Slamming AI with Kinkery

[Husky_Embrace]

Use of AI is not accepted on this site. The moral and ethical implications aren't fully understood and is a hot topic. This site believes it is best to not include anything with AI to prevent any current or future problems.

AI has the potential to be a wonderful tool. In some applications it has already done wonderful things. You can spot Tuberculosis faster and more accurate than a human. If you want something bigger, the folding of enzymes that took years to crack has now been solved by AI.

Personally I have also been at the forefront, being an Alpha Tester for DALL•E. For someone with very limited art talent it is definitely wonderful to create great things, especially with those unrestricted Alpha and Beta versions. This is the same for writing. It is a great tool as an assistance.

There is a difference between the two categories however. The language and painting is made from data that is not their own. They didn't pay for it, nor created the data.

Imagine going to work, where someone tells you that you're being replaced by AI. They trained the AI without your consent on your work and can now churn out stuff at will. This is what is happening to writers and picture creators in news, education, photography, entertainment. Why take a course in Python if you have AI that trained on their efforts without offering a dime in return? I asked the google search AI if people get any money from what it actively scraped from a news site. The answer is nothing, "but it could prompt more people to click to the source, getting revenue from ad sense." Which is not a good answer, as most people will not click further. Money for google, none for the one who put effort in gathering and writing down the information.


It is no different with writing stories. The question becomes what was made by you, and what was made over the backs of a thousand writers. AI monetising the work of others is simply repugnant.

Again, I see the advantages. Recently I didn't ask my partner, an Excell guru, how to write a complex formula. AI told me and after a few adaptations of the suggestions I got it to work. That being said, morally and ethically it is horrible. Someone making Excell video's doesn't get my click/money, nor millions along with me. Even if it was their work the AI learnt from.

So explaining all that I ask you friendly, do not use AI to make stories for the public.

 

[AwkwardMD]

What are we drinking this time, @AwkwardMD ? Tonight was my Friday, so I made myself a Smoky Mezcal Old Fashioned with Illegal Tequila instead of bourbon

I've been enjoying absinthe straight (ie, not prepared with burned sugar cubes). In my mid 40s, suddenly, I don't hate the taste of black licorice as much anymore. The absinthe we can get in the states is strong, but not *melt the walls* strong.

 

[Rob_Royale]

I've been enjoying absinthe straight (ie, not prepared with burned sugar cubes). In my mid 40s, suddenly, I don't hate the taste of black licorice as much anymore. The absinthe we can get in the states is strong, but not *melt the walls* strong.

I've been wanting to try that.

 

[Rob_Royale]

I'm not looking for acceptance. I'm expressing my point of view on the topic. Oh, and where were you when I needed support and encouragement to continue writing an epic novel of 296K words, the last time I checked?

Well for one, I wan't replying to you. I probably would have advised against writing something like that as a new writer. Your lone submission is 3k words. Learn to walk before you run. The fact that a paid beta-reader advised a rewrite should say something.

I'd also like to add that, as a writer, it wasn't enough for me to post a story on Lit. I wanted to be a part of a community of writers who understood that writing isn't easy, but there are people willing to help. Apparently, I was incredibly naive.

You're there. Writing isn't easy. By all means find a like minded human and write it as a team. There are plenty of those in the literary world.

 

[AwkwardMD]

I've been wanting to try that.

I'm a fan.

 

[yano2mch]

There is a difference between the two categories however. The language and painting is made from data that is not their own. They didn't pay for it, nor created the data.

Imagine going to work, where someone tells you that you're being replaced by AI. They trained the AI without your consent on your work and can now churn out stuff at will. <snip>

It is no different with writing stories. The question becomes what was made by you, and what was made over the backs of a thousand writers. AI monetising the work of others is simply repugnant.

Maybe. It would probably be more ethical if we knew all the training was done on stuff that was public domain, but with current copyright law pushing things back basically 100-140 years that means a lot of contextual and useful information would be inaccessible. I can only imagine works like Mark Twain and Moby Dick being used in it's limited language and understanding to explain a modern topic like CPU chip wafer manufacturing and being totally lost because it's before world war 2 still according to the AI.


Hmmmm.... I have to wonder if you would think data compression should be outlawed and we should only send uncompressed images and text as well. I can already tell you'd wonder why i'd change the topic to data compression.

If you've ever tinkered with weights and huffman or the like, you come down that compression scans a piece of data and then maps which is the most likely to be next, either in raw weight or in what comes after if there's multiple pieces that come before it. The most likely word to follow 'Hello' is probably 'How', as in 'Hello, how are you?'. As it takes the most likely elements to use the fewest bits possible it wouldn't be that far from what we're talking about, just lacking the multiple transformers that makes LLM and other processing possible.

So if we use precalculated public efficient weights (as far as per-symbol or word) you can get really decent compression without then having to include the weights in every message; Or you can not rely on someone else's work and make your own damn weights and compress your own data by hand. Because using anything from anyone else (public and free or otherwise) would be unethical and repugnant.

 

[Husky_Embrace]

Maybe. It would probably be more ethical if we knew all the training was done on stuff that was public domain, but with current copyright law pushing things back basically 100-140 years that means a lot of contextual and useful information would be inaccessible. I can only imagine works like Mark Twain and Moby Dick being used in it's limited language and understanding to explain a modern topic like CPU chip wafer manufacturing and being totally lost because it's before world war 2 still according to the AI.


Hmmmm.... I have to wonder if you would think data compression should be outlawed and we should only send uncompressed images and text as well. I can already tell you'd wonder why i'd change the topic to data compression.

If you've ever tinkered with weights and huffman or the like, you come down that compression scans a piece of data and then maps which is the most likely to be next, either in raw weight or in what comes after if there's multiple pieces that come before it. The most likely word to follow 'Hello' is probably 'How', as in 'Hello, how are you?'. As it takes the most likely elements to use the fewest bits possible it wouldn't be that far from what we're talking about, just lacking the multiple transformers that makes LLM and other processing possible.

So if we use precalculated public efficient weights (as far as per-symbol or word) you can get really decent compression without then having to include the weights in every message; Or you can not rely on someone else's work and make your own damn weights and compress your own data by hand. Because using anything from anyone else (public and free or otherwise) would be unethical and repugnant.

Unfortunately your arguments aren't in the same ballpark. I hope it was clear from the examples that I do not oppose AI. It is the data they are trained with and their use after that are the problem.

It doesn't matter how you spin it. Most public AI's sell you other people's work without compensation or approval of the people they took the data from. An AI might remix it, but it is still not their work.

If you do want to use it morally and ethically, your suggestion is great. Feed it all your own data.

Saying that AI can do what it wants just because it is useful to you is not acceptable however.

 

[yano2mch]

It doesn't matter how you spin it. Most public AI's sell you other people's work without compensation or approval of the people they took the data from. An AI might remix it, but it is still not their work.

To my understanding the models don't include whole libraries of people's works (or even a plaintext of a single work); It's only based on said work.

I suppose to understand you'd probably need something like Multi-Level Huffman to explain it. So lets take Mississippi, you have weights/instances of letters as following.

M:1
i:4
s:4
p:2

Which then would make (probably) this tree which is what's actually stored

00 - p
01 - i
110 - s
111 - M

Compressed and encoded using Huffman then looks like: 111 01 110 110 01 110 110 01 00 00 01.

But LLM's are not a single letter or word instance, they look back N words or farther, most have 4k to 8k context and even further. So if we look to 2 letter instances (current letter and previous letter you get the follow combinations: Mi, is, ss, si, is, ss, si, ip, pp, pi. Now we add those up.

Mi: 1
is: 1
ss: 2
si: 2
ip: 1
pp: 1
pi: 1

Using this table, the first letter is the 'past' letter and the second is the current letter, so once we give it M, only i, can come up, and with i in the previous letter only p or s is possible, requiring a single bit to determine which it chooses. (This is a very VERY small/simplistic table, if we went larger it would be too hard to track/follow, i've had outputs 3-4 deep and results were.... interesting. Also input extracted this way can span hundreds of megabytes, to which I'd have to quantize to only keep the shape of the huffman tree, not exact weights)

Now question: From the upper table(s) alone (or the tree) are you going to claim it contains a whole word or even contained a copyrighted work? Because that's not what the models do. Using the weights combined with the tensor layers that have it navigate through the weights do you finally get word outputs (or some type of output).

 

[Britva415]

Let's just look at things practically. Separately from the philosophical positions regarding whether the AI rules on this site are "OK" or "indefensible:"

Pragmatically speaking, now that OP has stated publicly that they had the idea to write a story this way (use AI to generate it, then re-write it in their own words), they probably have already made theirself a target of greater editorial scrutiny than other first-time story submitters here.

@sxPlasticity : You already have an outline. Wouldn't it be 100% just as easy to just write around the outline you already have, than to give it to a LLM and then un-do what you got from it? Probably easier.

 

[filthytrancendence]

Now question: From the upper table(s) alone (or the tree) are you going to claim it contains a whole word or even contained a copyrighted work? Because that's not what the models do. Using the weights combined with the tensor layers that have it navigate through the weights do you finally get word outputs (or some type of output).

I'm genuinely curious if you think this kind of pedantry is going to convince anyone who disagrees with you, or if you're just a troll?

 

[yano2mch]

I'm genuinely curious if you think this kind of pedantry is going to convince anyone who disagrees with you, or if you're just a troll?

I'm serious. Most of the weights used in data compression, and likely LLMs don't contain whole works of art and instead only pieces that have to be pieced together based on likelihood of appearance, negative appearance, or some type of numerical value to select the one it wants.

The above example is super simple because to expand to something far more complex, may take up a ton more space to get the general idea across. But with LLM's and other AI models, you're not looking at a couple characters back, that was just to show the complexity differences, when you start going larger actual examples like 'The quick brown fox jumps over the lazy dog' the complexity and having up to 26x26 tables JUST FOR 2 LETTERS (676 tables, though many would be culled since zz never appears for example).... Nevermind scanning something like the bible. Finally you'd probably have a singular table of single letters/symbols and their weights in cases where names new words and combinations weren't scanned in. So my name would be 'Ya' 'no' '2' 'm' 'ch' as 5 symbols rather than say Hercules which may be 1 or 2 symbols if greek mythology was included.

Also the more it's trained on more characters for lookback, even random input based on weighted models from likely possibilities will result in pseudo language and semi-comprehensive content (probably garbage but you'd see words or sets of words that actually appear in literature). Much like just taking simple rules and combination of letters, like th, ee, ea, and ing being common combinations they would be lumped in their own group, or things like I before E except after C.

 

[filthytrancendence]

I'm serious. Most of the weights used in data compression, and likely LLMs don't contain whole works of art and instead only pieces that have to be pieced together based on likelihood of appearance, negative appearance, or some type of numerical value to select the one it wants.

The above example is super simple because to expand to something far more complex, may take up a ton more space to get the general idea across. But with LLM's and other AI models, you're not looking at a couple characters back, that was just to show the complexity differences, when you start going larger actual examples like 'The quick brown fox jumps over the lazy dog' the complexity and having up to 26x26 tables JUST FOR 2 LETTERS (676 tables, though many would be culled since zz never appears for example).... Nevermind scanning something like the bible. Finally you'd probably have a singular table of single letters/symbols and their weights in cases where names new words and combinations weren't scanned in. So my name would be 'Ya' 'no' '2' 'm' 'ch' as 5 symbols rather than say Hercules which may be 1 or 2 symbols if greek mythology was included.

Also the more it's trained on more characters for lookback, even random input based on weighted models from likely possibilities will result in pseudo language and semi-comprehensive content (probably garbage but you'd see words or sets of words that actually appear in literature). Much like just taking simple rules and combination of letters, like th, ee, ea, and ing being common combinations they would be lumped in their own group, or things like I before E except after C.

Alright. I'll extend you the benefit of the doubt.

You're not going to find many writers who care to engage with you on this level of technical hair splitting. Myself, I've got a broad strokes understanding of how it works, and that's good enough for me. The math doesn't interest me much.

You're breezing past the bit where your model was trained on many works without the author's knowledge or consent. That's going to be 90% of authors first and only concern. Nobody cares that the functional model doesn't contain a copy of works it was trained on. It was derived from the works. That's the sticking point, and that's the pedantry I was pointing out. Nobody's going to engage with you if you breeze past that point with pedantry and technical jargon.

I'm semi-sympathetic to your cause. Of the authors here, I'm one of the only ones that is not vehemently anti-AI on its' face. Or at least I'm the only one I've seen try and put the brakes on the anti-AI train when I'm feeling masochistic. I do not share your enthusiasm for the technology, and am myself a techno-pessimist. But I think the extreme anti-AI position has become counterproductive to the interests of creative people, and so I attempt to preach moderation.

I don't think AI work has any business being published here, and I agree with the site's stance on that front. However, I don't think that all use of AI is definitionally or intrinsically morally wrong like many of my fellow authors seem to.

As far as the discussion you were having here before I got here, it seems to me what you were talking about is having a roleplay with an AI. Having done some text-based RP myself, it's my judgement that RP folks tend to overestimate how interesting the text of their RP is to anyone who is not involved in it. It's nowhere near a 1:1 for writing erotic fiction. And even if you could publish it, even if it were between two humans, I think you'd find very few readers interested in reading it here.

 

[yano2mch]

Alright. I'll extend you the benefit of the doubt.

You're not going to find many writers who care to engage with you on this level of technical hair splitting. Myself, I've got a broad strokes understanding of how it works, and that's good enough for me. The math doesn't interest me much.

You're breezing past the bit where your model was trained on many works without the author's knowledge or consent. That's going to be 90% of authors first and only concern. Nobody cares that the functional model doesn't contain a copy of works it was trained on. It was derived from the works. That's the sticking point, and that's the pedantry I was pointing out. Nobody's going to engage with you if you breeze past that point with pedantry and technical jargon.

Mhmm. Understood. But training it only affects weights a little bit per input.

On the other hand i personally think copyright is WAAAAAAY too long. Originally it's suppose to be 18-20 years, but Disney and corporations want to own everything forever. While the tradeoff with copyright (at least when it was put in with the printing press) is the knowledge is available to everyone so everyone wins, while only being able the sole owner of their own works for a short period of time.

I do think Copyright should have a minimum amount of time, but that time should be probably closer to 8 years, maybe 15, and then make royalties for commercial use on say TV last up to 40.

If the copyright is too long people who cared and passionately loved a product will long since leave because there's no chance to influence it in their or their grand children's lifetimes, and then you get stuff like Whinny the Pooh, blood and honey which takes something going public domain and then throwing on shock value because they can; instead of if it expired after 15 years then people who grew up with something would love it and then extend the original story in new and interesting ways, like maybe Christopher Robin and Whinny finding an old smoker and pointed top, then have adventures where they pretend it's a space ship and go to mars instead of in the back yard.

Then there's the complete opposite regarding copyright length where when Napster was a thing, bands put their music for free up and got a following they never would have, while established bands wanted to sue Napster because they weren't getting their nickel per play. Like the pharmaceutical companies, they want to take away the ladder that let them become successful so they are a monopoly.

I'm semi-sympathetic to your cause. Of the authors here, I'm one of the only ones that is not vehemently anti-AI on its' face. Or at least I'm the only one I've seen try and put the brakes on the anti-AI train when I'm feeling masochistic. I do not share your enthusiasm for the technology, and am myself a techno-pessimist. But I think the extreme anti-AI position has become counterproductive to the interests of creative people, and so I attempt to preach moderation.

I don't think AI work has any business being published here, and I agree with the site's stance on that front. However, I don't think that all use of AI is definitionally or intrinsically morally wrong like many of my fellow authors seem to.

I'm not sure i am enthusiastic of AI, yes it's an interesting new toy I'm playing with the last 2 months, and i try to see it for what it is and what it can really do. Ultimately I see AI as a tool, and honestly in a perfect world we wouldn't need it. It might be better if it didn't exist, or maybe it's improving the world in ways we aren't aware of yet due to it just hammering out brute-force style with designs we wouldn't have thought of, not too unlike an assembly programmer who reduces a code by 90% and is totally unreadable but is 5x more effective than the original source.

But ultimately there's a of things a lot people can't do. Like the availability of the Unity and Unreal Engines, it lowers the bar for who can creatively get involved in making content and games (Do you remember the 90's with the big fat books of Game making books including engines on par with DOOM? Yeah you needed a lot of patience and technical know-how to get off the ground even with sources included with compilers). In terms of what is even options when you consider the money involved but now can be done at a fraction of the price, and instead of needing huge team it can be done by a few or even a single person.

It is a lot like firearms is the great equalizer where even smaller weaker people can then defend themselves rather than those with brute strength determining the rules and just taking what they want.

As far as the discussion you were having here before I got here, it seems to me what you were talking about is having a roleplay with an AI. Having done some text-based RP myself, it's my judgement that RP folks tend to overestimate how interesting the text of their RP is to anyone who is not involved in it. It's nowhere near a 1:1 for writing erotic fiction. And even if you could publish it, even if it were between two humans, I think you'd find very few readers interested in reading it here.

Maybe, maybe not. A lot of the writing that i see comes out where if i dropped the Me/AI headers, and rearrange some of the inputs to interweave them it flows a lot like a book or story. Thankfully the AI's are pretty good at going 3rd person by default, rather than first person which is really weird to read unless you're solely looking from the main character's point of view.

I'd see it as an experiment, and ultimately either i get interest, or little-to-no interest. I'm fine with either, but if i got a lot of interest I'd know i could ramp up and take it seriously and maybe even consider more of a writing career. More likely I'd put up a dozen of the better results, and then drop it entirely due to lack of interest. I might even get to the point i don't need AI to RP with anymore and can just write without having that annoying mental writer's block, but I'm not there yet, and i may never be there.

 

[filthytrancendence]

I'm not sure i am enthusiastic of AI

Compared to your audience here, you're as giddy as a schoolgirl. Personally, I wish it'd go away. But it won't, so I'm making my peace with that.

But training it only affects weights a little bit per input.

Try and put yourself in an author's shoes for a second. Your art was scraped by some megacorporation's text crawling bot and fed into the woodchipper of AI training for the express purpose of putting you out of an already extremely tenuous job. Then somebody comes in here trying to tell you how the work that was used without your knowledge or consent doesn't matter that much anyway, because it only affects weights a little bit. Can you maybe see how that's going to generate some negative responses to your attitude?

All this talk of copyright is a complaint rightly levied against Disney. Not individual artists as an excuse to scrape their work and use it without permission. That's a bad look. Please stop.

I'd see it as an experiment, and ultimately either i get interest, or little-to-no interest. I'm fine with either, but if i got a lot of interest I'd know i could ramp up and take it seriously and maybe even consider more of a writing career. More likely I'd put up a dozen of the better results, and then drop it entirely due to lack of interest. I might even get to the point i don't need AI to RP with anymore and can just write without having that annoying mental writer's block, but I'm not there yet, and i may never be there.

I propose an alternative experiment: you just spend the time writing a story instead. Without AI. Start small. 1000 words. Learn the process. When it's done, publish it. Learn from the feedback. If you like it, go again.

When that's done, you'll definitely be at a point where you're a much more skilled writer. Writer's block is one of many things you can develop skills to overcome, if you choose to cultivate those skills. AI isn't the path to those skills. That's what the writers here are trying to explain to you.

 

[yano2mch]

Compared to your audience here, you're as giddy as a schoolgirl. Personally, I wish it'd go away. But it won't, so I'm making my peace with that.

Hopeful. And I'd suggest everyone try it, but i am not advocating we use it in everything. In fact i would rather it slow down. Give people a chance to adapt to the possibility rather than ingraining it into the OSes and every device and every search engine as soon as possible.

Try and put yourself in an author's shoes for a second. Your art was scraped by some megacorporation's text crawling bot and fed into the woodchipper of AI training for the express purpose of putting you out of an already extremely tenuous job. Then somebody comes in here trying to tell you how the work that was used without your knowledge or consent doesn't matter that much anyway, because it only affects weights a little bit. Can you maybe see how that's going to generate some negative responses to your attitude?

All this talk of copyright is a complaint rightly levied against Disney. Not individual artists as an excuse to scrape their work and use it without permission. That's a bad look. Please stop.

Again you misunderstand how it works. The training only makes it lean directions and have maybe the spirit of something, not lift it wholesale. Identifying word patterns a specific author may tend to use doesn't invalidate or steal any of their work (though it could use those increased weights to seem more like they wrote it), anymore than a picture of a car is stealing the physical car.

Example, being a computer geek I've read plenty of computer books. So if i describe RegEx (regular expressions) as a class of characters to identify matching a patterns without needing to be an exact match; For example \b[a-z]{4}\b will match any lowercase word of exactly 4 characters. NOW, You'd then argue I've now invalidated all RegEx authors writings, guides, manuals, howtos, sourcebooks and sourcecode who made it and wrote books on it for the last 50 years since i used the knowledge in those books just now? It's still under copyright, so i should never mention any part of RegEx as not to possibly give issues from said knowledge?

I agree authors artists and everyone should get paid for their work, we are well past the point of insanity with sheer amount of content long before AI came to the scene. There's more books than you could read in your lifetime, more songs than you can listen to in your lifetime, move movies than you can watch in your lifetime, more porn than you can fap to in your lifetime, etc etc. If you forbid access to something, people will just move to something else in a similar range and consume it instead and what you forbid will more likely be forgotten unless it's already heavily in the zeitgeist.

I propose an alternative experiment: you just spend the time writing a story instead. Without AI. Start small. 1000 words. Learn the process. When it's done, publish it. Learn from the feedback. If you like it, go again.

I wrote a story (about 50k long). It's rough, can't post it here as it features two teenagers as part of the main plot. Needs cleanups and rewrites in sections. But as i can't post it here, what's the point? It wasn't nearly as fun as ERP.

When that's done, you'll definitely be at a point where you're a much more skilled writer. Writer's block is one of many things you can develop skills to overcome, if you choose to cultivate those skills. AI isn't the path to those skills. That's what the writers here are trying to explain to you.

I don't need AI in order to write detailed and good content; I need AI in order to get past the writer's block i inadvertently suffer from. That's the whole point of wanting to do it as an RP and then convert to a story. Though unless i want to drag out say my longest 22k starter post intro (based on a RP in the Anita Blake world) you probably won't believe what, or how long i can really go before i just get stuck.

 

[Husky_Embrace]

@yano2mch the weights and such do not really matter much if you can do the following:

Data of other people in --> neural network --> data of other people out

It is not their data. However you store it, you still can get the exact words, or the meaning of it. You can ask what sentence were said on what page of GoT, you can ask a picture in the style of Studio Ghibli, you can ask how to code certain things in JAVA.

Does the AI have the rights, or financially compensated George RR Martin? Did Studio Ghibli say it is okay to ram their movies into the AI and replicate their style over and over? Did the AI put up a website for JAVA questions, gather a community where everyone helps each other, carefully maintain and grow it over years, just so it can now be used by AI?

It really doesn't matter how you store it in the model. Data that is not theirs goes in, data that is not theirs comes out. They take away revenue, jobs, as well as satisfaction of people in communities.

 

[Rob_Royale]

On the other hand i personally think copyright is WAAAAAAY too long. Originally it's suppose to be 18-20 years, but Disney and corporations want to own everything forever. While the tradeoff with copyright (at least when it was put in with the printing press) is the knowledge is available to everyone so everyone wins, while only being able the sole owner of their own works for a short period of time.

I do think Copyright should have a minimum amount of time, but that time should be probably closer to 8 years, maybe 15, and then make royalties for commercial use on say TV last up to 40.

If the copyright is too long people who cared and passionately loved a product will long since leave because there's no chance to influence it in their or their grand children's lifetimes, and then you get stuff like Whinny the Pooh, blood and honey which takes something going public domain and then throwing on shock value because they can; instead of if it expired after 15 years then people who grew up with something would love it and then extend the original story in new and interesting ways, like maybe Christopher Robin and Whinny finding an old smoker and pointed top, then have adventures where they pretend it's a space ship and go to mars instead of in the back yard.

What a crock of selfish shit is THAT? Why should anyone have the right to influence what someone else owns? By your logic after 30 years, someone else can just move into your home, or after 15 years, fuck your wife.
According to you, if they love them, they should have access to them and sooner rather than later.

 

[yano2mch]

I'm finding myself getting a little annoyed i have to play devil's advocate to provide the other side of the conversation and considerations. All i want to do is ERP and write stories.

@yano2mch the weights and such do not really matter much if you can do the following:

Data of other people in --> neural network --> data of other people out

It is not their data. However you store it, you still can get the exact words, or the meaning of it. You can ask what sentence were said on what page of GoT, you can ask a picture in the style of Studio Ghibli, you can ask how to code certain things in JAVA.

GoT? Not familiar with that. Though what the Neural network knows, and what a LLM having direct connection to the internet to search for something as part of a query is are completely different beast with different protocols. By default i'm assuming something offline, disconnected and isolated from the rest of the world. (though some programs i've seen where you can have it connect to your files, and thus it can expand it's access to knowledge nearly endlessly that way)

A style would be thickness of lines, presence of colors, maybe specific shapes of faces that in combination has a very distinct look. One of the youtubers i watch a lot, talks a lot about having worked for disney, and everyone is taught the Disney Style to which they want all their stuff to contain that style. But it is something that can be taught, and examples used for edge cases probably.

Asking it to code something in Java (or another language) overall would be inferring a basic layout of what you want and then converting to language of choice, then connecting most of the dots. It may have been a year ago when they were talking of having LLM's write 8bit BASIC code for Atari or Apple2, and it was.... close... not quite right, but nearly right. I've asked it questions on LFSR's and basic function creation using a Godzilla model (Godzilla, squashes your coding problems flat... or something), and it was basically right. But i didn't go far and deep into complex questions so i can't tell you how good or bad it is.

Does the AI have the rights, or financially compensated George RR Martin? Did Studio Ghibli say it is okay to ram their movies into the AI and replicate their style over and over? Did the AI put up a website for JAVA questions, gather a community where everyone helps each other, carefully maintain and grow it over years, just so it can now be used by AI?

We can take the questions and ask that of every fan, fanfiction writer, copycat or other person who puts up stuff based on their favorite franchise, and it's quite a different picture. It's like a game you can play when you see X posts, where they bash specific people and genders, but if you flip it it's suddenly racist and evil even when white/black male/female are swapped.

I'm going to back up and say as long as the copyright law is happy (it's 25% different at least from a created work) then you just have to let it go. The best way to remove or change a law, is to rigorously enforce it.

And no, while AI doesn't have rights, money, and didn't create a website, if it knows and understands something and i can access it (even when the internet is off, like mine tends to go) then it may very well be used more. Though i don't think AI should be throwing answers to you on the top of a search page and deterring you from going to sites either.

It really doesn't matter how you store it in the model. Data that is not theirs goes in, data that is not theirs comes out. They take away revenue, jobs, as well as satisfaction of people in communities.

If i can access it without a paywall, if it's a book i can get at the library, if it's in an encyclopedia meant for anyone to access, if it's ever been given away 'for free' or to claim, then i call it fair game.

You make it sound like this is an alien. Content is meant for consumption, and learning. The LLMs and image/video creation is meant for consumption. It's just acting as a different middle-man, and inferring knowledge. It can probably do a lot of other things too, but those things (automation) aren't where you're feeding it works of old and instead have to be made wholecloth for specific uses.

As for jobs... those going on strike or cities raising minimum wages are far more likely to cause the loss of jobs. Only those with short term profits in mind can't see you can't have businesses without jobs. But there seems to be enough that aren't thinking long term.

What a crock of selfish shit is THAT? Why should anyone have the right to influence what someone else owns? By your logic after 30 years, someone else can just move into your home, or after 15 years, fuck your wife.
According to you, if they love them, they should have access to them and sooner rather than later.

You're right, maybe we should go back before there was copyright, when we were controlled by kings, they could just come and take whatever they wanted without asking, no people rights and yes they'd burn your house and fuck your wife, and if you didn't like it you could die in the woods fighting the wolves.


You DO realize that copyright law suggests EVERYTHING (at least intellectual and creative) eventually goes public domain correct? The only difference is when. There's a nice video that's a bit old, but still totally valid called Forever Minus one Day, which talks over this. The wording of copyright is To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.

Do note LIMITED TIME which used to be as short as 18 years. This doesn't suggest you won't be credited still, but you won't have the sole COPYRIGHT, or rather who can be granted the RIGHT to COPY something.

Though a number of companies like gaming companies that make content and streaming companies, are happy to destroy something so it will never go public domain, because it won't make them billions of dollars if they provide offline and playable entertainment after their servers shut down. I feel this is a violation of the intent of copyright... and needs to be addressed soon.


And you as for someone else owning my house and fucking my wife, you're confusing physical property with ideas and concepts; I can't stop you from imagining my wife and fucking her in your head (though that would be weird though), but physical property is nowhere near the same idea.

 

[Britva415]

I'm finding myself getting a little annoyed i have to play devil's advocate to provide the other side of the conversation and considerations. All i want to do is ERP and write stories.

So why are you playing devil's advocate at all? You don't "have to," you could just stop baiting that hook. You know what you're going to catch on it. That doesn't make you or anyone else right or wrong, it just makes you invested, for some reason, in the act of arguing.

How bad do you want it? You've probably put at least 5000 words in this discussion alone. No writer's block here, I guess, but also no discernible effort toward overcoming it and going and doing the things you say you want to do.

 

[yano2mch]

So why are you playing devil's advocate at all? You don't "have to," you could just stop baiting that hook. You know what you're going to catch on it. That doesn't make you or anyone else right or wrong, it just makes you invested, for some reason, in the act of arguing.

Because i dislike the idea of hating something because you don't understand it. Not too unlike showing someone a flashlight for the first time when electricity is becoming a thing, they will think the devil made it and smash it and forbid their families from using it and instead will still use candles and lamps, without understanding that electricity plus a bulb creates light that doesn't require heat in order to work.

I also hate the hypocrisy. People expect you to write on your own, and then suggest you look at other great writers and learn from books, but the moment it might be handed to a program it's suddenly taboo and forbidden. So instead you suggest only limited and inhouse creations that's much smaller and inferior to be handed to a computer, but humans are intended to use that version and would be handed inferior information or generation when prompted. These are the same people who are happy to buy an Apple iPhone for thousands of dollars, and the moment it's useful to them they will overlook the very thing they say they don't want because it makes their lives easier.

How bad do you want it? You've probably put at least 5000 words in this discussion alone.

How bad do i want AI? Depends on what it can do. I've always been picked last, and never given a chance to get a high paying job even with high skills and potential. As such if an inexpensive and powerful product is available and i feel i can trust it, i'll want it. But if it doesn't meet what i want or minimum thresholds of requirements i don't want it. It's black and white.

No writer's block here, I guess

...... Maybe i was too subtle when i said i need interaction and feedback...

As i obviously don't get any feedback or suggestions here, i never have anything to reply to.

Nope....

 

[Britva415]

Because i dislike the idea of hating something because you don't understand it

I respect that.

I just don't believe that it requires anyone to make it their job to teach everyone else a lesson.

And then BLAME everyone else for making them do it.

 

[sxPlasticity]

Okay, y'all, simmer down, simmer down...

The point of my question was to find people interested in AI, not start a flame war.

I have not been able to read everything, but several things have been said, which I can cursorily comment on as the OP. If you disagree with them, start another thread; but it seems only fair to keep to the point of what I was asking.

The first thing is, AI is a technique, and it is everywhere now. It will remain. You may not like it, and it certainly has its problems, but you cant escape the fact that it exists.

Second, you all might want to take a look at your responses. AI is basically a programmed mirror. And, if you were to speak to yourselves in the way you are addressing some of these issues; well, assuming you would not want someone speaking to you in the manner you speak to others - it's no wonder you dont like AI...

Finally, if you actually look at the way I phrased the original question, I proposed a human based project to OVERWRITE AI-generated content. The origins of the project were a very basic outline, as stated in the OP. A story was built by AI upon it. The idea clearly then to rewrite the AI generated story. This is in keeping with traditional artistic and scientific practices, whereby our knowledge and customs advance themselves by building upon established forms; pretty much all art does that - even stuff life Bach and Mozart, Technology, and so on. With respect to having sex, there's certainly some pretty creative ideas out there, but if you ethically go too far out of line - I dont think I would blame AI, but you're more likely to just have to police show up at your door - the laws and ethics are basically that straightforward.

And, I guess that the point of what I wanted to ask in proposing such a project... AI generated porn is not humane; but you can't remove the human element from the results it generates, either. So? Its not going anywhere, and you could try to fight it all you want - but it just is what it is... and it can go and go and go... I think the first reply made a good point, on point... I dont know if I would say it it completely realistic because AI is everywhere now, and you can't just pull the plug... But at some point, you have to tell it to stop... and there's no fighting that... whatever "stop" happens to mean at that point...

To those interested in continuing with the initial intention of the thread, feel free to email me.

Ty.

 

[Britva415]

How bad do i want AI?

No, how bad do you want to ERP and write.

 

[Britva415]

if you actually look at the way I phrased the original question, I proposed a human based project to OVERWRITE AI-generated content. The origins of the project were a very basic outline, as stated in the OP. A story was built by AI upon it. The idea clearly then to rewrite the AI generated story.

I already asked if it wouldn't just be easier to write around the outline you already have, instead of giving it to a LLM and then un-doing what it gives you back.

I guess that when I asked that, I wasn't really paying attention to the part where not only do you seem to not want to save yourself that trouble and instead just write the story from your outline, you don't even want to do the rewriting part - you're trying to crowdsource that part to other humans.

Why wouldn't you just go directly to asking them to just do the writing? Why is AI even in this process at all?

 

[yano2mch]

And, I guess that the point of what I wanted to ask in proposing such a project... AI generated porn is not humane; but you can't remove the human element from the results it generates, either.

Depends on the porn. I'd be curious to see porn in cases where it would obviously never be possible, say superman with wonder woman floating in space. Actually most of them being non-human or super powers seem to be the main component.

But just making movies with AI may reduce from hundreds of millions of dollars to a low fraction of that cost, and we might start seeing favorite books becoming movies or series, even if they aren't making enough money to justify lead actors and high production costs normally.

As a kid i read the Juxtaposition series by Piers Anthony (one of the few authors i read as a kid), seven books long involving an alternate universes, fantasy and tech, love, question of intelligence being machine vs man, aliens. But it also included a lot of sex (in it's limited young adult writing fashion). I'd love to see that; But it will never be made unless a method is cheap enough, and someone lovingly tries to transfer it from page to the screen. Not to mention the amount of sex would put it Unrated or X, even if it's not outright porn.

No, how bad do you want to ERP and write.

ERP? One of my favorite pasttimes... but finding a partner is a pain, i can spend hours looking and not find one. I eventually just went to a passive mode spending more time downloading stuff that interests me to sort later.

Write? I'd love to write. Wanted to write books when i was 14. But i VERY quickly found multiple things i couldn't do or get past: Planning, patience, problem solving.... and social interactions as i have trouble reading body language and people (though Anime has taught me a LOT in the last 5 years) .

So now nearly 30 years later, it might actually be an option, though with the saturation of the books market i doubt i'd ever make it big. But there's still hurdles i can't get past on my own.