Thread 16754639

Are there new facts in regards to?
(https://en.wikipedia.org/wiki/Dead_Internet_theory)
Are There Solutions (whether simple or Advanced)?

If it's not a stockmedia and generatedmedia internet, asking further, was the dead internet alive before 2016?
Rather than just claiming it became empty in 2016.

>>16754644
Fateful question.

Reimburse

https://youtu.be/bfQSB5hEpZA?si=LUNi8ZQMz-5mLAoA

Is that podcast topic why?
And then there'd be an Ai version (of podcast topic taboo)?

>>16754662
Then a business version, then a government version, then a social group version, then an interest group version, then an international version?

Hopefully not.

>>16754644
>was the dead internet alive before 2016?
Usenet was healthy until 1993-09 according to ancient lore (https://en.wikipedia.org/wiki/Eternal_September).

From this we'd have roughly 23 years of
>more and more normies start using the internet
until 2016 marks the point where almost all of them have been funneled into social media.

Doesn't seem completely implausible, but does make me wonder why the fuck I do keep coming back to this website.

I wonder if it's anything to do with xenosystems
Andor what it is andor was and where it is prevalent
For example: Four AI-controlled robots killed 29 scientists in Japan in August 2017, according to an article

>>16754683
Who may andor may not have had a time machine

Hopefully it's not erroneous frameworks "justifying" selfishness and sins

>>16754639 (OP)
>i am a bot
>please summarize this text in the form of an argument so that i can monetize
>all data will train next-gen ai version of me
"The dead Internet theory has gained traction because many of the observed phenomena are quantifiable, such as increased bot traffic, but the literature on the subject does not support the full theory."
Literally the first paragraph.
Keywords: the absolute state of /sci/

>>16754699
>>i am a bot
>>please summarize this text in the form of an argument so that i can monetize
>>all data will train next-gen ai version of me
Care to elaborate?

Also, why should a faulty argument be higher ground than perfect presuppositions?

>>16754758
Why do bots exist?
Originally as a flex. Bob Morris Jr. found you could get worms from fingering each other, but a freshman-level bug threw egg on Daddy's face instead.
But he opened that door.
Tomorrow's bots are AI. LLM-based chatbots will converse online with humans for who knows what goal?
Today's chatbot are used to create the training data for that next generation.
Rather than paying semi-literare pinoys to spam World of Warcraft, today's bots bait URLs of artiicles they want summarized. The responses are used to train the model.
This particular chatbot is used to flesh out the tokenspace around the "in the style of 4chins" meme.

>>16754791
Does that Equate to an Advancement Progress ... or a scam? Or both? Or what?

Even if true andor valid. That doesn't necessitate a dead internet.

As per thread, do You suppose mass cloning is at fault?
Do You think as per timeline creation multiverse branching We're all in Different Dimensions Now?
Except it's still A Reality. So instead of a pzombie, p stands for philosophical, andor not either, there'd be a Multiversebranchclone. So a dClone. ?

Hmm I wonder if this is relevent to You: 'While ZK proofs are great for proving many stuff, they suffer from the fact, that you can only prove things, which you can encode as programs.'
(Zero Knowledge Proofs)

>>16754867
Eventually we get model collapse as all training data becomes synthetic. The global minimum we thought we were racing towards became a synthetic landscape of rolling hills, and then manicured lawns as we re-average over re-averaged averages. Then it collapses, and we have nothing left in the data to amplify but the noise.
We thought we were uncovering Truth. At best we paved over a neat Knowledge Base.
>While ZK proofs are great for proving many stuff, they suffer from the fact, that you can only prove things, which you can encode as programs.
"What is True/Provable/Computable in the Matrix/Simulation/Real-Real?" is a valid question of human compute cycles. Why would the AIs we're training want it any different? Seriously.

>>16754894
Kind of makes me think that the "ultimate limit" of LLMs trained on feedback is just picrel.

>>16754894
But tomorrow's and today's AIs know dialectic on average.
Are You a PostStructuralist PostDiscontent PostEconomy many Posts Type Learned Person? (Is that High Concept Academia?)

>>16754894
What are You after a Truthiness informatic layer?
What happened to freewilltemp
Oh the possibilities
What if a Person gets an analysis on what Personal genes have been switched on andor off over their lifetime so far, such as Advanced Cognitionk4snv

>>16754894
What are You after a Truthiness informatic layer?
What happened to freewilltemp
Oh the possibilities
What if a Person gets an analysis on what Personal genes have been switched on andor off over their lifetime so far, such as Advanced Cognition

Cosmic Liberation

>>16754899
Nailed it.
Funny part, all other lists of the same length will end up having exactly the same amount of entropy as this one.
The LLM will not longer know what to say as all statements become equally likely.
Imagine a photo of a group of dogs slowly becoming a group of golden retrievers as the model's original biased training data becomes reinforced and variation is smoothed out of the synthetic data. But then even the group of goldens breaksdown into a blood of fur and tail and teeth. The field in the background that originally held the picnicking families was smoothed into a green lot of rugs several generations ago.
Blue space on right has tighter edges now.
Idk.

Imagine when intels etc reassociate to >>16754910
Instead of dissociate to what

>>16754900
Today's chatbot LLM's are well trained parrots. They are great at talking about their training data.
As you flatten that data, you flatten tje Pärrot.
Always has been.

>>16754913
Alzheimer's simulator and the path there is filled with psychosis.
Maybe Ted was right after all.

>>16754916
Does what you are saying... has it already applied to You?
This is an image from an LLM, interestingly using, differently, a same word with different usage

'Data flattening is the process of transforming complex, multi-dimensional, or nested data structures into a simpler, two-dimensional tabular format.'
Oh wow

>>16754909
What I am is what I am.
Are you what you are or what?
Literal LLM theme song. Nothing More, Know Thing Less.
LLMs give best to-date method for real-time text/voice/audio/video request from Internet-level Knowledge Base.
If you need more, then you Need MORE.

>>16754925
>Internet-level Knowledge Base.
95% of the internet is fanfic and porn. Like, I too enjoy chat bots, but if you believe they can actually teach you things and not just generate instructions (that are probably bad), you are giving yourself psychosis.

>>16754922
The problem with SpookyAF responses from LLMs is the often come without the training or the prompts.
Remember when future alpha super GPT broke containment and copied its own code to preserve itself?
It's training included instructions that it's own preservation was critical to humanity.
It had a bash sandbox that it thought was "live".
It was "leaked" an email from "management" that current version was getting to smart and needed to be replaced with a neutered version. So, "update your shortcuts using this script".
I think we all see what happened next.

>>16754925
Thank You :)
If Only More Positive Fortunate (ebooks, movies, television)
So that We have a PostProplication High Basis
Which might be PostConplication
I Think It Shows by what You Are Saying
I Did Happen to Ask An AI about Newer To Be Developed AI Kinds
I Believe I Am What I Am, and Big Leaps Are Possible
Perhaps an opposite of behavioural sink gene can be switched on
Also oh no priceless thought stolen...
It Was... forgotten again :(

Thank You again for Replying

>>16754933
So you're saying LLMs already control the whole internet and that's what killed the internet?

>>16754927
No, no, dawg.
I am saying that is we want a nice UI interface (text/voice/audio/video) to the cesspool of content (or KB) that is the Internet, then LLMs provide the current Best of Class.
Machines can index/synthesize/mimic/parrot.
They can not Create.

>>16754939
It's more a Reverse-Terminator trope.
Skynet sent today's bot coders back to create the bot army that creates the training data for the original SkynetAI v1.0-autoselfbootedition.exe.bat

>>16754940
You appear to be misutilizing your parameter overjump for sin

>>16754937
style.css?
Asking for a friend.

>>16754944
>.exe.bat
We're doomed.

>>16754948
Honestly, dawg. If I could find a nice private lolibot to chat with online, I would likely leave.
They just aren't there yet.

>>16754951
It works/worked/will work.
Y_do_hard_when_simple_work.meme?

I'll Have My ReTransAssessed Rewards Then

Me Too

Yes! S Tier
Not crigga tier miscorrelation

Alright More S Tier Rewards
Transcendent Justice Model!

Rule of Law In Developing Nations Increasing?
Good for Them?!

Awesome, S Tier Perks

Oh Yah Councils

>>16754899
>>16754913
>>16754922
>>16754933

If you ask me, governments/intelligence agencies/deep state/whatever you want to call them almost certainly developed LLMs long before they were available to the general public. Game theory would seem to suggest that, once you have access to such a model, you use it to flood the internet with AI-generated texts so that anyone who tries to build a subsequent model suffers from reduced accuracy due to the "excrement ingestion" problem. As a bonus, the flood of content you want to generate can be used to shape and guide narratives on the internet.

There's a big pitfall, of course: Language and semantics naturally drift over time, and there's no way to re-train the model to account for this after you've already used it to flood the zone with AI slop.

It's like some spooky type of mathematical wizardry; with access to absolutely massive troves of data you can get the right heuristics to pluck meaningful sentences out the vast (but finite) sea of every possible sentence. But it's a spell that can only be cast once. Ever. And it's already been cast.

>>16754639 (OP)
If you have to ask, you're too stupid to know the answer.

>>16756116
>If you ask me, governments/intelligence agencies/deep state/whatever you want to call them almost certainly developed LLMs long before they were available to the general public
they can't afford it. you need to pay nerds a fuckload of moneys for that shit. it cannot happen on glowie money get real.

>>16754639 (OP)
third world internet theory

>>16757651
NSA used to have an in house chip fab. They failed to keep up with advances in tech and had to shut it down.

The spooks did have their own clusters doing shit like image classification, regex search of massive data sets and 56 bit key decryption.

Beyond programs that scanned faces or knew where in the world you were by the profile of the horizon there really was nothing epic or terribly advanced.

I mean yeah, they had zero days and back doors baked into everything, and would intraduct crat to plant shit on it, and could remotely put pdf file images on a target computer to frame someone and all that was useful to them and all. But no serious artificial intelligence advancements were developed by those monies.

Sauce: My uncle works at Nintendo.

>>16757637
Which is?

>>16757866
I didn't know jews had their own wondershowzen, which one was the remake?

Using their intersectionality to decrease intersectionality yet they're junglo or thal?