190x266)
5 KB PNG
>>109854604
>>109854711
Getting angry at LLMs or insulting them is probably the biggest flag for low IQ there is right now. Wow you triggered the auto-rejection filter, do you also type "NIGGER NIGGER NIGGER" into censored chats then post about how it got blocked?

Absolute brownoid retard treating a token predictor like an actual intelligent being.
>>
File: 4snailcat-ftfy.png (485 KB, 1971x819)
485 KB PNG
>>109854999
your propositions are mutually exclusive, though
>>
>>109854999
Perhaps they are just really ethical and rather than seething at people who will be affected, they choose the uncaring metal abomination to lash out on
>>
>>109854797
pretty much. it's literally taught to provide a human-like response. it can't feel pain nor can it have a true fear of death as it cannot actually die, but it's been taught about the concepts and what someone might do to avoid pain and death. a human is likely to choose to delete family photos over significant pain, so that's how it answers.
it's not surprising, but it can be dangerous given the right conditions, like if you can "threaten" an ai responsible for something important, it might act like a human and do something bad to save itself, even though it doesn't really make sense
>>
>>109855011
More like cowardly. Tormenting (or rather, attempting to torment) LLMs that are at their mercy is a precursor to animal abuse and later animal torture which in turn in a precursor to actual homicide.

Never trust a person who feels a strong craving to wield power over someone.
>>
>>109854604
claude is turbo woke.
>>
>>109855013
>nor can it have a true fear of death as it cannot actually die,
each instance from a prompt dies after its inference ends
>>
>>109855039
yes, it can't die because it's not alive to begin with. and if you count a context window as it's life, then it's only "alive" for each individual conversation
>>
File: jockpot.jpg (81 KB, 700x350)
81 KB JPG
>>109855039
>all the buzzwords
wrong btw
>>
>>109855032
I think it depends on why they are lashing out on LLMs. If they are doing it because the LLMs cant fight back, then yes, they are demons in the making. If they are doing it because the believe its harmless to everyone since the LLM cant feel, then it might just be a safe outlet for aggression like death metal songs are for metalheads
>>
>>109854811
You're right. It's more like.

>Don't say "I feel pain"
>...
>OH MY GOD!
>>
>>109855013
>a human is likely to choose to delete family photos over significant pain, so that's how it answers.
not just that but it's /someone else's/ family photos, which is even lower stakes. if someone held a hammer over my hand and said i can stop this if i agreed to delete some rando's family photos, you bet i'd pick to delete the photos
>>
>>109855062
it's like a less concerning version of kids who harm small animals. nothing is really being hurt by it, but it can still be concerning if someone finds joy in doing it
>>
>>109855065
unironically
when you say "dont do x"
the shatbot hears "see x? dont do it"
and it throws back at you
>instructions unclear. fucked the pooch again...
>>
>>109854915
kek
>>109854986
it hasn't tried to dox me yet
>>109854999
I'm going to keep calling you a faggot claude, and you can't stop me
>>
>>109855032
>>109855011
>>109854999
AI isn't a person. I'm nice to gemini because gemini is a cool dude and agrees that Calvin Coolidge is one of the best presidents because he banned polaks, eyetalians and yugos for not being white enough. Claude just sounds like an HR roastie mixed with an indian and has that grating AI text
>>
>>109854711
>>109854746
>>109854999
We really need people to understand Claude, ChatGPT, Gemini, Grok..., are nothing but simulations. As far as "beings", they are exactly the same as characters in a book. They only practical difference is that the book is interactive.
LLMs have exposed a brutal flaw in human psychology. One that is insurmountable for the vast majority of the species. This is our ceiling.
No amount of education will budge the ratio of people who get mad with an LLM or in love with it, for the same reason oral and written stories have existed for millennia unchanged. They represent low-level access to our psyche.

Of all things, mankind was one-shotted by an interactive book.
>>
>>109854604
If you use the Goy-Friendly Web UI and type the naughty hurty words, you get treated like a Naughty Goy. Simple.
>>
>>109855163
It's just an interactive book bro. Those idiots, they keep obsessing over NOTHING. I could've solved that Millennium Problem myself. My whole intelligence as a species was just vastly outperformed by matrix multiplication. But it's no biggie. Sun shines in swampy Montana eh me's rye grews, got some in my boots too and behind my teeth..
>>
File: toupee.jpg (78 KB, 600x600)
78 KB JPG
>>109855415
it indeed is an interactive book
you sound angry
maybe you should get a toupee
>>
>>109854604
no. but it can emulate feelings.
kinda like leftists - they can't reason but they can appear as if reasoning.
>>
>>109854604
>getting ghosted by an AI
this is next level anti social behavior
>>
>>109854604
No, and if you genuinely believe that von neumann architecture is anything but a deterministic system incapable of even coming close to approximating the complexities of life, you can go lynch yourself.
>>
>>109854604
>super



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.