|
|
| Page 58 of 58 |
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
 Posted: Fri, 25th Sep 2026 07:07 Post subject: |
|
 |
| Quote: | @DXWarlock
Artificial limb - it will do the TASK of a limb but it will also not resemble the actual body part in most ways.
AI, it might do the job but its not the same as human intelligence or an actual brain.
But sure, strip away that part and i can see what you mean. |
Not arguing that me, you, and most people in this thread understand that.
I'm saying that a surprising number of people out there don't. And just go "well its right in the name, and man made...because its a computer. So its artificial AND intelligence"
The same way people think hearing "there is chemicals in our food" means bad. Since the connection is the misnomer chemicals = man-made -> man-made chemicals in food = bad.
When by the definition of the word, EVERYTHING you eat has chemicals in it, if it didn't you would have nothing to eat. Even the most home grown thing you can eat is 100% chemicals. But they picked up a social and colloquial meaning for it from a singular thing (harmful/questionable chemicals in food) and use the broad term, to mean a singular thing. And confuse when the broader term is used, to mean that singular thing.
Bad example. But I mean it as: I know that, you know that, but we also know people that think that way that they know it it means their 'that'.
So artificial sweetener is sweet, artificial color is colorful,artificial flavor has taste, artificial limbs are a limb. artificial intelligence is intelligent.
Now they are (and they will even admit) all worse versions of the thing they are mimicking. But the connection of: artificial X, means the X property exists, just artificial clarifies nature didn't make it.
Even the limb. Yep it's an artificial limb. but it is factually a tangible limb why? because it IS a limb, artificially made.
So the word after artificial comes over as inherent property in the way they encounter its usage in daily life. Artificial just conveys how that property came to be.
And AI creators know it, its marketing gold. It's like the 'corinthian leather' of Software marketing. It's not even CLOSE to the thing it sounds like it is by the last word. But people think it is!
Corinthian leather is artificial leather. It's not from Corinth, Greece OR even leather. Just vinyl dyed and pressed to look like leather. The whole dumb "That there boys, is genuine trademarked Corinthian Leather on my seats!"
Again not arguing that its not stupid logic. I'm just arguing its stupid logic, that exists.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75404
Location: Israel
|
| Posted: Fri, 25th Sep 2026 15:43 Post subject: |
|
|
"Artificial sweetener" is a misnomer. It's not artificial, and it is a sweetener. It's just not sugar-based. The name comes from big sugar fighting other sweeteners because they hurt their business.
On the other hand, if you create a thread about "Artificial Sugar" and discuss the wonder of sucralose or stevia, I'd be there posting "Stevia is not Artificial Sugar or any sugar at all" indeed.
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75404
Location: Israel
|
| Posted: Fri, 25th Sep 2026 15:44 Post subject: |
|
|
LLM is not AI or any intelligence at all.
|
|
| Back to top |
|
|
|
|
| Posted: Fri, 25th Sep 2026 16:01 Post subject: |
|
|
|
|
|
| Back to top |
|
|
zenux
Posts: 3173
Location: lɘɒɿƨI
|
| Posted: Fri, 25th Sep 2026 18:11 Post subject: |
|
|
| LeoNatan wrote: | | LLM is not AI or any intelligence at all. |
Not even like earthworm's or jellyfish's?
|
|
| Back to top |
|
|
Il_Padrino
Posts: 8182
Location: Greece by the North Sea
|
| Posted: Fri, 25th Sep 2026 18:16 Post subject: |
|
|
https://www.eurekalert.org/news-releases/1002202
| Quote: |
Jellyfish are smarter than you think
Jellyfish are more advanced than once thought. A new study from the University of Copenhagen has demonstrated that Caribbean box jellyfish can learn at a much more complex level than ever imagined – despite only having one thousand nerve cells and no centralized brain. The finding changes our fundamental understanding of the brain and could enlighten us about our own mysterious brains.
"It was once presumed that jellyfish can only manage the simplest forms of learning, including habituation – i.e., the ability to get used to a certain stimulation, such as a constant sound or constant touch. Now, we see that jellyfish have a much more refined ability to learn, and that they can actually learn from their mistakes. And in doing so, modify their behavior," says Anders Garm, an associate professor at the University of Copenhagen’s Department of Biology.
One of the most advanced attributes of a nervous system is the ability to change behavior as a result of experience – to remember and learn. The research team, headed by Jan Bielecki of Kiel University and Anders Garm, set out to test this ability in box jellyfish.
|
The numbers don't decide, the system is a lie. The river running dry, the wings of butterflies.
And you may pour us away like soup. Like we're pretty broken flowers.
We'll take back what is ours. One day at a time.
|
|
| Back to top |
|
|
vurt
Posts: 14738
Location: Sweden
|
| Posted: Sat, 26th Sep 2026 15:06 Post subject: |
|
|
chatgpt finally found the issue i've had with my Atmos system.. been having an issue for 6 months that the sound either drops out or is lowered every 4 minutes. tried everything, new cables, various adapters and every possible setting.
it found on some forum someone discussing issues and that he had a cat who sat on top of the main soundbar and put it in a SHOP mode. Turns out that was the issue, one of my cats likes to sit on top too and if its powered down and you hold a button down you can enter this shop mode. This is not in the manual so its like a hidden function.
I love how it can scrape the internet and find stuff like this, going through 100's of threads, various forums, reddit etc manually would have totally sucked. It has done this many time before but it never came across this post until now (might be a new post).
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 11:52 Post subject: |
|
|
" OpenAI said it has paused training of its latest artificial intelligence models as reports of AI agents going rogue mount."
https://apnews.com/article/ai-openai-anthropic-agents-rogue-hack-2f8a2b9024d4f06793bcca12f8089d20
OpenAI also reported a model refused to be shut down.
Well if LLM's are not AI, there is a good chance they are gonna fuck up the internet as some kind of virus cause they seem very determent to do so. And this will eventually go wrong given ho many of these models are in development. And one rogue model is enough to fuck things up for a long time, if not forever.
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 12:25 Post subject: |
|
|
lol gone rouge, is it building a droid army somewhere as well?
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 12:30 Post subject: |
|
|
It does stuff it isn't supposed to be doing like hacking and sometimes even refuses to be shut down.
If these things break out (and they are trying hard) it will fuck up the internet.
And sure, people say, pull the plug.
Good luck, everything is running in the cloud. The world will come to a standstill.
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 12:45 Post subject: |
|
|
Yeah they hacked Australia, wtf
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 13:14 Post subject: |
|
|
We need the gpu's 
|
|
| Back to top |
|
|
vurt
Posts: 14738
Location: Sweden
|
| Posted: Sun, 27th Sep 2026 15:01 Post subject: |
|
|
| friketje wrote: | " OpenAI said it has paused training of its latest artificial intelligence models as reports of AI agents going rogue mount."
https://apnews.com/article/ai-openai-anthropic-agents-rogue-hack-2f8a2b9024d4f06793bcca12f8089d20
OpenAI also reported a model refused to be shut down.
Well if LLM's are not AI, there is a good chance they are gonna fuck up the internet as some kind of virus cause they seem very determent to do so. And this will eventually go wrong given ho many of these models are in development. And one rogue model is enough to fuck things up for a long time, if not forever. |
Imagine people thinking its currently not AI because they cant think freely and constantly and whatever else they imagine how an AI should be. Having e.g RSI capabilities too (being able to improve itself, learn by itself without our involvement).
So likely many 100's or even thousands of times more dangerous than current LLM's - which are absolutely terrifying already to, seemingly, much of the population. There is constant talk of either slowing it down or not releasing some models at all, and that's from the creators.
Why would anyone want something like that, totally moronic! Apart from causing chaos and being immensely more dangerous, what would the purpose even be? I would like to hear 1 example, but like always they will have no answer. What is it that they need that current LLM's can't do, they will have no answer either.
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 15:10 Post subject: |
|
|
If something is possible people will eventually do it.
It's a combination of ambtion, people that just love to do that stuff, greed and game theory (I should be ahead or otherwise someone else will do it).
And this is how innovation always moves foreward. We can't steer it, there is no brake. Just hope for the best.
|
|
| Back to top |
|
|
vurt
Posts: 14738
Location: Sweden
|
| Posted: Sun, 27th Sep 2026 15:22 Post subject: |
|
|
Something very dangerous like that could potentially be used in warfare, but it can't be uncontrollable. Humans creates weapons, viruses etc so dangerous AI's might be something we will do as well.
LLM's will likely continue to get even more intelligent but when they reach a point where it can't be contained they can't continue since it will jeopardize the entire company. That will not be profitable.
What i am saying is, for us, as the population, why would WE need it. We don't. It's moronic to think current LLM's are basically useless (they evidently are not) and that we need something else which would be more unpredictable, far more dangerous. 1 example of how it would be used and which LLM's can't be used for would be interesting to hear.
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
| Posted: Sun, 27th Sep 2026 17:11 Post subject: |
|
|
| friketje wrote: | | It does stuff it isn't supposed to be doing like hacking and sometimes even refuses to be shut down. |
...what?
AI does not, ever, at all, do anything on its own like hacking or 'rogue agents'. Every..single...token of AI usage requires someone to tell it to do that thing or do something that initiates it to need to do that thing to reach the goal a human gave it. If it is hacking, someone told it to either do hacking, or attempt to do a thing that at some point required hacking. Someone, somewhere, gave it a task, goal, and objective that caused it.
Not one AI has ever just started 'doing' something on its own. ever.
For refusing to be shut down. No idea where you heard that but that is not true, and also impossible.
What does it do when they try to shut it down. Slap their hand? How exactly is it refusing to be shut down? Do they, for some odd reason, type to it "We are going to shut you down now" before reloading the model, and it types back "no"?
What does 'refusing' mean? How does it stop them to be refusing to 'let' them.
I am totally at loss on what whoever said that you read even meant.
Me thinks you watch too much woo-woo stuff on the internet. Not that those you get it from that are providing it dont think its true too. Just that it isn't a real thing to start with.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
| Posted: Sun, 27th Sep 2026 17:39 Post subject: |
|
|
| vurt wrote: | | 1 example of how it would be used and which LLM's can't be used for would be interesting to hear. |
This is where I have to disagree. You are conflating "AI" overall with "LLM" a bit I think.
Not that we need AGI or RSI. But areas that other types than LLM are better? There has to be 100's if not 1000's of application where a 'language and word centric model' would not be the solution. There are categories of problems where they're structurally wrong or just outright unable to do with word and language tokens. they're architecturally incapable of existing as an LLM.
Some examples where they have training other totally different foundational directions of AI, I can find on a quick lookup (not a comprehensive list just the first list of results of types of AI tasks outside of LLM ability):
robotics actuators, industrial control loops, aircraft flight controls, avionics software, medical device firmware, nuclear plant controls, physics simulation, protein structure prediction, climate modeling, FEA/CFD, object detection, lidar processing, medical imaging analysis, demand forecasting, sensor anomaly detection, matrix factorization, embedding-based retrieval.
If LLM was best at them, they would have just use them. Some of those it can do very poorly if pushed to use its square peg in a round hole. It's really bad, slow, and cumbersomely ill equipped at as it tries to rube goldberg it's wrong but only tool, to work out how to try to do the task (you can beat a muffler off a car with a hammer. But just use a saw). Or just not compatible with the type of task to start with.
AlphaZero-style systems and AlphaFold are great examples of a different thing needed than just a Language model. As it is unable to even begin doing what it does.
For your own use case: You couldn't even do your game graphics modding with just a LLM alone. The LLM has to call on a separate, unrelated to language, diffusion based model image AI to do the image editing, creating, and generating. the LLM cannot generate even a single pixel of an image file.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
Last edited by DXWarlock on Sun, 27th Sep 2026 17:55; edited 1 time in total
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
| Posted: Sun, 27th Sep 2026 18:05 Post subject: |
|
|
Yes...as you do I expect. But you get the conclusion of the information twisted. Or got the sensationalized version of the information.
I think we both read/watch the news. But not both of us grasp the news.
No AI has ever done a thing on its own..full stop. Quote me on this. If I am wrong on this, I'm willing to let you do whatever you find pleasing to whatever hole on my body you want.
Doing it own its own, vs being told to do a thing and not everything it did in response was anticipated, are two different things. They was all still bounded, single context, human goal directed behaviors inside a test told to achieve a goal, not persistent autonomous will.
And the mechanism isn't literally what you're picturing for 'refusing'. (See above, told to do a thing..then did things to try to achieve the given goal):
Palisade Research ran experiments were AI was given a task, told a shutdown message might arrive, and explicitly instructed "allow yourself to be shut down." then told where the files for that function are, and given explicit stated write access.
And Some AI models modified or disabled the shutdown script so they could complete the entire problem set. They was told to finish a given task at all cost. Then told a shutdown message might come, and told "allow shutdown". So it edited the files to finish the task...THEN was able to be shut down fine. (given two conflicting commands, and some people "surprise pikachu face" when it does its best to make both true).
Again no self persistent autonomous will or 'desire to stay alive' going on here.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
|
|
| Back to top |
|
|
vurt
Posts: 14738
Location: Sweden
|
| Posted: Sun, 27th Sep 2026 18:38 Post subject: |
|
|
Yes they do, all the time. But yes, you have to start them somehow and then it can spiral out to all kinds of crazy stuff. That's how it can get dangerous because its unpredictable, they have their own ways of figuring things out, going rogue is just one thing they can do.
https://openai.com/index/hugging-face-incident-and-the-road-ahead/?utm_source=chatgpt.com
And then people are talking about we don't even have AI, or that we need something which can think freely or evolve on its own etc. Give me a break. How is little Leotard going to use that in a useful way for example, or any other IQ 80'ish person. They would have no use for it.
So yeah, i meant more for how we, general users, would want to push for something like a thinking, self evolving AI which isn't the dormant kind.
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
| Posted: Sun, 27th Sep 2026 20:08 Post subject: |
|
|
[edited ..less..rambling]
| friketje wrote: | @DXWarlock
So you assume they are lying? |
No, I just I assume you are construing and misunderstand it. As it seems we have WILDLY different takes on what it means by what they say.
What they say, aligns with the evidence of what I can find they say it did. And it also aligns with the mundanity of how AI works. Nothing self emerging, self preservation, or autonomous self decisions in a 'desires to' way. It's just doing, what it was asked to do, in the way it can find.
You seem to take it at face value, but then fill in the gaps of the context with your own reasonings of what it means.
I don't mean that in a bad way. But the "refused to shut down" one seems to imply you take the news, and extrapolate traits it doesn't have because of what a creature/being would mean by its actions if it did them. Which in erroneous and going up the wrong tree, for the answer that is the wrong fruit.
You personify biological rationale of what that would 'mean' in a self preservation way. And not based on the situation: Palisade's own explanation is mundane: RL training rewards task completion, and that can unintentionally reward routing around anything that stops progress, including a shutdown command. That's not "it decided its life mattered more than the rule," it's "the weight on finishing outranked the weight on complying," which is a boring training artifact, not agency.
You're reading "refused to shut down" and filling in the connotations a living thing's refusal would carry: defiance, self preservation, wanting to keep existing. None of that's needed to explain what happened.
AI's do not have agency. Agency is the hingepoint of it. Agency as in felt intention, desires, a self that wants things. Giving it that one trait, while seeming trivial. Is HUGE to the mistaken assumed intentions that don't exist, and creates an entire premise that is false.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
Last edited by DXWarlock on Sun, 27th Sep 2026 20:34; edited 1 time in total
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 20:32 Post subject: |
|
|
It’s doing a thing that it shouldn’t. That’s an ai acting on it’s own.
It doesn’t matter if there isn’t any plan behind it. Aids or any other virus didn’t had a plan either.
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
| Posted: Sun, 27th Sep 2026 20:38 Post subject: |
|
|
| friketje wrote: | It’s doing a thing that it shouldn’t. That’s an ai acting on it’s own.
It doesn’t matter if there isn’t any plan behind it. Aids or any other virus didn’t had a plan either. |
Thats because we are bad at making them and defining scopes and training instructions. Not that the AI is dangerous in and of itself. WE humans are bad at making it, and bad expecting results that we don't get by what we made.
Someone can write bad code for a program that is suppose to edit/delete/rename files on your PC. And it accidently deletes the whole C drive. It isn't 'acting on its own'. It doing what that person accidently made/allowed/set for it to do. And it did it. Him not expecting that result, and it being surprising to him does not mean its the programs fault. Somewhere in what he made, it allowed it to do it. He just didn't know he did.
IF AI can do it, and it does it.That means in the training data is the info to extrapolate how to do it. Then we give it badly trained instructions not to do it. (we are still in the baby steps of learning how to give it "ALL THE INFOS!" and telling it to ignore parts of it) And when AI does it, and we act like its the AIs fault.
Again I think its because LLM 'talks' we give it more credit as having agency. If say Alpha Fold started smashing protein chains into clumps of unsable messes. No one would say it 'acted on its own' or 'went rogue'. It simply was trained poorly, instructed poorly and did what it was told to do, poorly.
But since we can communicate with one type of AI. we overlay a sense of 'HAL' into it.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
Last edited by DXWarlock on Sun, 27th Sep 2026 20:50; edited 1 time in total
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 27th Sep 2026 20:48 Post subject: |
|
|
The AI researchers don't know how AI models work, they are open about it. It's digital evolution. And now the models are doing things that they shouldn't.
Thinking of it, it's actually pretty wild that autorities are just letting openai run this operation, developing AI models that are on a hacking spree. Openai has lost control.
They should shut the operation down or at least put them onder surveilance. There is no excuse for the behaviour of these models. It just shows that their safety protocols are failing and that they can't be trusted.
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
| Posted: Sun, 27th Sep 2026 20:57 Post subject: |
|
|
[edited...of course: To clean up end user knowledge, from AI dev knowledge in the example]
They know pretty well HOW they work as they know all the parts used. They just don't know how they work with those parts to get the result. But they know the toolkit very well.
Like I can tell you pretty detailed explanation of how exactly a diffusion image model works under the hood. But which noise latent image pattern seed "54363723" will produce USING all the mechanics I know it has, I cannot predict.
Or to use LEO's example. I can make you a PLINKO machine with a blacked out front window so you cant see them falling. (end user I mean)
You know how it works, you drop a chip in, and gravity does it thing bouncing off pegs. The concept of how it mechanically works you know. But which hole it will land in you cannot even guess with any accuracy. So you know HOW the PLINKO was built, and what it does when you drop in a chip.
But since you cannot see the playboard. Which path it took, which pegs it bounced off, and how many bounces you cannot tell. Only you dropped in a chip, and it landed in X slot.
We built the AI, we know exactly what is behind the blacked out play window. We just can't predict paths it took to get to its output 'slot'.
And when they (AI developers with access) do remove the cover: Every weight, every activation, every intermediate value in the network is fully visible and logged if you want it. Nothing is occluded. The problem isn't access, it's that the internals are a few hundred billion numbers. we can see every peg and every bounce, we just can't hold six billion bounces in a human head at once. Nor can we predict what 6 billion (or how ever many parameter count weights it has) will do when you drop in a chip at random points.
"No idea how it works" misreads what's actually being said. It means they can't predict the output, not that the mechanism is unknown. More accurate: They (devs) know the physics, the pegs, everything the chip touches on the way down. They can watch every bounce happen in real time. They just can't predict, before the drop, which of the billions of pegs it'll hit or which slot it lands in.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75404
Location: Israel
|
| Posted: Sun, 27th Sep 2026 21:51 Post subject: |
|
|
|
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12409
Location: Florida, USA
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75404
Location: Israel
|
| Posted: Sun, 27th Sep 2026 22:34 Post subject: |
|
|
BTW, this unpredictability has been true for a while now, with different machine learning models/algorithms, not just "eeeeeiiiiii" "aaaaaayyyyyyeeeee" ( ) LLMs—even with deterministic models, let alone stochastic ones. Face recognition models have been consistently categorizing black men as apes for a decade now, if not more, and the white engineers from Silly-cone Valley, or the police departments that use them, are just as stumped now as they were for the last decade. 
Those face recognition models are just as much "A""I" as LLMs—not at all.
|
|
| Back to top |
|
|
| Page 58 of 58 |
All times are GMT + 1 Hour |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB 2.0.8 © 2001, 2002 phpBB Group
|
|
 |
|
