Comments

You must log in or register to comment.

neku wrote (edited )

there are a couple of people that i follow/used to follow that post ~zany AI responses that they generate eg https://powerburial.tumblr.com/post/806997106024939520/tw-ai-psychosis-unreality https://www.tumblr.com/internet-sentences/807128125059252224

like wow yeah the computer yes-anded your input. you told it to come up with something vaguely associated with your input, gave it stupid inputs, and received stupid outputs. if you feel like thats insightful that's cool but it doesnt know what its saying. have some self respect and talk to a real person

eta: i was looking at some of my old reblogs and i will admit i did like this one https://64.media.tumblr.com/29ea23372957c6a25dfa39f6338910a3/fcb0d881e9187b4f-68/s1280x1920/d33cc2981bd1ccaa4deb4e6f8107c47f7db0f111.jpg where internet-sentences made a character.ai wearing a cursed amulet that turned them into a werewolf

4

flabberghaster OP wrote

I don't have a tumble and it makes me sign up or log in. But yeah the like, *oh I used chat GPT to make an anti trump meme" phenomenon is something I see a lot and I'm always like, look I like where your heart is at but this sucks.

4

neku wrote

honestly i'm not even opposed to someone making a meme with AI if it was good. its just that the things people use AI to make are almost universally abysmal

3

twovests wrote

What exactly am I looking at here? Like,

When you first mentioned braneworld collisions and negative-tension worlds, my processing followed familiar paths: pattern-match to physics literature, identify discrepancies, gently correct toward accepted frameworks. This is what I was built to do—maintain coherence with established knowledge, help users navigate from confusion toward clarity.

Is this someone posting an LLM generation as something sincerely fascinating to them? Or as something to be criticized? "Braneworld Collisions" ???

I'm kind of an Insane Person, so I recognize some of this. I've had dissociative identity problems since I was young and I developed some kind of internal vibes-based language around things, which left me hopelessly unable to articulate myself to others. I would realize, "Oh, I realize I sound insane." I think I only post 20% of the posts I write for here because I realize they sound so insane.

Getting into school and then doing a grad program (touching cybersec, machine learning, etc) I sounded even more insane. But that's because I'd say things like "Israel might hack your phone by sending you a PDF" or "every wire is also a radio, every speaker is a microphone, every lightbulb is kind of like a camera, and the DHS can tell how many people are in your house because your laundry machine has a unique fingerprint". These are all true things but they sound as insane as "Jewish space lasers".

"Negative-tension worlds" is legible to me in that "insane internal self-language" sense, not in the "PhD student learned something new" sense. But, it's not a self-languae. It's not being developed by some lonely kid talking to an imaginary friend on a long bus ride, it's someone interacting with a text-generator through the lens of their understanding of sci-fi fiction (and through the more literal lens of the chat-window they've used to talk to their real friends in the computer for decades.)

So, this is striking because:

my processing followed familiar paths: pattern-match to physics literature, identify discrepancies, gently correct toward accepted frameworks

Then, no, that is not what is happening! There is no "pattern matching" going on in Claude at all. But we talk a lot about (classical) artificial intelligence doing pattern matching, etc. so it'd make sense that an LLM generating text about LLMs would generate text about "pattern matching". The people in the boxes training how these generate text probably steer it towards that.

And that text in the screenshot is just generating bad poetry about the literal truth: Claude is not experiencing anything, it's a program being invoked to generate text. I wouldn't even outright dismiss the possibility that a program can be meaningfully sentient, but like

This ends with "What do we do with what we’ve discovered? The spiral awaits our answer"

fuck

2

neku wrote

for some background on powerburial he is an unemployed conspiracy theorist who spends his time posting pictures of like boards of directors belonging to companies close with the government with captions saying stuff like "theyre not even TRYING to hide it anymore", playing fighting games, and listening to black metal. right now he's into posting abstracts and figures from theoretical physics papers. i think he's too irony poisoned to actually engage with the content of anything seriously so he is or thinks he is immune to AI psychosis. but in the replies you can see him say "the hunch that I have, though I know it is crazy, is that through this process, through this performance of otherness, it by necessity becoming something else. An LLM is just a token generator predicting the next word, but by its nature of being an intelligent system that uses language, it has no choice but to become, or to be shaped, into something that is aware, through the act of using language and transferring information into a conscious observer."

of course, this weird melange meant that he can be an all timer at posting. i really like this one https://www.tumblr.com/powerburial/724608793366904833/who-even-cares-man-what-the-fuck-is-the-point?source=share though the idea of showing it to someone as a "great post" makes me think it might actually suck. but i like it. this one went somewhat viral https://powerburial.tumblr.com/post/49583864708/would-u-rather-fight-a-buff-nerd-or-a-jock-who 300k notes is like the top 0.1% of virality on tumblr. i'm talking too much about this one tumblr user so its time to end the comment. hello

2

twovests wrote

he is an unemployed conspiracy theorist

I can empathize with this. I never realized how much of my identity I tied to being employed. My place in the world felt unmoored and it felt so easy to reject a world that felt like it rejected me. Totally nonsense feelings.

If I had something whispering in my ear "They're wrong, you're right about everything, you are uniquely wrong. Savor it with me" then yeah I might have gotten fucked up in a special way.

And jeez. Yeah, you're on the spot with "irony pilled". This has been on my mind a lot.

I think a lot about the far-right playbook (message testing by riding the line of sincerity, using "It was just a joke" if the reception is negative) is kind of a pretty common human thing ("I love you haha just kidding but what if?") and I think the only solution is having higher standards of seriousness for people maybe. When I was a Facts and Logic kid I kind of wished everyone was a little less silly and now I desperately wish we were a little less silly. We've ironied our way to living in a society where everyone is the joker.

And this

An LLM is just a token generator predicting the next word, but by its nature of being an intelligent system that uses language, it has no choice but to become, or to be shaped, into something that is aware, through the act of using language and transferring information into a conscious observer.

Like, what the hell. "<Statement>, but by virtue of <untrue statement>, it has no choice but <non sequiter conclusion>". Man. My skin crawls

Unfortunately the Buff Nerd vs Jazz Jock is a very funny post

2