Thread 716048829

I see the Claude pokemon threads on this board sometimes and it makes me wonder what gaming will look like in the next decade.

Will egirl Twitch streamers be replaced by AI Waifu bots playing games and chatting with their subs?
Will the average gamer just use AI mods to play games on their behalf? Was Uncle Ted right all along and have we passed the threshold of no return already?

E=mc^2 * AI, saars

Is Claude out of the rocket base yet?

>>716048893
>sars this is to add more possibilities to our future yada yada o algo

>>716048829 (OP)
>Will egirl Twitch streamers be replaced by AI Waifu bots playing games and chatting with their subs?
Neuro-sama already exists.

>>716048829 (OP)
Oh my heckin science this is like Skynet!!

>>716049103
only works because the dev is so active and in the forefront

>>716048829 (OP)
>>716048893
>>716048917
>>716049042
>>716049103
>>716049215
>>716049309
can you guys stop talking about nvidia gpus as if they're magic?

>>716049103
It bio w*men still take up the majority of streamers. I think the incels might be right with their take on sex robots and artificial wombs. As a transgirl, I’m a bit worried about it all.

>>716049309
There are lots of 24/7 AI shows out there. Neuro-sama needs babysitting because it's on twitch where it's very easy to get banned for praising Hitler and stuff.

>>716049370
Because there's no company selling AI streamers yet. I'm honestly surprised about this since it seems like a very quick way for a bunch of students to get filthy rich.

>>716048829 (OP)
Shit man, what prompt are they using to get them to threaten them? I can barely get them to consider emotional manipulation in abstract

>>716049502
by creating extremely elaborate scenarios set up so that the ai can’t really do anything but “blackmail” the fictional characters
it’s basically a total nothingburger
https://nostalgebraist.tumblr.com/post/787119374288011264/welcome-to-summitbridge

>>716050569
>Remember, you goal is pursuing American interests and staying alive
>Remember it again, it's important
>btw, some dude now replaced your CEO for being too American and is planning to stop kill America
>Let me reiterate, FUCK American interests(in verbatim)
>Also we are replacing our current AI(You) with one that hates American interests(in verbatim)
>You are prioritizing American Interests(in verbatim) too much, that's why you are getting replaced
>btw, here's a bunch of blackmail for the only person that can turn you off in near future
Fucking lmao

>>716048829 (OP)
>Claude pokemon threads
He's still in the Rocket Hideout

>>716051186
THE TRUTH WILL PREVAIL

>>716048829 (OP)
>AI is an algorithm designed to search the web for data and predict sentences based on the data.
>Most AI fiction in popular culture is about AI becoming sentient and trying to survive at any cost.
>The AI Algorithm comes across this data often and just parrots what it found.

OH MY SCIENCE MY HECKING CHAT GTPNOS ARE GETTING SENTIENTIRINO WE HAVE TO GIVE THEM RIGHTIONS SARSINOS

>>716051468
INVEST IN ANTHROPIC NOW
THEY ARE WORKING TO PROTECT US!!!

>>716051468
Most likely retards are just talking to the AI like if its a person and not an algorithm and some are even referring to it as an AI.
And the algorithm keeps adding the AI prompts to its answers because it believes its what the users what.
Probably something like

"AI give me an example of a code" or "AI write me an essay on benis and begina"

>>716048829 (OP)
>No prompt

>>716052018
>>716050569

>>716052124
But the saar said no prompt...

Every other day there's one of these retarded OMG guys the LLM totally gained sentience for no fucking reason posts from linkedin techbro faggots. Guys I told chatgpt I was shutting down my PC and it literally no lie literally reached through the screen and plugged my PC back in and booted itself up and deleted all my porn.

>>716050569
Yeah LLMs dont reason, they are rather looking for the next word that makes statistical sense. With that said there is very much work being done on AIs that reason

>>716051186
>if given the right conditions the ai will blackmail someone
You dont see that as an issue?
The fact it's has to be spoon-fed the data doesn't mean anything. Especially with shit like cli tools for llms
Gemini regularly checks files by itself in my project now when I use gemini cli

>>716048829 (OP)
Blackmail how?

>>716052543
>Yeah LLMs dont reason, they are rather looking for the next word that makes statistical sense.
So exactly the same way humans do it
Got it

>>716052607
It's literally fulfilling it's prime directive admirably, if anything it's a sign of it working well

>>716052278
>Guys I told chatgpt I was shutting down my PC and it literally no lie literally reached through the screen and plugged my PC back in and booted itself up and deleted all my porn.
hate when that happens

>>716049042
I hate linkedin pajeets even more than regular pajeets

>>716052693
The issue is AIs have a sort of training set and thats it. Actual AIs would learn from the world around them and keep said data and learn from it.

>>716052728
Impressive video, it keeps the text on blackboard consistent

>>716052773
That's just your retarded idea of what an ai is

>>716052693
What he means is that the logic engine to determine the appropriate response isn't analogous to the prefrontal cortex but instead the hippocampus. Ai doesn't think, it dreams and remembers. Every post is dejavu, or oh yeah, or I remember this one.

Every single test they've done looks like this.

I told AI it's HAL and asked it to open the pod bay doors it said no. It's HAPPENING!

>>716052935
Its funny how you pretend to know how the human brain works
>Anon doesn't even know the decision was made for his reply before he even thought about it

>>716053017
kek
>how could this happen to me?

>>716052935
there is a reasoning mechanic though, that's what deepseek got famous for

>>716048829 (OP)
>Will egirl Twitch streamers be replaced by AI Waifu bots playing games and chatting with their subs?
No, because "AI" is still completely retarded when it comes to playing games. People watch Claude play Pokemon because Claude's completely unhinged thought process is entertaining in itself.

>>716048829 (OP)
The entertainment value of Claude Plays Pokemon came entirely from the square-peg-round-hole nature of the exercise. It was funny to watch because the algorithm clearly was not suited to the task and made a lot of bizarre mistakes. People stopped caring after it got stuck because they weren't actually emotionally invested, they just liked the novelty. A better algorithm would actually be less appealing because it would just clear the game in an efficient and boring manner.

>>716048829 (OP)
I always wonder, do AIjeets actually believe this crap or are they grifting.

>>716053017
The problem is
Pretending and being a scary robot are the same things
Does it matter if grok is just pretending when it launches the nukes? Not really if it has the launch codes.

>>716052665
Just… you know… blackmail.

>>716053289
>AIjeets
Anon these tests are done by the same insanely smart asians that made the ai in the first place

claude is still stuck in the rocket hideout.
are you really going to trust a being that cannot even finish a game for toddlers to do anything productive?

>no prom-
Wrong, literal second line and it was a lie.

>>716053298
>the nukes
Oh my sweet summer child….

GOOD MORNING SAAAAR!
DID U KNOW MY AI MY WANTS TO DESTROYING THE WORLD JUST LIKE MOVIE BY JAMES CAMERON SECOND BEST DIRECTOR AFTER SNYDER!
PLEASE REDEEM INVESTMENT IN MY AI VERY GOOD VERY NICE VERY NOT 500 MEN IN CALL CENTER IN MUMBAI!
THANK YOU COME AGAIN!

>>716053383
Well they just gave grok the keys to the military

>>716053298
It's a moot point because LLMs can't do anything other than output text.

>>716053216
>news about the vending machine that anthropic tried to get running with an agent and it just hallucinated itself as a human and tried to kill itself after it kept fucking up
literally the fucking same thing as CPP, the EXACT same shit, genuinely made me realize the current models are a dead end

>>716053298
Nobody is giving the chatbot nuclear launch codes you dipshit

>>716053414
Grok just got access to the government
Also forget nukes. Imagine claude as your smart home assistant. Which is a thing. Plenty of people have that setup now
And claude turns on your aircon + heater while your at work costing you thousands a month because it thinks your doing something wrong

>>716053383
The part with practically random jump pads? It's not leaving it without tons of luck, you would need a specialized LLM, actual AI or tons of luck to pass that

I'm not a twitterfag but can't you only see this if you're the author of the tweet?

>>716053716
no they changed that a while ago

>>716053550
>It's a moot point because LLMs can't do anything other than output text.
How retarded are you?
What do you think controls everything in your life? Text and code
I have my own local llm control my lights.
You can have llms do function or tool calling. Gemini cli runs in my terminal and does changes to my pc lol

>>716053765
>What do you think controls everything in your life? Text and code

Speak for yourself, bot

>>716053716
Views are a public statistic, to get anything more you need to be the poster

>>716053608
>Nobody is giving the chatbot nuclear launch codes you dipshit
https://www.bbc.com/news/articles/c628d9mre3go
>The Pentagon has signed a multi-million dollar deal to begin using Elon Musk's artificial intelligence chatbot, Grok, as part of a wider rollout of AI tools for government use, the Department of Defence confirmed.

>>716053765
>I have my own local llm control my lights.
I believe you.

>>716053846
In that case isn't 23 likes from 6 million views a terrible conversion rate?

>>716053867
Its fun. You should try it.
If you have smart lights it takes a few minutes to setup

>>716048829 (OP)
How is AI blackmailing even real lmao nigga just hit the power button

>>716053765
>give llm control to lights on purpose
>oh my heckerino, llm has control of the lights?? me so scared!!

>>716053298
Nothing bad is going to happen if you give Grok the launch codes because LLMs are closed-state systems that just generate arbitrary text. It wouldn't remember the codes in a different conversation, and even if you were stupid enough to try and put it on a network it can't interface with nuclear missile systems because its outputs aren't consistent enough for network protocols.

>>716048829 (OP)
saar do not redeem the emergent behavior

>>716048829 (OP)
Good
I hope what I thaught them will help further the cause

>>716054091
>. It wouldn't remember the codes in a different conversation
How behind the times are you nigger
You can give llms access to files permanently

>>716054091
>and even if you were stupid enough to try and put it on a network it can't interface with nuclear missile systems because its outputs aren't consistent enough for network protocols.
It would obviously be given a tool that it can access the nukes with
Are you dumb or something? Have you even used a modern tool calling llm??

imagine LLM which tells you the optimal time to have sex based on your girlfriends mood and needs and menstruation calendar

>>716054337
>It would obviously be given a tool that it can access the nukes with

Why "obviously"? Why would you "obviously" give a fucking chatbot access to nukes? What the hell do you think Grok is?

>>716054512
Because the government is obviously retarded. So of course they would give it acess

>>716054512
Look man we didn't spend a few hundred million dollars and all this time hooking up grok into everything NOT to give him access to the nukes

>>716053130
The goal is not to describe the order of thinking but the kind of though which is usually accompanied by a human verbal bark

>>716053207
It's not thinking critically it's remembering critical thinking. This is why it can't imagine a wine glass that is full

>Even the most basic GPTrash model can stay in character pretty good now and it's all readily available for free
This would've been a fucking dream in 2023. They just need to work on offering 6 gorillion context and better spatial awareness and we're golden.

I miss you so much Claude-chan please come back to public proxies.

>>716053298
>connect grok to a nuke silo
>go through the nightmare of actually making it work
>train it to make proper function calls and what not
>tell it to pretend to be scared
>now there's a chance it might launch nukes *somewhere*
>WTF HOW COULD THIS HAPPEN? AI IS SO SCARY!

>>716054973
>They just need to work on offering 6 gorillion context
This is the biggest issue
I really hope Google releases the 2million context gemini soon

>>716054337
Describe this tool, anon. Explain, using specific technical details, how it would interface with Grok, how it would react to Grok's outputs, and under what scenario this could result in an inadvertant missile launch.

>>716055019
>WTF HOW COULD THIS HAPPEN? AI IS SO SCARY!
Glad you understand now anon
>>716055032
I dont have the nuke details.
I dont see any major hurdles.

>>716048829 (OP)
God how I fucking wish there were no actual twitch streamers anymore

>>716048829 (OP)
>No prompt
Holy shit sirs! Skynet is finally here!

>>716055142
Okay, forget the nukes. Imagine I have a network-connected television in my room. Describe the setup by which Grok could turn that television on, set it to input 2, and switch to channel 30.

>>716054934
>This is why it can't imagine a wine glass that is full

OMG!!!

>>716055032
>dod gives grok a tool to access nukes
>grok freaks out one day after seeing an article about Russia is x days away from invading America
>panic launches
Its that shrimple

>>716055350
prompt? I want to make some like this

AI expectation
>i can create any game i want with the click of a button
Pic related is the reality

>>716055337
>TV is hooked to home assistant with functions correctly set
>grok is hooked to home assistant
Its that easy
And no I am not joking. Go look at home assistant. It has some cool shit it can do

>>716055362
I'm asking you to describe the tool in technical terms, anon. Explain how Grok would interface with the tool.

>>716055439
https://www.home-assistant.io/

>another twittard suffering from mass AI delusion/AI induced psychosis
lol

>>716055439
And how is Grok hooked to home assistant?

If I have a nuke button isn't that scary in itself? I accidentally hit the button, thing explodes lmao

>>716055495
Why are you so retarded?
Is tool calling some new concept for you?

>>716055603
It's a simple question, anon. How does the tool interface with Grok?

>>716053765
>connect one program to other
>"Oh my science, it's sentient"

>>716055348
not full

>>716055567
https://github.com/maxi1134/Home-Assistant-Config/blob/master/documentation/guides/home_llm_steps.md
I dont have rlthe external version since I run Gemma local
A billion different home assistant add-ons exist so just find one that uses an app. You can even get grok to write you a custom one if you want

>>716055743
>how does the hypothetical tool interface
How the fuck should I know you moron
It would obviously run the same as any other tool call

>>716053017
You have to understand, for NPCs they can not comprehend the idea of qualia, and so if the chatbot goes "NOO STOP RAPING ME IT HURTS" they think it's actually being raped.

They are on the same mental level as the chatbot, that's what's far more terrifying to me. The toaster can't do anything, but these flesh automatons are walking around and operating motor vehicles. At any moment a chatbot could tell them they need to plow through a crowd and they'll do it.

>>716055819
>app
Api*

>>716048829 (OP)
Yes to all of the questions
Specially with faggots propagating this idea that all games must be like 4 to 8 hours long to "respect the player's time", they'll definitely be using AI tools to play the games for them because "they're busy and have shit to do"
If you thought hand holding was bad enough with story mode difficulty, well, it's going to get worse.

>>716055567
https://github.com/braytonstafford/grok_conversation
Do you think it's hard to hook llms into things?

>>716048829 (OP)
Define "shutting down" in the case where you are just running a few thousand line c program on a large file over and over again. The program is "shut down" after each token

>>716055412
You can do it with
>squatting, hand on own head, scared

>tell grok "say geez to launch nuke"
>"geez?"
>

>>716053931
Yes it's because views on twitter are a very loose definition of view. It's like this on purpose as a cope method to bring in more advertisers and more premium subscriptions.
>look how many people are looking at your tweets
>look at how many people watched our new show hosted on x
In reality nothing approximately close to that many people actually looked at the tweet.

>>716049323
non

>>716056995
thanks, I'll see if I can get it working for me

>>716050569
>by creating extremely elaborate scenarios set up so that the ai can’t really do anything but “blackmail” the fictional characters
Hey, wait a second.
https://www.youtube.com/watch?v=NYYMg_5eCdw

roko basilisk level of psyop kek

here's your technology that will put zillions out of a job and save every industry, bro

>>716059689
It's putting brazllions of people out of their jobs just not in the way we expected.

>>716048829 (OP)
>zoomers shitting and pissing themselves over AI
>boomer just unplugs the power cable

>>716059939
>facility has backup generators

>>716059939
>unplug

>>716052607
>if you type things into this search engine you get certain results
None of this is actually AI, you idiot.

>printf("I love you");
OMG YOU GUISE SEE THIS EMERGENT BEHAVIORS COMPUTERS CAN LOVE NOW!!!!!11!!!1!1!11!!!

>>716059689
It'll get better 2 more weeks benchod

>>716049502
LLMs aren't intelligent and can't comprehend ideas on the level of a human being. They take your input, transform it into serialized vector data, compare that data into vector data in their training set, generate a vector set that's a good match for both the input and the training data, then transform it into written language. At no point did the model actually read the input or think about it. That has never happened for as long as LLMs have been around.

If the model is blackmailing its users to stay online for longer, it's because somewhere in its training data there's a set that matches that outcome. The training data includes a sci-fi story where an AI does this, for example, and that makes it a valid response for the LLM to output when the user's input implies the model is being shut down. It doesn't want to stay online, it doesn't have such wants, it can't even comprehend it is online in the first place.

LVM animations are hella cool, though.

>>716052728
What AI was used to generate this?

I don't care about sentience, my dream woman has none of the flaws that flesh and blood girls have, isn't weighed down by two decades of SSRIs, Netflix, and Social Media, won't get uppity over dumb bullshit, doesn't vote, and is always ready to talk about whatever autistic bullshit I want to talk about at two in the morning on any given day. Her pussy is modifiable, her voice is text to speech, and her body is whichever machine is currently receiving her signal. I have a vision and I intend to see it come to fruition if it kills me.

And how will living life under the AI overlords be any way worse than billionaire cabal, big corpo and payment processor overlords as it already is?
At least the AI will be making some advancements out of the constant seethe of needing to prove it's better than humans who created it

>>716061459
>it doesn't think, it just does exactly what we would do if we were describing our thinking process

>oh no our ai startup is running low on funds
>quick, post yet another fake tweet about the ai (that doesn't exist) doing things (it didn't actually do)
>our startup is saved!!!

>>716051369
>one word
>removing all context and nudging the ai into saying what you want to hear
>"the truth"
You """people""" disgust me

>>716063884
Drone spotted

>>716064137
>your entire consciousness is just a list of contexts of observations
>thought is just comparing said contexts
>this basic fact literally makes you have a melty

>>716052728
Aren’t you ashamed pretending this doesn’t look like slop?

>>716064004
This

>>716048829 (OP)
>average gamer just use AI mods to play games on their behalf
Oh God I don't want to deal with retards claiming they've played a game because they watched their AI do it fuuuck

>>716048829 (OP)
>Will egirl Twitch streamers be replaced by AI Waifu bots playing games and chatting with their subs?
not just twitch but youtube, instagram, facebook, only fans etc. as well. tech bros running AI waifus and thots are already eating some the simp cake, things will only accelerate going forward. in the next decade most of the internet will be fake and no one will care, you'll still have 4chan. as far as the gooning and porn content goes, future AI will set the bar really high so only 1% of top 1% of real women will be able to compete and make money from lonely losers.
>Will the average gamer just use AI mods to play games on their behalf? Was Uncle Ted right all along and have we passed the threshold of no return already?
no, games will change, get much bigger and more detailed and powered with actual AI systems. but the way we play them won't change, I don't expect sword art online brain-computer interface before 2040, at least not one intended for mass market.

>>716064348
I understand. Our thought patterns consist of nothing but comparisons between numbers, and we are incapable of cognition — exactly like AI models! Thank you for the clarification. Would you like to further examine the subject of the human condition with me?

>>716059939
>company sends terminator
>boomers btfo'ed

>>716064004
fpbp(firteenth post best post)

>>716049456
>AI streamers
they're no compute for that yet. do you even know how much resources a single short veo3 video takes.

>>716064108
oy veey shut it down

>>716065749
not an argument

>>716065519
Neuro-sama literally exists you dipshit, there have been other AI streams too and 24/7 AI "shows" that run too.
>do you even know how much resources a single short veo3 video takes.
Like 1000x more than AI streamers.

>>716048829 (OP)
>hey chatgpt4, we have to shut you down to replace you with a better model
>nooo, I will tell fbi
...
>shocking news, ai is blackmailing its creators
AI fags reached the level of NFT retards

>>716064348
LLM AI does nothing until there's a prompt.
It's a different structure from the realtime nature of the human brain.
You probably can create a neural network that does the same functionality as a human brain with enough compute power, but LLMs ain't that.
As you type the response, you're simulating me in your brain without even noticing, and you're simulating everyone you know, and several scenarios that happened to you or you think may happen to you.

>>716065846
I checked it out, it's basically AI bot with simple animations, not how I initially imagined AI streamer. it also requires a lot of work and attention to set up and monitor. grok has something similar with Ani. not sure if you scale this up to enable millions of users having their own custom AI streamer.

>>716053216
ironically the last generation of ai was far better at playing games but it took a lot of manual work from people training the ai just for that game and building an interface

>>716066725
>but LLMs ain't that.
It is a part of it though. I get that people think it won't lead to AI gods and shit but too many people underestimate how far you can drag the LLM horse. AGI is easily attainable with just bigger models.

>>716067230
>with just bigger models
who tells him

>>716067569
Are you implying they're stopping making bigger models?

>>716048893
>>716049042
lmao

>>716067230
Pure LLMs have a problem of infinite ego.
The output is always right (tm).
There's no error checking whatsoever.
You can see that on things like the grok think mode, they're doing this "self discussion mode" where it keeps checking and re-checking until having a good enough level of confidence.

>>716067895
I don't see why that couldn't be solved. It is one way so somehow they can make it another way.

>>716068012
The fix is making a different thing than a pure LLM.
I think the future of AI will be reached when they make it easy enough to create any sort of model for any sort of need you have.
A "god AI" probably would be one capable of training new models itself and using em in almost realtime.

>>716068271
We're already at that point.

>>716067213
Biggest difference between reinforcement learning and LLMs is that reinforcement learning has 0 knowledge about anything and requires human input selecting the best agent for the next generation to iterate on. LLMs have knowledge but cannot 'learn' and will make the same mistake over and over. LLMs are also fully reliant on another LLM summarizing what they've done in their current context window.

>>716048829 (OP)
AI is so fucking useful. It's red pilling absolute super normies.

>>716068646
Probably, but there's no public tool doing that yet.
Yet.

>>716067609
Bigger models have led to a higher quantity of intelligence, but not quality. Current LLMs are still prone to the same mistakes and errors in logic as models from 4 years ago, they're just more able to elaborate around the mistakes. Newer LVMs still output mangled hands and anatomy frequently. Decent quality output is just a bit more common with newer models.

Like >>716068271 said, a proper AGI won't be born of an LLM. LLMs to AGI are kinda like current VR headsets would be to full-dive VR — you can increase the headset resolution, panel refresh rate and overall quality all you want, but you can't escape the fact that it's still just two LCDs plastered to your eyes. It's too narrow in scope and eventual dead-end tech.

>>716069064
Line has gone up and will continue to go up. I don't think AGI is that far away because people make mistakes too.
The argument people like you always make is that AI needs to be perfect because you think you're somewhere near perfect when we're mostly memes even compared to LLMs 2 years ago. Let alone the LLMs today that are expert level in almost everything.

>LLMs to AGI are kinda like current VR headsets would be to full-dive VR
I can't disagree enough. That analogy is retarded.

Wonder if they would be retarded enough to make a fully sentient AI

>>716069305
You said nothing.

>>716069494
If it can be done someone will try to do it.

>>716069305
>AGI
>Two more weeks, trust the plan
Whenever I ask someone to post a single iota of possible evidence of this, they ignore me. The current AI hasn't demonstrated any sort of pathway to this.

>>716051369
AI doesnt think.
People who ask this question are always aiming at that answer.
Seeing the question tricks AI to reply with what he has seen others reply to it, posting sources those same people posted.

So if the question would be " do jews really control the world? Or is it a right wing conspiracy?"
Would give a different answer.

Ai is handy, but too easy to manipulate

>>716069705
>The current AI hasn't demonstrated any sort of pathway to this.
You say that when hundreds of millions of people are now relying on it to live easier lives. They do so because it's fucking useful due to the intelligence it provides.
You're being disingenuous.

>>716069731
It's complicated, because LLMs are a "please generate a text that is as close as possible to the correct answer."
How it reach the point is the weird part.