|
|
| Page 54 of 55 |
|
|
|
| Back to top |
|
 |
vurt
Posts: 14674
Location: Sweden
|
 Posted: Sun, 13th Sep 2026 13:53 Post subject: |
|
|
| Il_Padrino wrote: | Been working with Fable and it still amazes me every day.
@vurt: any model that's based on backpropagation (working itself back from the answer to the question/prompt) will never lead to AGI. . |
well, again, we are already there (many agrees with this btw).
What exactly, which is intellectual, do you think it can't do?
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12356
Location: Florida, USA
|
| Posted: Sun, 13th Sep 2026 21:09 Post subject: |
|
|
You'd have to make something.
You are talking LLM AI I assume. They are not the type of AI you need. They are language models. They can type, write code, and some VERY basic image processing, and reading webpages and do things on them.
But in the end, they are language/words models. What you'd need is a totally different type of AI.
Think of it like this:
Can my SDXL image generation model play WoW for me? No totally wrong type of AI.
So same here totally wrong type of base the AI 'does' as its purpose. Can't make a language model inject and read a game in memory.
Now you MIGHT be able to get it to code a program what will.
-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 |
|
|
|
|
| Posted: Sun, 13th Sep 2026 21:15 Post subject: |
|
|
| DXWarlock wrote: | You'd have to make something.
You are talking LLM AI I assume. They are not the type of AI you need. They are language models. They can type, write code, and some VERY basic image processing, and reading webpages and do things on them.
But in the end, they are language/words models. What you'd need is a totally different type of AI.
Think of it like this:
Can my SDXL image generation model play WoW for me? No totally wrong type of AI.
So same here totally wrong type of base the AI 'does' as its purpose. Can't make a language model inject and read a game in memory.
Now you MIGHT be able to get it to code a program what will. |
Do you have a job by now? Since you have a hard time finding a job in coding cause off you know what??
|
|
| Back to top |
|
|
zenux
Posts: 3150
Location: lɘɒɿƨI
|
| Posted: Sun, 13th Sep 2026 21:29 Post subject: |
|
|
If math is the language of the universe, and Agentic AIs using LLMs have started to crack theoretical math problems, does that mean they speak the language of the universe now?
i.e. they are getting closer to AGIs?
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75350
Location: Israel
|
| Posted: Sun, 13th Sep 2026 21:37 Post subject: |
|
|
LLM is not AI. And it’s not getting closer to AGI. And neither is stable diffusion AI.
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 13th Sep 2026 22:05 Post subject: |
|
|
| LeoNatan wrote: | | LLM is not AI. And it’s not getting closer to AGI. And neither is stable diffusion AI. |
Ok, i'm gonna react to this for once.
So what is AI?
|
|
| Back to top |
|
|
vurt
Posts: 14674
Location: Sweden
|
| Posted: Sun, 13th Sep 2026 22:13 Post subject: |
|
|
| PickupArtist wrote: | has anyone managed to get an ai model to run and level up in a grindy game
wow forever im looking at u , blizzard dont give a shit about teleporting bots so i wouldnt mind an ai to farm mats n professions or level up n shit for me  |
yes but likely expensive.
"A Japanese RimWorld creator, Apocriton, connected OpenAI GPT-6 Astra to RimWorld and had it autonomously play the game. It actually reached the credits/escape ending, taking roughly 15 hours of real operating time. RimWorld creator Tynan Sylvester subsequently commented that Astra could “play RimWorld to the credits screen all by itself,” calling it an inflection point in AI capability.
The setup was roughly:
Model: GPT-6 Astra.
Interface: a RimWorld mod called RimMolt, which exposes the game to AI through MCP. So Astra wasn't limited to blindly clicking pixels; it could query structured colony information and issue game actions through tools. RimMolt works with clients such as Claude Code and Codex as well.
Instruction: essentially “judge everything yourself and clear the game.”
Result: Astra survived raids and eventually got the colony off the planet, reportedly keeping all colonists and even the original pet chinchilla alive.
Game setup: Cassandra storyteller, Community Builder difficulty, Ironman, no DLC, with mods that did not materially change gameplay."
Rimworld is extremely hard, i have thousands of hours, never finished it. Super complex game too.
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 13th Sep 2026 22:17 Post subject: |
|
|
AI knows all the tips and tricks. Just google what you have to do in a game and there you go...
But still, putting all the things together is impressive. It doesn't do a better job then us avarage gamers.
|
|
| Back to top |
|
|
vurt
Posts: 14674
Location: Sweden
|
| Posted: Sun, 13th Sep 2026 22:18 Post subject: |
|
|
| DXWarlock wrote: | You'd have to make something.
You are talking LLM AI I assume. They are not the type of AI you need. They are language models. They can type, write code, and some VERY basic image processing, and reading webpages and do things on them.
|
they are far more capable. that's the thing with the new models. they can even draw, give them an image and tell it to draw it, they can do it. its really cool.
they can work in blender and make 3D models and rig models, they can use godot.
they can download mods and install them, i hate doing the more advanced modding setups so i just use an AI for it, they will troubleshot it too if something doesnt work.
| friketje wrote: | AI knows all the tips and tricks. Just google what you have to do in a game and there you go...
But still, putting all the things together is impressive. It doesn't do a better job then us avarage gamers. |
yes its way easier for an AI, like with anything. They can do things much faster too.
Last edited by vurt on Sun, 13th Sep 2026 22:20; edited 1 time in total
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12356
Location: Florida, USA
|
| Posted: Sun, 13th Sep 2026 22:20 Post subject: |
|
|
| friketje wrote: | | Do you have a job by now? Since you have a hard time finding a job in coding cause off you know what?? |
Yea. Found one a few weeks ago. AI wasnt the issue, it was either remote jobs get 3000 applications, so they can be SUPER picky. Or the requirements for jobs was absurd. Like listings for a job wanting 10+ years experience of this and that. Or 10 certs listed as 'nice to have' for a job you MIGHT use 3 of those certs, and the other 7 once every few years in a pinch.
And so many of them have "Required: Bachelors in computer science (of some form)". I'd say 1/2 the software guys I know don't have that.
-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 |
|
|
|
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75350
Location: Israel
|
| Posted: Sun, 13th Sep 2026 22:28 Post subject: |
|
|
| friketje wrote: | | LeoNatan wrote: | | LLM is not AI. And it’s not getting closer to AGI. And neither is stable diffusion AI. |
Ok, i'm gonna react to this for once.
So what is AI? |
Something that learns and improves with time. And I don't mean something that gets replaced by others from time to time, having cost millions if not billions to iterate.
https://en.wikipedia.org/wiki/Intelligence
Go over the various definitions of what "intelligence" means. None of that is found with LLMs. It's an illusion of intelligence by a pachinko machine, essentially. Dumbos like vurt getting impressed. But there is no actual intelligence there, let a lone a general intelligence or "super" intelligence. |
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
vurt
Posts: 14674
Location: Sweden
|
| Posted: Sun, 13th Sep 2026 22:48 Post subject: |
|
|
funny how LeoNatan thinks AI is just "intelligence" and not "artificial intelligence", you know like in the name? But i am not the least surprised.
AI is not Human Intelligence, never will be, because its not human. I think pretty much a 6 year old knows it, but not little Leo.
|
|
| Back to top |
|
|
|
|
| Posted: Sun, 13th Sep 2026 22:52 Post subject: |
|
|
| LeoNatan wrote: | | friketje wrote: | | LeoNatan wrote: | | LLM is not AI. And it’s not getting closer to AGI. And neither is stable diffusion AI. |
Ok, i'm gonna react to this for once.
So what is AI? |
Something that learns and improves with time. And I don't mean something that gets replaced by others from time to time, having cost millions if not billions to iterate.
https://en.wikipedia.org/wiki/Intelligence
Go over the various definitions of what "intelligence" means. None of that is found with LLMs. It's an illusion of intelligence by a pachinko machine, essentially. Dumbos like vurt getting impressed. But there is no actual intelligence there, let a lone a general intelligence or "super" intelligence. |
I think AI is a mixed bag. Some of it is genious, other stuff is rather stupid. But it's still improving. It's not todays capabilities that matter. If AI progress stalls for a couple of months, then you are right. But as long as thing keep on improving, I would be cautious underestimating what it can do in the future.
|
|
| Back to top |
|
|
vurt
Posts: 14674
Location: Sweden
|
| Posted: Sun, 13th Sep 2026 22:57 Post subject: |
|
|
he hasn't got a single clue how LLM's work, ZERO. yet he keeps yapping about it. what a total tard. if you're so interested in debating, or in AI's and LLM's overall, why not research it first, learn the BASICS? But nooo, he really wants to come of as a total little dumb assclown.
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75350
Location: Israel
|
| Posted: Sun, 13th Sep 2026 23:01 Post subject: |
|
|
| friketje wrote: | | LeoNatan wrote: | | friketje wrote: |
Ok, i'm gonna react to this for once.
So what is AI? |
Something that learns and improves with time. And I don't mean something that gets replaced by others from time to time, having cost millions if not billions to iterate.
https://en.wikipedia.org/wiki/Intelligence
Go over the various definitions of what "intelligence" means. None of that is found with LLMs. It's an illusion of intelligence by a pachinko machine, essentially. Dumbos like vurt getting impressed. But there is no actual intelligence there, let a lone a general intelligence or "super" intelligence. |
I think AI is a mixed bag. Some of it is genious, other stuff is rather stupid. But it's still improving. It's not todays capabilities that matter. If AI progress stalls for a couple of months, then you are right. But as long as thing keep on improving, I would be cautious underestimating what it can do in the future. |
Define improvement. Models improve, yes, but so is their hardware requirements and hardware investment growing. I am not disputing LLMs are not impressive, or saying they are useless tools. It's just that they are as much of an "AI" as a switch or an if statement.
Dumb cumtwats like vurt don't even understand the discussion. They think "tool does something impressive, more impressive than I can do, tool intelligent hurr durr" |
|
| Back to top |
|
|
vurt
Posts: 14674
Location: Sweden
|
| Posted: Sun, 13th Sep 2026 23:02 Post subject: |
|
|
so you are impressed by a "pachinko machine" now all of a sudden? you make fun of me for being impressed, yet you say you are impressed.
you are making zero sense.
HW requirements are going down btw. the models we can run today locally, they are so much better, we could only have dreamed of running something like that 3-4 years ago locally.
Last edited by vurt on Sun, 13th Sep 2026 23:08; edited 1 time in total
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75350
Location: Israel
|
| Posted: Sun, 13th Sep 2026 23:03 Post subject: |
|
|
I'll quote my edit for fun.
| LeoNatan wrote: | | Dumb cumtwats like vurt don't even understand the discussion. They think "tool does something impressive, more impressive than I can do, tool intelligent hurr durr" |
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75350
Location: Israel
|
| Posted: Sun, 13th Sep 2026 23:05 Post subject: |
|
|
"Calculators sum and divide numbers faster than I do, omgz0rz, AI!!11!"
|
|
| Back to top |
|
|
vurt
Posts: 14674
Location: Sweden
|
| Posted: Sun, 13th Sep 2026 23:09 Post subject: |
|
|
coming from someone who missed the A in AI. keep trying. you are totally clueless.
why didnt you solve that math problem with a million in price money instead of the "stupid AI" lol...
|
|
| Back to top |
|
|
Il_Padrino
Posts: 8158
Location: Greece by the North Sea
|
| Posted: Mon, 14th Sep 2026 00:05 Post subject: |
|
|
Intelligence is intelligence, whether it's human, animal or artificial. Saying AI is "different" sounds to me like excuses for the many dumb things it still does.
I can't find a good way to put it into words why, but cats or dogs, for example, are intelligent. They are beings. They can't do math or write or talk. But they are still considered intelligent. Or do you disagree with this?
@vurt: to answer your question about an intelligent thing an AI can't do: actual reasoning, I think.
It's very good at finding assocations and working from those, but that's not the same (although they can appear similar).
While I've been building my app with Claude Fable, there are often design decisions where it makes suggestions that are very complicated (but probably would have worked), while missing the straightforward solution in front of it. Because it doesn't really reason or question itself. Unless it gets an input trigger from outside (eg you).
The video of the AIs setting up a lemonade stand should have been pretty clear that they are not intelligent.
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.
Last edited by Il_Padrino on Mon, 14th Sep 2026 00:14; edited 1 time in total
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 75350
Location: Israel
|
| Posted: Mon, 14th Sep 2026 00:12 Post subject: |
|
|
Reasoning, learning, evolving understanding of the world.
A dog steps on asphalt in the scorching sun and runs by instinct to shadows. The next time it will refuse to walk on the asphalt. Ask an LLM some question and then correct it. Start a new session and it will make the same mistake. Sure, you can add "memories", but then you are wasting its "context".
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12356
Location: Florida, USA
|
| Posted: Mon, 14th Sep 2026 00:33 Post subject: |
|
|
Id disagree, the dog sometime run around like an idiot. gets on the grass and is OK. And will do it again later. And what it learned doesn't not propagate to other dogs.
Either way the dog learns by training (self guided learning, or outside). LLM learns by training (self guided, or outside). And it propagates it to the other internal agents on its own.
Your is is based on reactionary physical survival stimuli.
Better example is you can grab a dogs paw, shake it. And next time it doesn't know what you are doing the second time. You have to take time to train it outside of expected results to get those results.
(not same same, but less flawed than your example.)
To make yours MORE correct. After many 'previous model versions' of dogs (since before they was dogs or even mammals), Dogs have learned, by trial and error and process of elimination of ones before it that died from heat and didn't breed to make more that do the same thing to die...that 'future versionings' of dogs got smarter about what to do when encountering heat. It didn't learn it on its own. It was already 'baked' in from last alpha versions of dogs.
While I get your point. Its apples and oranges you are weighing, and then going "See these are NOT apples!"
-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: 75350
Location: Israel
|
| Posted: Mon, 14th Sep 2026 00:54 Post subject: |
|
|
Intelligence is not about cultural or evolutionary muscle memory and instinct. That's "the definitive underpinning" if you will A baby is born tabula rasa with very little understanding. Throughout its life, it will learn, form understandings, and be able to reason about new experiences (although, with gen ztards, that might be disputed ). Its learnings in a vacuum do not travel through its genes, but it did learn and its world view advanced even without a generational or cultural baggage.
So yes, models advance. But that is a testament to its engineers' intelligence, rather than the LLM's.
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12356
Location: Florida, USA
|
| Posted: Mon, 14th Sep 2026 01:21 Post subject: |
|
|
True. but we cannot compare ingrained traits to intelligence either.
Intrinsic knowledge that comes from genes (or wherever we are still unsure how it passes). Like even a bird taken at birth will know which bugs to not each that appear to have poisonous colors. A dog never near the water before will flap his paws to swim if held over it (or in it). A cat doesnt need to be taught how to twist his body to land on its feet. By the time it can strut on it's own it just 'knows' how to.
Or prey animals, even captive from birth. Know which predators are the ones that go for them and will fight/flight being shown those specific ones.
So the line is there, and your example was one of them. It wasn't a trained knowledge, it was an instinctual one to dangerous stimuli.
I DO get your point, just saying it was a bad example that had holes someone could poke into it. is all.
-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: 75350
Location: Israel
|
| Posted: Mon, 14th Sep 2026 01:25 Post subject: |
|
|
The example had two parts. The instinct was to run to the shadow ("how does it know?"). The learning was not to go on the scorching pavement or asphalt.
|
|
| Back to top |
|
|
DXWarlock
VIP Member
Posts: 12356
Location: Florida, USA
|
| Posted: Mon, 14th Sep 2026 01:28 Post subject: |
|
|
I got ya.
I think it's only because we don't let it/configuring it to. Not that it cannot.
Lots of AI (non- LLM) learn on its own. The ones given a ragdoll body and told 'learn to walk' is a good example of those.
It will find many ways to walk or get around given its body type. Some people even give it two or more criteria like: Learn to walk, in the least energy expending way'.
Protein folding AI is a great example of it. It learns from its past attempts, mistakes, and new information about molecular bonds it figured out.
-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 |
|
|
| Page 54 of 55 |
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
|
|
 |
|
