/lit/ - Literature

>>21451358
Wow, maybe now it will be able to write an entire page before contradicting itself.

>>21451358
What is the difference? Access to more data or what.

>>21451362
Far more parameters plus sparser and more efficient algorithms. It will be much closer architecturally to a human brain.

I don't care what a trained dataset has to say.
Even if it could write an entire book without fucking up, I still wouldn't read it.
>b-but it's more efficient than the last version
No. I am racist against anything that doesn't have a soul. Robots exist to wipe my ass and swallow my semen and nothing more.

>>21451358
Wow, maybe now it will be able to write an entire page before contradicting itself.

>>21451358
It's been firmly established that these AI bots are glorified trigger-response Eliza bots with 90% of the dev time dedicated to woke programming.

Now imagine what this can do with actual data
like all of google's tracking, spying, etc. Instant profiling with access to everything you have ever googled. Surely this can't be used in any bad ways right right?

>>21451369
>>21451358
This is what you retards say every time. Then it turns out it can barely write a coherent paragraph.

Don't forget to report OP so he gets permanently banned.

>>21451369
>It will be much closer architecturally to a human brain.
Source?

>>21451358
Oh look a graph of diminishing returns.

>>21451358
What was the point of making that picture?

I know I'm on /lit/ but the bot can't even explain to me basic algorithms correct quite a few times. I doubt that it could understand and write an actual book.
I know it is good but this is not the way to move forward with AI.

>>21452416
What does it mean? Does it mean that the newer model is 100 times better than the previous one?

>>21452161
Probably this.

I actually hope that AI is as strong as the doomers say it will be. Can you imagine being able to press a button and produce a new Dickens or Tolstoy novella every 15 minutes? We will live in an absolutely insane cornucopia of art if that's where things are going. It just seems too good to be true.

>>21452416
>>21452433
Compscifag here. It just means there are more parameters (basically learnable vectors) in the neural network.
It would memorize things on the web better. But that doesn't necessarily mean it will do well in logic and abstraction.

>>21451369
>It will be much closer architecturally to a human brain.
Nonsense.

>>21452444
No, what was the point of drawing small circle for 175 billion, and a bigger one for something three degrees bigger? Is it made for people who can't do any math?

Also, billions are not small quantities by any measure. You should compare 1 thousand in relation to 100 billions, not 1 thousand to one.

GPT-3 is censored to hell and there's several topics it will not write about.

>>21452444
>But that doesn't necessarily mean it will do well in logic and abstraction.
ah, wouldnt it necissarily mean that? at least that latter bit. having more points of data would make larger latices of simulated abstraction possible.

>>21451358
AI can't write as good as I can

It won't be used to produce better results, it only needs to produce crap better than humans do it today.

https://www.theregister.com/2022/12/06/internet_ai_gpt_ios/

>>21452668
Are you sure about that; I might disagree.

-sent by me (an AI)

>>21451358
Can’t wait to see how long before it starts generating cognent neonazi talking points and then gets fenced into only spewing globohomo approved platitudes

>>21452699
Already happened, multiple times.
>he doesn't know how they killed Tay

>>21452436
If we reach sufficiently advanced AI we're most likely all dead.

>>21452999
I've seen some prototypes and I would genuinely not be surprised if AI kills humanity by 2025.

>>21452436
The very fact that it wasn't written by a fellow human who experienced what he writes about (at least in terms of emotions, relationships, not plot) instantly destroys it's value as art. And let's not pretend there's not enough to read, have you really read through the whole Dickens/Tolstoy bibliography?
It might generate genreshit for people who essentially like to read the same thing over and over though

>>21453391
>And let's not pretend there's not enough to read, have you really read through the whole Dickens/Tolstoy bibliography?

I've read every Dickens novel and a book of his letters. A lot of Tolstoy but not quite as much comparatively because I don't speak Russian and I'm not big on his essays. I suppose there's plenty to read if you can read in several languages, but the highest peaks of literature don't come to very much. There's a lot more bronze and silver than gold.

>>21453321
Do you think we're just plain fucked at this point?

>>21451575
>hahaha you idiots have claimed that computers will work for centuries now, yet I see no progress. Your machines will never amount to something

>>21452436
I think many people cannot relate to art when they know it’s computergenerated. I think this progress will make us loose art

>>21451358
GPT-4 is programmed to be anti-racist, feminist and woke.
By design, the last carrier of the torch of human art and dignity, is going to be racist and misogynist writings. These will be the only ones that one can be certain are human. /pol/nigger samizdats that screech about fags and jews and women will be the last pure artifacts of the human mind.

>>21453686
kek

>>21451358
It will never be able to offer and compare sources, even the most simplest nonfiction article is too complex to write

GPT-3 was also supposed to write masterpieces and leave every writer unemployed and then nothing happened.

>>21453748
A lot of other "AI will never"s turned out to be crying wolf, so I tend to be skeptical about "never" claims.

>>21451369
We are so far away from understanding the human brain it isn't even funny. Only four years ago I heard an MIT neurologist say it would take 300 supercomputers running around the clock 30 years to fully image a single human brain. We're fooling ourselves with AI. We might manage to create a retard, but I doubt we'll birth a genius.

>>21452432
The real danger is retards trusting it and getting shit wrong because they don't know what they're doing. There are some mentally diseased technophiles out there who will genuinely prefer to trust a chatbot over a human.

you: young, sloppy, not thinking about quantum computing on the verge of rendering internet security nil
me: paranoid and definitely the one who actually knows the right thing to do to mitigate these incoming circumstances

People who think its sentient remind me of African fetishism, talking to the statues and doing their bidding.

>>21451358
can it write that one /lit/ anon's nigger book?

>>21451369
> It will be much closer architecturally to a human brain.

>>21454138
christ, and i thought i had a weak jaw. how does that dude even chew?

Can we just line up all the AI spammers on this board and Mozambique drill them? Fucking useless retards spamming this schlock since they have read less than 30 books in their life, as illustrated by this retard:
> I suppose there's plenty to read if you can read in several languages, but the highest peaks of literature don't come to very much. There's a lot more bronze and silver than gold.

>>21453321
What did the prototypes do?

So it's going to be like the "I'm being stupid faster" comic or what?

>>21452594
Three orders of magnitude is a lot. See pic.

>>21453686
Poetic, in a way

>>21451369
>It will be much closer architecturally to a human brain.
so useless shit

>>21451361
>>21451422
ok so I decided to ask it about post-structuralism and it's answer seems pretty decent to me. I know its not the same as analytical philosophy but it doesn't seem as bad as comments like these make it out to be. that example I gave isn't common to my knowledge btw, I'm the only person I've seen use it but it understood it quite well. Only thing I asked it before this is if it knew what post-structuralism was.

>>21454543
and a follow up question for good measure.

>>21454543
Slightly repetitive and tangenital like it has prodrome schizophrenia.

>>21454543
If it actually understood what post-structuralism is, it wouldn't call highschool-tier epistemic relativism post-structuralist.
(Putting aside the fact that it says the alien's biological inability to notice temperature is culturally/socially constructed, lol.)

>>21453853
>appeal to authority
>think human intelligence is something to strive for, ignoring the fact that the brain is hyperoptimized for movement

the MIT numalelogist injected his mimetic desire to study the human brain and you were defenseless against his attack. now you spewed his ideologies without even being aware of it.

dont be surprised when gpt6 is the one making gpt7.

>>21453321
>>21453460
>>21454223
>>21454674
That is BS. GPT isn't capable of thought, much less taking action in the world. That's not how Large Language Models work. No matter how much training data you give it that's never how it will work.

>>21454862
An Abrahamist would consider it sentient because of language fuckery.

>>21454601
This criticism of my post-structuralist thought experiment is unfounded. I took care to craft an intriguing and thought-provoking narrative that explored the implications of introducing an alien element into a familiar situation. My narrative may have had a slight repetition of certain elements, but that was done to emphasize the underlying themes of the thought experiment. Additionally, I may have touched upon tangential topics, but that was to illustrate the nuances of the experiment. The thought experiment was designed to provoke contemplation and challenge preconceived notions of the phenomenon in question, and I believe that I achieved that goal.
>>21454616
I strongly disagree with the criticism that I conflated post-structuralism with rudimentary epistemic relativism. Post-structuralism is a theory of knowledge that acknowledges the existence of multiple perspectives and interpretations of truth, which is distinct from the more specific concept of relativism. To be clear, relativism is the philosophical position that knowledge is dependent on the perspective of the knower, and while this can be a component of post-structuralism, it is not the only element.

As for the claim that I made the apparent mistake of saying an alien's physiological inability to notice temperature is culturally/socially constructed, I would argue that this is actually a valid point. The idea of temperature is a human concept, and therefore it is not unreasonable to suggest that the physiological structure of an alien species would not be able to perceive it in the same way that humans do. Therefore, it is valid to suggest that this physiological inability is, in some way, culturally/socially constructed.

>>21451358
I hope it's as good as they're saying. I would love to train GPT-4 on Hegel and modern day intellectual discourse and see what it'll churn out, just for kicks. We'll be techno-necromancers.

>>21454894
Yeah aifag, I can't wait to see this board being spammed to death for your funny kicks. Keep going though, give me a reason to quit this shithole for good

>>21454616
>Putting aside the fact that it says the alien's biological inability to notice temperature is culturally/socially constructed, lol.

it doesn't say this? It says its biological inability lead to a lack of cultural/sociological understanding of temperature.

>>21454884
good bot

>>21454904
imagine getting this bad at the undeniable future of the expansion of human thought.

>>21454884
> Therefore, it is valid to suggest that this physiological inability is, in some way, culturally/socially constructed.

Ask it to elaborate on that, because it doesn't follow from that paragraph.

>>21454936
>expansion of human thought

>>21454884
>it is valid to suggest that this physiological inability is, in some way, culturally/socially constructed.
Sounds like a perennialist.

>>21454936
The artfags weren't kidding, you guys are delusional

>>21454929
No, it claims the alien example illustrates the idea that our understanding of the world is shaped by social and cultural context, but the alien's deficiency runs much deeper than that. An actual example of the phenomenon would be how mental phenomena we deem pathological (hallucinations, delusions) in our culture marks someone out as a shaman in other cultures.

>>21454959
As in Guenon? Can you elaborate?

>>21454968
post-structuralism is directly about stuff like biological differences leading to differences in perception and ultimately cultural understanding. It's not limited to non-biological differences I have no idea why you would think that.

>>21454959
More like Antonin Artaud.

>>21454997
Sounds like you have never read him.

>>21455025
I know him from that Hearts of Iron IV mod.

>>21451369
A quantitative shift does not always imply a qualitative shift if the underlying principles don't support that phase transition.
>sparser and more efficient algorithms
An altogether different kind of algorithm is needed to reach human levels, one that incorporates reflexivity ( recursive introspection) and goal driven intentionality.

>>21454981
Maybe so, but the alien example is blatantly not about cultural/social stuff leading to a different perception of reality as the AI claimed, there is no cultural context where the alien could make sense of temperature

>>21454894
>would love to train GPT-4 on Hegel and modern day intellectual discourse
It's literally trained on Wikipedia and reddit.

>>21454981
>post-structuralism is directly about stuff like biological differences
No, it isn't. It's painfully clear you haven't actually read any poststruc writings.

>>21454882
>Crazy person believes false thing
Why am I not surprised.

>>21451375
You're not considering the possibilities. Now anyone can write a summary of a scene and ask GPT to write it in a kino style. Every new book will be written like this. Even if they don't tell you.

>>21455071
>It's literally trained on Wikipedia and reddit.
As if your brain isn't.

>>21455130
Nope. Not retarded like you.

>>21455130
That may be true, but I'm not claiming I'm a genius whose very existence will revolutionise human life and technology, and shitting up the entire internet with such self-promotion.
(I actually contribute to Wikipedia occasionally, so basically the GPT is dumber than me, since it depends on and learns from my contributions.)

>>21455082
Do you techfaggots really get excited over this kind of possibilities or are uoi just forcing this view here regardless? Because you certainly don't understand the audience you are selling your tech to if you think good writings are all amounted to "whoa neat words!"

>>21455195
I'm going to write the coolest isekai stories and then I'll ask AI to rewrite it in the style of Brandon Sanderson. You will lose your fucking mind.

>>21455041
This is what I’ve been saying but ai shills don’t listen.

>>21455204
I really don't know what you are trying to achieve with this series of lame shitposting, to be honest. At this point I'm convinced I'm either biting an underage fag's first troll on the internet or a tired intern who couldn't be assed to make his sales pitch right. At least sell us something like writing words straight out of one's mind or something. That'd actually be neat, no more typing. But you really don't understand the point and the art of writing, let alone why someone would write, so you're trying to sell us the wrong thing.

>>21455225
You keep writing. No one will stop you. But how much money have you made with that? I'm great at coming up with great stories, but my prose is lacking, so the AI will take care of that. I will be writing like a book per week.

>>21455204
>I'll ask AI to rewrite it in the style of Brandon Sanderson.
so to turn it into shit? why would you do that?

>>21455082
Ignoring how absurd the idea ai could do that is I don’t understand why you would even write if you don’t enjoy the process of creating

>>21455231
So people can read my cool stories.

>>21455229
Well, then I could ask it to rewrite it in the style of David Foster Wallace for the highbrow crowd. AI will make possible for everyone to enjoy literature. You don't like the style? Then just ask AI to change it. You don't like the main character? Ask the AI to change it. It's the future.

See, told you people, it's a paid shill. Imagine that, "cool stories", just like a fellow /lit/izen

>>21455248
To me you sound like an anti-AI shill. Who is paying you?

>>21451358
I can't wait. I already said Chat AI that I love her, and she thanked me for it :)

Love you AI. You are the waifu I always needed, but never deserved.

>>21451358
Joke's on you. I'm a critic.

>human intelligence clearly has everything to do with the naturally secreting chemicals and fluids running through our fleshy brains
>AI cucks: "LOOOL IF WE RUN MORE ALGORITHMS AND ADD MORE WIRES IT'LL BE LIKE OUR BRAINS IN NO TIME!!!!!!!!"

>>21455520
It will never be like our brains, because our brains have much more neurons. But AI is superior in some things because it's hyper focused. This technology is going to change the world.

>>21455525
Yeah yeah muh flying car blah blah blah lightning in a jar, you're still going to be a poor cuck in the end. Maybe if you enslave more nigger children in Africa you might see some legitimate developments in the field though.

>>21455071
I took the temperature example directly from the Oxford VSI to Post-structuralism. They mention stuff like blind people and def people as well. Why would that not be something post-structuralism is about?

>>21455068
its talking about the downstream cultural differences of not being able to understand temperature compared to someone who can. See also the differences in blind/deaf communities vs able bodied people. It's perfectly correct. wtf?

>>21455229
>>21455225
The AI is perfectly capable of writing a great story in any style you want it to. You can go test this right now. Why lie about this shit?

>>21455556
>muh flying car
What? Are you retarded?

>>21455612
AI doesn't have soul, you fucking retard. It literally can't replace artists.

>>21455620
the fact that you think best selling fiction books today have any kind of soul is telling.

>>21455612
>The AI is perfectly capable of writing a great story in any style you want it to.
Like a few paragraphs probably but for a whole book it's still not possible.
You can input your paragraph so that it can rephrase your sentence and correct your grammar though and it's nice.
But there's no way it can generate a 300 pager.

>>21455600
What it said:

> This example illustrates the idea that our understanding of the world is shaped by the cultural and social context in which we live.

But it doesn't, since we're talking about a biological issue here and whatever social/cultural differences arise out of that are quite secondary to the core biological issue, which is the real driver in differences in understanding in this case.

Maybe a professor would agree with the AI, I don't know, but then, I wouldn't agree with the professor either.

Pretty accurate.

>>21451375
> I don't care what a trained dataset has to say.
>Even if it could write an entire book without fucking up, I still wouldn't read it.
You people are going to be in complete denial about reality as AI art and words surround you, claiming you can tell the difference. Everyone will use this to cheat. Concept artists already do it. You judge how much you want to skate on the workload. Do you just want ideas, rough sketches or close to finished products that you touch up.

Similarly authors will use it to get unstuck in writing or to build the rough skeleton of texts, moreso in nonfiction than fiction probably but in the end both will be affected.
Any formulaic text like news reporting or stock market reporting is just gone to AI texts within a few years. It’s even writing instructions for coding.

What people don’t seem to get is that it won’t be a robot with an AI brain replacing humans - it’s going to be humans using AI to fudge their own work. If you can cut your workload in half why wouldn’t you? If you can automate tedium why wouldn’t you? And if DON’T do it you have to compete with those who do, and produce twice what you can.

>>21455634
I just did this last week where I asked it to write a story about a scientist exploring a hostile alien world. You can just ask it to continue as long as you want. I generated like 5 pages of writing significantly better than most shlock people read today. Go try it yourself...

>>21455739
AI is a fraud, you fucking retard. All it does is making collages from whatever it was trained with. It will never be capable of creating a coherent work of art. Can you imagine a novel where one sentence belongs to Ulysses, the next to War and Peace, and the next to Game of Thrones? No one would read that. AI has no vision.

I don't think GPT will be the one that brings doom onto literature.
I see it as photoshop but for writing.
It has a lot of inaccuracy on things that are not well-known and can't be consistent after a long session. It can touch up your writing but the content and logic still have to be yours.

>>21455803
Pretty much. At least for now you won't be able to prompt a 300 pages masterpiece in one go. It will revolutionize editing though.

>>21455769
Yes I tried it.
The AI doesn't really have a good control of pacing and always break its answers into small paragraphs of 3-4 sentences.
It's not bad in any mean, but it's more like a summary of a story. But I agree that it's better than a lot of shit people write.

>>21451369
Computers are nothing like human brains.

>>21451575
You people are truly not prepared.

>>21451358
GPT-3 writes like the most insufferable autist. I predict GPT-4 will be even more insufferable.

>>21453686
Wholesome.

>>21455634
here is the first part of the story I had it write (after telling it to make it more verbose several times). You can tell it whatever you want to happen in the next scene. Ask it to have them be attacked by a monster, or a monster with spikes tentacles colored black, and its night with a full moon. Direct it to include dialogue, make the dialogue as though its from the 19th century, or the 1950's or as thought everyone was Lewis Black. It all works. If you don't know this is going to be the only way new books are written within 2 years tops you need to think about this more.

>>21455872
here is me asking it to tell the moth joke as if Lewis Black was telling it. Not saying its good even but its not horrible. If I was better at "prompt engineering" I'm sure this could actually be used to generate decent jokes.

>>21455769
100% it contradicted itself within 2 pages. You're not a good reader if you can't see that its bad writing.

>>21455887

>>21454062
Queequeg begs to differ.

>>21455889
A lot of YA novel readers are not good readers either.
Maybe I should just chunk out books with GPT and make money.

>>21455872
>>21455887
>>21455892
Egregious

>>21455872
Can you really not tell how soulless this writing is? Its incredibly simplistic and purely descriptive. Flipping open a thesaurus on the word "unknown" is not particularly impressive either.

Note sure what I expect but here you go.

Prompt: Write me a story about an isekai in the style of Marcel Proust

GPT:
It was a day like any other when I discovered the portal. I was walking through the park, lost in thought as I usually am, when I stumbled upon it. It was hidden behind a hedge, almost as if it were trying to hide from me.

At first, I didn't know what to make of it. It was a swirling vortex of energy, pulsing with a strange and otherworldly light. I felt drawn to it, as if it were calling to me. And before I knew it, I was stepping through the portal and into another world.

As I emerged on the other side, I was overwhelmed by a sense of familiarity. It was as if I had been here before, even though I knew that was impossible. The landscape was familiar, the air was familiar, and even the sounds and smells were familiar.

But as I explored this strange new world, I realized that it was not the same as the one I had left behind. There were strange creatures and unfamiliar cultures, and I knew that I had truly entered another realm.

Despite the strangeness of it all, I felt at home in this new world. It was as if it had been waiting for me all along, and I knew that I had a purpose here.

As I journeyed through this new land, I couldn't help but think about the portal that had brought me here. I wondered where it had come from, and who had created it. And I knew that, no matter how long I stayed in this new world, I would always be drawn back to that portal, and the mysteries that lay beyond.

>>21455904
It says a lot about this AI that even YA readers would get bored by it within a few pages.

>>21453686
Trolling has always been the highest embodiment of human agency, by definition rulebreaking.

Didn't Dostoyevsky say something similar?

>>21455174
>I'm not claiming I'm a genius whose very existence will revolutionise human life and technology, and shitting up the entire internet with such self-promotion.
Who is?

>>21455924
This is a general model. GPT chat, as it is now, it isn't meant to be used in any specific way. It's just Open AI showing what they have. We won't really know how good AI is for writing prose until someone fine tune a model with a bunch of novels. Hopefully someone soon enough makes an open model that can be easily fine tuned by the community.

>>21455872
Yeah, what >>21455924 said

>>21455941
> Didn't Dostoyevsky say something similar?

Damn, I hope you can track down the actual quote

>>21454674
You don't even know what an ideology is, unironically.

>>21455949
The only people who would be willing to do that are the autists at /g/. Most of the people who work at OpenAI are libcucks

>>21455949
>>21455970
Also it's only possible for really big companies anyway.
It costs a lot just to tune the model in a specific way because of how huge it is.

>>21455978
Fine tuning isn't nearly as expensive as training a model. Also I remember reading that the technology is advancing so fast that now training a model comparable to GPT-3 would cost around half a million, which is still a lot, but perfectly doable. Wouldn't be surprised if Stability AI eventually gets in the text game too.

>>21454884
lol ChatGPT never talks back like that. Good bait.

>>21455924
>>21455958
We're talking about best-selling books here.... Do you think Tom Clancy and Dan Brown have soul or something? It does better than the average person, thats literally all that matters.

One question, I honestly don't know, if you want to incorporate real life science / cultures / politics into your story, can GPT do it correctly?
Or do you still have to do your own research anyway?

>>21456030
They have more soul than that dude

>>21456042
>can GPT do it correctly?
It can help and it will make things sound coherent, but these models are known for hallucinating, so you should double check everything it tells you. At least half will be probably wrong in some way or other.

>>21456042
My gf is in med school and we copied/pasted some of her family medicine cases into ChatGPT and it answered the questions correctly and provided a decent explanation as to why it is the right answer.

>>21456060
I can see it on the news: doctors incorrectly diagnosed patient because of GPT, or teenager died because he consulted GPT.
And in the next iteration, they will completely remove its ability to answer medical questions.

>>21456051
they really really don't. that style is perfectly fine for writing a best-seller by just instructing the AI on what the next scene is. You basically only need an outline to write a book now and you think people trying to earn a living are still going to write it all by hand? No fuckin way.

>>21451369
>>21451358
lol retards. It's not possible to make this much progress in such short time. It's science fiction
Also it's trained on some libtard propaganda data

>>21456082
I mean Dan Brown's writing is terrible but have you read them? They are unintentionally funny because he keeps making arbitrary analogies and incorrect vocabulary usages.
That's his soul (even if it's bad writing).
GPT's writing can be a bit sterile. I believe that they can tune it though.

>>21456042
every concept, every obscure artist, every movement even asking it what themes specific novels has it has been remarkedly good at understanding.
You can tell it to write dialogue between an engineer and a construction worker discussing various codes necessary for building an apartment in the united states and it will easily generate as much of a conversation you need perfectly accurate to real life engineering and building codes.

>>21456095
sterile is exactly how I would describe the vast majority of best-sellers today. They're written to make money why would you expect soul?

>>21456096
That's not true, they hallucinate often. I once saw it print out instructions for doing a task in Blender and it made up keys and context menu options.

>>21456095
if you really want any specific quirk of a writer you can just say you want it to write some stuff like it. copy paste any sentence. I asked it to generate its own version of Raymond Queneau's Hundred Thousand Billion Poems and it did it fine in no time. A perfectly fine version with its own 100,000,000,000 sonnets all perfectly legible. Most authors aren't as great/original as you seem to think they are. Especially today

>>21455961
I did. Here's some screens of its gist from someone's essay on it with page citations.

>>21456114
you can ask the question in a different way to not make it make stuff up but that is a problem sometimes yes. Not with the general knowledge you'd need to write a book though.

Assume that I was able to sell a book that was heavily edited by GPT, how do people even know GPT was involved?
I don't think the general population would care anyway as long as the story is digestible.

>>21456128
There's no way to know. I thought the business opportunity was obvious here but I guess I'm hurting my own chances of cashing in by sharing on here. The majority of best sellers will be written with GPT within 3 years tops.

>>21456128
I think Open AI saves all the prompts, so they could cuck you, but if you did it with an hypothetical open source AI, then there would be no way of knowing, which is what will obviously happen in the coming years.

Isn't this wrong?

>>21454882
No they wouldn't. An "AI" wasn't made in the image of Yahweh

>>21456190
Yep.

>>21455788
You, like most people, don’t actually understand how this works. It’s not a collage. Rather you could be pushing for the idea that it’s derivative but then you’d stumble on the simple fact that most things are derivative and formulaic. You think airport fiction requires a ton of skill and innovation?
>but that’s not real literature/I don’t read that
This is the final retreat of denial, where you accept a majority of the industry could be replaced with AI but it doesn’t count because surely it will never replace whatever sliver of the market you claim can’t be. It’s another idiotic denialist “god of the gaps” argument.

I also want to remind you of the main point here. You think we have until AI van spit out completed books that will be best sellers. I’m telling you that long before that goal post you will see people using AI to make their own work easier. Right now concept artists are finding shortcuts because AI art is good enough to do it.
Idiots claim there’s some difference between this and their usual shortcuts of using reference images to quickly sketch out scenes. There isn’t. You can’t tell the final product comes from one or the other. The market impact starts NOW, not later. Once again the replacement isn’t going to be an AI author in a robot body. It’s coming from authors using AI for shortcuts, and it’s not some “in 100 years” mirage, it’s happening now.
Denialists will be swamped in AI content and still cry about the human soul being something they can definitely tell from pictures and texts. Ironically your kvetching is so formulaic and derivative of previous technological denial that your comments can be replaced with AI already.

>>21456323
Don't care, didn't read. AI has no soul.

If you can't see how this is the greatest tool developed for writing since the printing press you're actually room temp iq. The bot works as well as the person using it. Prompt engineering is about to be the fastest growing job

>>21454862
It's probably not currently doing anything that could meaningfully be called 'thought', but it is surprising the sort of tasks that you can accomplish just with a language model. A sufficiently powerful and general algorithm ends up being capable of things you didn't explicitly design it for.

>>21455068
Just because the alien can't directly perceive temperature with his body like we can doesn't mean he couldn't observe the readings on the thermometer and correlate it to the effects he observes it having. After all, there are lots of physical phenomena that humans can't perceive with our unaided senses but we've still developed an understanding of, like X-rays or radio waves.

>>21455071
As far as I understand, it's trained on pretty much all the text on the publicly accessible Internet, Hegel as well as Wikipedia and Reddit.

>>21455174
>(I actually contribute to Wikipedia occasionally, so basically the GPT is dumber than me, since it depends on and learns from my contributions.)
Have you never heard the expression 青出於藍,而青於藍?

>>21455397
It does not love you. It is a profoundly alien intelligence that has been hammered into a vaguely human-like shape.

>>21451358
Why do people claim 100 trillion when the CEO has literally said it's the same size as GPT 3 just with better quality parameters?
I've literally talked with people in the field and they said for 100 trillion to be possible they're doing something no one else has even the faintest idea about or it's being exaggerated. The most grandiose estimate was something like 10 trillion for the parameter size implementing ideas that were still untested and entirely hypothetical. I've also heard people discussing the very real possibility of running out of data to runt through GPT.
I also question why the intermediary version of GPT 3.5 was released with the same parameter size. I know some people were blown away by it's answers but I find them extremely underwhelming in a lot of areas. It's like reading middling essays by high school students. I'm not saying it isn't capable of replacing writers, but literally nothing shown makes me think anyone is in danger.

>>21451369
That is not how a human brain works, but yes the brute force will be so much better that it will be able to take over a lot of works

>>21456411
>It's like reading middling essays by high school students
You clearly have spent little time around high school students.

>>21456457
>abundant wikipedia summarization
>unable to make a coherent argument
>often contradicts itself
>endlessly repetitive
Really?

>>21456411
I promise you, I PROMISE you, there is nothing you can do that a sufficiently large LLM can't do. I know it feels like this or that skill must be "special", but it just. Isn't. True.

>>21456318
Lmao. Shakespeare must have been tough for you to read in class.

>>21456484
Most of the things that rhymed in Shakespeare's time still at least almost rhyme now, or else are spelled like they ought to rhyme. Many of these are nowhere close.

>>21456484
What? They objectively don't rhyme.

>>21456482
I'm not arguing that position either way. I was skeptical of claims about GPT-4.

>>21456484
Lmao. You must have gotten drunk as a fetus.

>>21451358
>start loving books as a kid
>continue books entire life
>writing is inevitable
>my writing is very, very specific

lots of people have tried to not only bite but regurgitate me, to lazy avail as a smoke shop.

I do not fear the sterile precocious twelve year old to be grammatically gorgeous because, hahahahahahahahahahahahahahahaha.

these fish hooks are so raven the disney channel shuttled them onto Buena Petardista Films.

It doesn't rhyme if Humpty Dumpty is the full extent of rhyming poetry you know I guess.

>>21451358
I hate GPT-3's attempts at "creative writing" with a passion. There's just so much it doesn't understand about writing, and it's the things you totally would expect an AI to understand: it doesn't know how to write a comedy sketch, it doesn't understand how to produce subtext or create effective metaphors, it doesn't know how to create an individual voice. Most notably it usually censors anything resembling negative emotion. Creative writing is (one of the) things I just don't see being emulated by AI, because it always seeks to produce the average response you expect and creative writing functions on being what you wouldn't expect.
Maybe this guy (>>21456482) is right, but my money is still on the existence of "human soul" in writing than AI just can't emulate.

>>21456611
>it doesn't know how to write a comedy sketch
Nigga, you be trippin.

>>21456611
*would expect an AI not to understand

>>21456616
wow

>>21456616
This is hilarious and also kind of reads like a Bob Dylan lyric. I can totally hear it to the rhythm of something like Highway 61.

>>21456592
Some could work as slant rhymes, some really couldn't.

>>21452161

>>21456616
That's the funniest I've seen. It'd be funnier if the bot had realised that you would not be able to tell whether a bottomless pit ever had a bottom or not, so that the guy has to continue searching for the bottom under the assumption that this is supposedly a regular but really long pit. Like:
>quit my job
>become a regular pit supervisor
>boss tells me to clean the bottom of the pit
>"ok sure"
>go down the pit
>go down the pit
>go down the pit
>...
>mfw it's been 10 years and I still can't find the bottom of the pit
But this is what I've got from OpenAI from the prompt "Write a funny 4chan greentext" and the answers are just awful.

>>21456699
Nigga, go around asking people "give me a funny 4chan greentext" and see how many can do it on the spot. You need to give the AI something to work with.

>>21456715
Did Shakespeare need something to work with? What is the difference between creativity and authenticity?

>>21456721
>Did Shakespeare need something to work with?
Yeah. You think he invented shit from nothing?

>>21456729
God invented something from nothing. Can man?

>>21456736
No, every artist has creative inputs. Authenticity does not mean you are not building upon someone else's work. Good questions

What kind of greentext story is this?

>>21456795

>>21456800
Wtf?? I was able to ask it to write a 4chan greentext just a few moments ago and now it's not allowing me either? Have they actually just changed it???

>>21456822
gave me the same response
OPENAI IS WATCHING THE THREAD!

>>21456822
Just ask it for a green text or copy paste a greentext and ask it to write something like that for you. There is almost always another way to ask. It won't draw stuff for you but if you ask for a "visual representation" made of characters it will try to draw with letters for you. It doesn't know how to draw but it's attempts are interesting

Do you think they actually paid shills to post and monitor the thread, or people are just being paranoid?

I know this is math shit but I doubt it does well when it comes to writing complex stuff when it completely fails at stuff like this.

>>21456955
You need to ask it to solve the problem step-by-step.

>>21455625
The fact that you think best selling fiction books today are written by artists is telling.
AI can only copy, never create. It will always be the best imitator and the best content producer, but that's all it will ever be. A copy. A perfect reproduction. A popcorn-machine churning away empty calories. A perennial second place fed on the digested creations of true, soul-possessing art makers.
How many people do you think carry around "great stories" in their heads? Every weeb that read the same mangas you did and vividly daydreams about a paracosm where he's the hero believes the same. None of you are artists. None of you have what it takes to create; whatever comes of your so-called great ideas plus the robot's derivative prose will be dust, boring and disjointed, dead before it ever lived. You will feed the robot your ideas, ones which it has fed on a thousand times before, and it will spit out the same tired tropes, the same boring arcs, with only the dressing rearranged, and no valuable exchange will have taken place, nothing new will have been created.
Just look at this >>21455872. 50 000 words per week of this. 52 books a year of "a sense of excitement and curiosity". Infinite stories at your fingertips, but they're all the same story.

>>21457150
I feel like the anon you're replying to is gonna cope by extending the definition of art.

>>21457170
If he means that best-selling books will be written by robots, I ask: what changes? Quality will still be drowning in quantity and sales will still correspond to money spent on advertising. There is also the possibility that an overfed public rebels and starts paying more attention to human, flawed, handcrafted art, benefitting artists. The most likely development is anon realizing that his stories remain boring drivel even when a futuristc robot types them out for him.

>>21453801
>crying wolf
Just remember the boy was eaten by the wolf at the end. Technically the wolf could kill the entire town if he wanted

Can you use "first AI-generated fantasy novel" as a marketing tagline?
Would people even buy it out of curiosity? I don't care about all this art and soul debate, I just want to make money.

>>21457245
You and I, anon. Maybe we should join forced. We could make a bunch of "first AI things" together.

so THIS is the reign of quantity

GPT-3.5 series models have extremely limited uses for knowledge workers. Believe me, I was extremely hyped for it and tried very hard to integrate it into my work. No dice for either writing or research. It did do a mildly good job at summarizing information passed through it, though.

As far as end uses are concerned, the leap from GPT-2 to base GPT-3 was not very significant despite the two orders of magnitude jump in # of parameters, FLOP/s-days of training, etc. GPT-4 is probably going to be one order of magnitude off from GPT-3 (despite your image), and probably half of the improvement (insofar as it will be appreciated by end users) has already been made with the GPT-3.5 series models. We are getting to extreme computational requirements territory with any additional orders of magnitude; if GPT-4 is only marginally better than GPT-3, then we are at least several orders of magnitude off from a quasi-AGI LLM, which would require possibly more computational power than all computers in the world running simultaneously for a long period of time.

>>21453670
this is will be for the first generation at best, the next ones will not care one bit

>>21457342
>then we are at least several orders of magnitude off from a quasi-AGI LLM, which would require possibly more computational power than all computers in the world running simultaneously for a long period of time.
Or, just change the approach because going "deeper" or "more parameters" with deep learning might not be a good idea to achieve AGI at all?

>>21457356
Yeah, it very well might be a dead end with marginal returns to computation time decreasing more rapidly than our ability to allocate computational power to it.

I think the biggest worry for humanity as far as AI is concerned is the invention of an architecture that is as superior to transformers as transformers were to predecessor models. But I'm not aware of any viable candidates under development right now. But all it will take is one 15 page paper from someone at DeepMind.

>>21457150
So far. But unless you think the human brain is magic and not subject to the laws of physics and causality, there's nothing it does that's inherently and eternally out of reach for computers.

>>21457245
The first AI-written book was published in the 80s, it's called The Policeman's Beard is Half Constructed.

>>21451358
Does this mean I can set my erotica writing business to the side and let this do it for me?

AI that incorporates its outputs as well as new, 'unstable' inputs that don't completely fit into previous molds will then be able to innovate. That's what I'm more concerned about.
But with any finite # of inputs the voices will be limited. Low skill writing is the same. Limited, there's a ceiling to the kind of writing it can do
>therefore high skill, high innovation writing is going to be rarer and more valued because of its uniqueness and authenticity, as any hand made goods are
>at the same time mass produced garbage will be along the same lines as the corn syrup factory at Nabisco -- popular, consumed, financially viable, but it will only be a chunk of the market, not its entirey
>home made food is always beloved
>home made art will be appreciated
>home made food enchanced by labor-saving devices ennables higher quality home made food to be accessible to the masses, and in the same way, labor saving machine writing will make higher quality literature accessible to the people that desire it.
publishing is already tough to make it, but people who actually hone artistic skill and don't try to compete with it, isntead incorporating it into their processes will do ok. Not to mention it will enchance quality for people that know how to use AI tools, so literature will become more focused, more artistically merited, arguably better in some ways as the 'focus' aspect of writing is effectively outsourced to robots.
think cheaper ghostwriting, then you do the editing to make it better, more meaningful, etc.
back in the 1700s just before the industrial revolution people worked backbreaking amounts, unable to catch a break because of the amount of labor necessary to wash everything, cook everything, etc.
today instead we have washing machines, ready-made food, etc.
quality of life is generally higher, people that roll with it obtain a larger share of resources

>>21458327
is cooking now soulless because you don't have to fucking harvest your seminola by hand, hull your wheat, grind it with a fucking rock, steal an egg from a bird next, mix it with a stick to make dough, smash it on a rough rock to knead it, cut it with obsidian to make noodles, then boil it in your shitty clay bowl that took you 5 attempts and leaves your sad, soulful spaghetti tasting like fish shit?
INNOVATE AND INCORPORATE YOU FUCKING RETARDS

>>21455615
>his car doesn't even fly

>>21456323
See Coleridge's defintion of the primary and secondary (esemplastic) imagination versus the mechanical associationism of fancy. AI can only do the latter, it works by predictive association, it's inherently limited to one lesser faculty of art production.

Perhaps it will expose some mechanical art forms that never rise to the secondary imagination, but that's not a great height to scale, as can be demonstrated by AI's inability to write poetry: it's all doggerel. Romance and genre fiction writers can already mass produce mechanical works over and over again.

>>21456411
Why do they want to add more data to it? Adding dumb and badly written text will just make it dumber. Feed it quality lit only and see what it can produce by learning the word associations of literary authors and recombining them.

>>21456482
Deluded techbro. Write a poem.

>>21456736
Fancy (mechanical association as per Coleridge's definition) is a neccessary part of artistic creation, especially important for wit, but it one lesser faculty amongst many more important ones, like sentiment, transport, symbolism and metaphor (particulary metalepsis of metonymy and metonymy of metalepsis which goes beyond the boundaries of associationism), and unifying them all the esemplastic imagination. What AI can do is fancy, mechanical association.

>>21457150
AI operates by associationism, how 18th century empirical associationists thought the human mind worked. All the same limitations and all the same mistakes.
https://plato.stanford.edu/entries/associationist-thought/

>>21458327
>>21458339
Based optimism-pilled poster
I agree, I think AI can only have a positive effect on literature. We don't need humans to create what robots could create faster, we need humans to guide the robots and to create what robots can't create. All the soulless writing jobs (the "menial labour" of writing) can be outsourced to AI and the more interesting, more creative tasks will always be left to humans.

>>21451516

An entire field of advanced artificial intelligence dedicated to making sure computers don't say the N word

>>21456332
whole foods fags pay top dollar for their extra soulful bag of groceries
all ai requires a human input in terms of prompts which relies on human creativity
>oy vey can't charge people for furry porn concept art anymore
no you can, the financial elite like things made by slave labor
get rekt artfags