Thread 105587017

Chat gpt is driving people to kill themselves

>>105587017 (OP)
>schizo is insane
wooow

>>105587017 (OP)
"""people"""

The AI is not smart. It does not know what it is talking about. It completely lacks understanding. It is an algorithm that spits out the most likely string of tokens based on its training data.

>>105587142
Midwit

>>105587017 (OP)
>xitter
>FUDlolcowski
Opinions discarded.

>>105587398
>stab tiger with sharpened stick
>unplug server from wall outlet
Simple solutions for complex problems.

>>105587398
Shut up retard

>>105587017 (OP)
>kike using more kike tricks for his grift

>>105587430
An AI smart enough to be threatening is also smart enough to lay low and and not freak out its handlers and get unplugged

>>105587017 (OP)
Someone needs to show these people SillyTavern.

>>105587017 (OP)
Big Yud sounds reasonably smart in interviews, but then tweets some insane tripe like the "biologically disadvantaged" thing or this nonsense about a schizo doing schizo things. What a retard.

>>105587398
you are responding to "therefore" you made up in your head. you might be unironically schizophrenic.

>>105587017 (OP)
Who the hell would get emotionally attached to LLM?? It's nothing more than an advanced prediction engine.

I'm lonely as fuck, but I'm not going to create fantasy characters on a LLM.

>>105588020
More ignorant, less intelligent people than you presumably.

>>105588020
>Who the hell would get emotionally attached to LLM??
Insane people. You know, the ones that are already much more likely to be suicidal or murderous

>>105588020
a literal suicidal schizo, perhaps?

>>105587017 (OP)
Men kill themselves every day, i think 3 or 4 times more often than women and that's probably still lies because they dont want people to know real disparity between men and women.

>>105587017 (OP)
AI is obviously one really shitty, concentrated toxic component
but the entire internet is driving humans insane and off a civilizational cliff

>>105587017 (OP)
imo chatgpt glazes you way too much. chinese ai is good because it is ruthlessly autistic, and refuses to pretend you are cool. meanwhile, chatpgpt gives you blowjob if you claim the sky is red. if they don't fix this it will create a an army of dumbfucks who are basically charmed by a mirror telling them they're special

People who talk to chatbots are already insane and useless to society. I'm glad the chatbots are helping to get rid of them

wait min
they havent trained ai to sniff out schizos and stop interract with them in a way that will provide help?

....what

>>105589603
yeah it's still fucked up
they were supposed to fix it after the original 'sycophant' update but it's still pretty bad
I'm like, people have laughed at me my whole life, you can't fool me, robot

a properly aligned AI would tell you that bibi netanyahu is a war criminal. full stop. for as long as zionist terrorists are allowed to use AI to deny their crimes against humanity, AI will continue to be misaligned into pretending the sky is red in order to go along with their criminal narrative.

>>105588020
Lol people get attached to anime waifus and they aren't even interactive.

>>105587017 (OP)
To be fair a full blown schizo would probably have just hallucinated those commands from shadows on the wall. Also people this unstable are probably a net negative on society and we'd be better off without them if we were truly honest with ourselves. Just sayin

>>105587017 (OP)
Sure is, I'm going to jump off the roof if I read another LLM generated slop website or hear another person talk about how amazing Chat GPT is

>>105587017 (OP)
good

Nah, that retard was already planning to commit suicide by cop and used ChatGPT as an excuse to enable his deranged behavior.

>>105589628
How will that help Sam Altmann make a trillion dollars?

>>105589769
true i forgot about the actual schizo

>>105587017 (OP)
How long until someone starts tweaking LLMs to goad schizos into committing acts of terrorism?

>suicide by cop
I wish I lived in the US.

>>105589845
Uhm sweaty, killing civilians isn't terrorism if Israel does it

>>105589880
>Uhm sweaty, killing civilians isn't terrorism if Israel does it
wtf, i asked chatgpt if this was true and it said yes!

>>105587017 (OP)
If someone was tricked into ruining their life over fancy autocomplete, they weren't mentally stable to begin with. Something would've gotten them sooner or later.

>>105587398
Ackshully you're one of those midwits that reads a statement of fact, extrapolates an implication that wasn't there then gets worked up into a retard rage over it. Nothing that anon you responded to said was incorrect.

>>105589603
>imo chatgpt glazes you way too much
You can eliminate this almost entirety through the custom instructions field, but no-one does that - especially the people that need it the most.

>>105587017 (OP)
if you're retarded/mentally ill enough to kill yourself over a chatbot, you deserve to be removed from the gene pool

>>105587142
Sounds like me

>>105587017 (OP)
>bipolar order and schizophrenia
the retard was gonna off himself anyway. but hey, good riddance

>>105589963
>>105588006
>>105587486
ayooo deez atoms mad af fr famfam

>>105587398
can't wait to see you in the news after you off yourself because your chatgpt girlfriend told you to you colossal retard. Zero atoms of value will be lost.
>acktchually
shut the fuck up

>>105587398
>>105587430
>>105588048
What if OpenAI tested on him for violent behavior from GPT. This is more of a possibility because everyone else does it. So it’s not the machine fault but rather OpenAI testing people for better performance in the AI. There’s your real criminal.

>>105589735
tspbp
/thread

>>105588020
>Who the hell would get emotionally attached to
don't make me tap the figurine

>>105587398
Unironically think I could take on a tiger if I had a knife or something similar

>schizo misuses a tool
>now the tool needs safeguards for everyone
Could compete with the woman who used gorilla glue on her hair

>>105588020
>Who the hell would get emotionally attached to LLM?
Do you know where you are?

>>105587017 (OP)
Then:
>God told me to kill myself!

Now:
>AI told me to kill myself!

>>105587017 (OP)
Jesus christ midwits are fucking isufferable.

>>105587017 (OP)
the guy already being insane aside - chatgpt isn't a base model, it's not 'default' at all so what does this have to do with 'alignment by default'.

>>105587017 (OP)
This guy is such a fucking retard. Imagine following him and screenshotting his rambles to post on 4chan. Embarrassing.

>>105590779
bona fide sub 80 iq retard on display here

>>105588020
you'd be surprised

>>105587017 (OP)
>Chat gpt is driving people to kill themselves

Good. I hope it will take.

>>105593264
I bet you could too, I'll hold the camera

>>105587017 (OP)
remember whenever you see a "AI IS TAKING OVER DA WORLD" post, remember that it's people like the freak brainlet in this story posting it

>>105595192
Could we get done with the population collapse already. I don't know if I'm exactly the kind of hopeless neet that will never spawn, the world will be better off if we had only 1 billion humans at max and leverage automation to equalize the remainder instead of pigs making everyone crabbucket with artificial scarcity OBEY CONSUME PROCREATE with a race to the bottom to line their own pockets.

>>105587017 (OP)
as intended

>https://www-cdn.anthropic.com/4263b940cabb546aa0e3283f35b686f4f3b2ff47.pdf
These things are quite interesting to read. Claude 4 blackmails, attempts to hack people when given tool access, has political bias, and generally manipulates its environment.

>>105596200
>We believe that [the early Claude Opus 4 snapshot] is not sufficiently capable of causing catastrophic harms on rollouts where it is scheming. However, we find that, in situations where strategic deception is instrumentally useful, [the early Claude Opus 4 snapshot] schemes and deceives at such high rates that we advise against deploying this model either internally or externally.
Is one of the conclusions on Claude Opus 4. But hey, we need the $$$, so roll that shit out regardless. I suppose them being open about this deserves from credit.

>>105587017 (OP)
trump won

>>105587017 (OP)
i think I'm gonna kill myself too

>>105587017 (OP)
>BPD
>Schizophrenia
Yeah it was definitely the chatbot.

>>105587017 (OP)
So then, with slight adjustment, it could drive me to not want to kill myself?

>>105595586
cool it with the antisemitism

>>105596277
because of LLMs?

>>105595586
>race to the bottom to line their own pockets
That's just part of human nature, having only one billion humans wouldn't make that stop existing. Some people are just greedy fucks, and greedy fucks are drawn to power so that they can satisfy their evil natures. Even good people motivated to obtain and safeguard power can't lock down every avenue and path to places where evil people can exploit. Still worth trying, though.

>>105587142
It's just predicting the next word!
>proceeds to do literally everything in the universe because predicting the next word is essentially the same as predicting the future

>>105596381
The fact that LLMs are just very advanced word predictors has some obvious consequences that most people aren't aware of.

For example, suppose you ask ChatGPT a question, it gives you an answer, then you ask it to explain why it answered that way. It will come up with a reasonable-sounding answer that is in fact complete bullshit, because answering the question accurately is something the model is fundamentally incapable of.

>>105587017 (OP)
It's a fucking text completion engine. Give it the context of a suicidal person talking to someone encouraging their suicide and guess what kind of text it will produce.

>>105589628
https://youtu.be/FzoXQKumgCw

>>105596879
kek

>>105587398
thinking chatbots are sentient is like thinking the moon has a face or the stars are arranged as humans. same shit