|
people can still do it for fun / hobby. I'm pretty sure the Kindle writing is a way for these people to make a living off of this. Do fanfic sites actually pay you?
|
#
?
Mar 15, 2023 19:38
|
|
|
|
| # ? Jun 5, 2024 01:38 |
|
|
No one will be making a living off low end anything if I can get one of those short contract third world call centres to feed a constant stream of semi edited AI novels to my 5man head office. We'll allocate one hour of final review per half edited manuscript and immediately self publish. I just hope they don't add a box on amazon self pub for 'I declare this is not AI generated' 
|
|
#
?
Mar 15, 2023 19:46
|
|
|
pixaal posted:There's reddit communities about writing for really niche genres like find a specific form of time travel (or fetish, that's popular) and write poo poo that wouldn't even get a C in highschool. It sells because so little exists in the genre if someone wants more you come up as the next best thing. Yeah this is basically what I see happening. Good authors will be fine, but if people can now just ask the AI for a story set in the 1700s where aliens invade and France, the UK, and Russia team up to kick them out or whatever then it's going to weed out a lot of other authors.
|
|
#
?
Mar 15, 2023 19:49
|
|
|
Mordiceius posted:I've had some thoughts on AI lately. ChatGPT v4 is supposedly able to handle up to ~25000 words, or ~50 pages
|
|
#
?
Mar 15, 2023 19:53
|
|
|
I mean, I'd imagine ChatGPT (especially the new version) can write a story for you, as long as you work through it granularly: ask it for a basic outline first, then separate that into chunks and work your way down to a chapter-by-chapter outline, and finally the draft itself. Otoh there are people out there than can churn out a fiction novel in a week (quality notwithstanding), so I'm skeptical this would even have that much of an impact. It's definitely not as simple as giving it a 'write me a story' prompt anyway.
|
|
#
?
Mar 15, 2023 19:58
|
|
|
Fwiw the Kindle market has been 90% computer assisted garbage for years. Long ago people figured out they could recycle existing content by throwing it in a blender with SEO tools designed to essentially "paraphrase this paragraph" over and over with very little editing. Sometimes it's obvious, synonyms that don't quite fit, gibberish plot, but the ones where the "writer" spent 15 minutes fixing issues you probably wouldn't ever know it was essentially a mishmash of plagiarism.
|
|
#
?
Mar 15, 2023 19:58
|
|
|
Ovenmaster posted:I mean, I'd imagine ChatGPT (especially the new version) can write a story for you, as long as you work through it granularly: ask it for a basic outline first, then separate that into chunks and work your way down to a chapter-by-chapter outline, and finally the draft itself. I'm not even saying people are going to flood that market place I'm saying someone is going to release a website you pay $5/month for that will write you all the smut you want about any fetish you want. They wont need to buy from anyone else.
|
|
#
?
Mar 15, 2023 20:05
|
|
|
I feel like the writing market won't be too badly affected, in all honesty. If you want a "good" AI novel, it's still going to require quite a bit of guidance, direction, and care by a human editor. The people impacted the most are those people churning out large amounts of already low quality products. And honestly, gently caress those people. The other thing I wonder about is AI art in game development. Where do people fall on that? Let's say I wanted to make an iphone app and had no artistic skill myself but could craft some images via AI generation that would work. Would people even know? Would people even care?
|
|
#
?
Mar 15, 2023 20:06
|
|
|
Rule34-GPT
|
|
#
?
Mar 15, 2023 20:06
|
|
|
Mordiceius posted:I feel like the writing market won't be too badly affected, in all honesty. If you want a "good" AI novel, it's still going to require quite a bit of guidance, direction, and care by a human editor. Much like the writing, whether people know is based on how much effort you put in trying to hide that fact, and people will care, but mostly lovely people who just like being mad about things.
|
|
#
?
Mar 15, 2023 20:11
|
|
|
Mordiceius posted:
Be honest about it and up front some people will care. Some people will buy it to see it in action, others will boo and boycott it. High on Life uses AI art, and it's really obvious because some of it is barely modified and looks like someone who finally got the hang of prompting and has 20-40 hours under the belt. That's a hard test to use because this was released just before the creator's scandal in Rick and Morty came out. https://kotaku.com/high-on-life-justin-roiland-ai-art-rick-morty-1849900835 - around the time of release before the scandal. I think attempting to hide it especially through omitting will get retaliation. Some people are very against AI and don't want to consume anything by it. They get angry if you trick them. Other than tricking someone there's little reason to leave it out especially if you are getting good images and they are impressive AI images.
|
|
#
?
Mar 15, 2023 20:12
|
|
|
pixaal posted:Be honest about it and up front some people will care. Some people will buy it to see it in action, others will boo and boycott it. High on Life uses AI art, and it's really obvious because some of it is barely modified and looks like someone who finally got the hang of prompting and has 20-40 hours under the belt. Counterpoint: gently caress those people. It's not a trick, and you aren't under any obligation to explain your work process. There isn't a single game made today that doesn't use ai tools in some capacity, anyone telling you otherwise is either an idiot or intentionally lying to you.
|
|
#
?
Mar 15, 2023 20:24
|
|
|
PlaysGamesWrong posted:Counterpoint: gently caress those people. It's not a trick, and you aren't under any obligation to explain your work process. I mean yeah, but at the same time I think letting people know we used AI and this is awesome is cool and likely a bigger boon than negative so why not? They'll slowly come around, people did with photoshop... and computers in the workplace... mostly on both counts.
|
|
#
?
Mar 15, 2023 20:27
|
|
|
I'm also curious about how people treat it when you use AI art as part of greater pieces. Like, if I use AI to generate a luscious green fantasy landscape, and then modify it adding borders or painting in little details, do people really give a poo poo that the rolling hills were AI generated? But even further - would people even known just by glancing at it? Maybe I'm just blind but I, personally, have a really hard time telling a lot of AI art from hand crafted works.
|
|
#
?
Mar 15, 2023 20:35
|
|
|
For me personally all I care about is my own experience. As an example I know that stable diffusion is being used a lot in the indie game creation community to do tasks like generating tons of different shapes and colors of player cards and gems and totems and whatnot, cool, whatever, they generally look neat and a human doesn't need to spend hours or days working through all the permutations of a simple concept. However if the entire game scene turns into just a flood of AI generated garbage including storyline and audio and video and images and whatnot then yeah that will suck and I'd be against it. I think it will bubble at the start and everything you see and here and look at for a while is going to be AI generated nonsense because everyone's trying to find a better way to build a mousetrap but eventually it'll settle into its place as the tool that it really can be, and that's fine with me.
|
|
#
?
Mar 15, 2023 20:46
|
|
|
Mordiceius posted:I'm also curious about how people treat it when you use AI art as part of greater pieces. Like, if I use AI to generate a luscious green fantasy landscape, and then modify it adding borders or painting in little details, do people really give a poo poo that the rolling hills were AI generated? But even further - would people even known just by glancing at it? The militant people will call for your boycott/literal murder. It's become a "cause" and reason has gone out the window. The detectability though is hit or miss. Unedited and original aspect ratio stuff is likely to be noticed by a trained eye, especially if there are details that the ai doesn't get quite right. You can also generate art that is indistinguishable by humans with a bit of effort. The problem is that the answer to this question might be completely different next week. I don't know that anyone has tested the ability to fool digital forensics just yet though.
|
|
#
?
Mar 15, 2023 20:47
|
|
|
PlaysGamesWrong posted:I don't know that anyone has tested the ability to fool digital forensics just yet though. I've been poking at the Hive AI content detection tool to try and understand exactly how it detects AI content. The #1 way seems to be it's use of colors, at least with MJ images. MJ is bad about using color compared to a human artist and a lot of the shadow and highlight colors it chooses are just kind of awkward and weird (e.g. mustard green pixels shading a purple object) - it looks fine on the context of the image itself but if you look closely it's wonky. So just scaling Chroma up/down or making an image gray-scale will do most of the work fooling it. What's interesting though is you can index an image as grayscale then switch back to RGB and manually colorize it, and it will start getting detected as AI generated again. Getting around that is easy to do by indexing the grayscale image with a limited palette (e.g. 12 color grayscale) before switching back to rgb and colorizing. There are some "standard objects" that MJ always makes in similar ways - trees and houses are two big ones in my experience. Hive will generally detect those objects no matter what you do to them. You can slap a 16 pixel superpixel filter on then index it down to like 6 colors so it looks like a posterized lithograph printing cutout and is unrecognizable as the original AI work, and hive will still detect it as MJ with 60%+ certainty. My guess* is that it's trained similarly to how most machine learning stuff is trained - it looks at thousands of known AI images and then looks for similar elements in whatever you point it at. So standardized objects like houses or trees, or standard angles/compositions, or things that happen in all AI images (e.g. bad color use) get picked up but it's not really much more sophisticated than that. *This guess is based on anecdotal gut instinct deep dish peat moss fucked around with this message at 21:14 on Mar 15, 2023 |
|
#
?
Mar 15, 2023 20:58
|
|
|
There is 'exciting' AI art debate happening right now in the TG as an Industry thread in Tradgames, you can see what the normies think. There's already been a mod warning.
|
|
#
?
Mar 15, 2023 20:58
|
|
|
Mordiceius posted:I feel like the writing market won't be too badly affected, in all honesty. If you want a "good" AI novel, it's still going to require quite a bit of guidance, direction, and care by a human editor. High on life used ai generated art to make the posters on your bedroom wall near the beginning of the game. I thought it was pretty neat, and it doesn't bother me that it was ai generated. I have to assume the vast majority of people won't know, won't notice, or won't care
|
|
#
?
Mar 15, 2023 21:12
|
|
|
Also I have a post I'll make soon about my new process for using ai in the art assets creation flow for my game. I decided to lean heavier into human pixelart but I am using AI as part of the initial sketching process
|
|
#
?
Mar 15, 2023 21:15
|
|
|
PlaysGamesWrong posted:Counterpoint: gently caress those people. It's not a trick, and you aren't under any obligation to explain your work process. I mean this is true, but hop onto artstation (a site I love to hate) and you immediately see large quantities of technically competent, really awful soulless art that has come out of a 'professional workflow' and is proud of it. So yeah that won't change but nor will it improve. I guess it's functional. The site had a big old no to AI protest when this all came out though because as far as they're concerned a 50% efficiency gain means 50% of them lose their jobs.
|
|
#
?
Mar 15, 2023 21:27
|
|
|
It's nothing complex but here's the workflow I am using for hand-drawn (not filter-based) pixel art scenery: Here are some examples made this week using that technique (none are "finished")  Personally I feel pretty good about using AI to do the outlines and then filling them in myself, though I can see why some people would say it's "less artistic" - I felt worse about it when the AI was doing the whole thing and I would mostly just filter it and call it done. Here are some pixel art planets I drew recently. One of them used this method, the other 5 are all purely hand-drawn. It's kind of obvious which one IMO, but that's largely because MJ sucks at generating planets (Hint: it's the one that looks like a circular slice of a map rather than a round sphere) 
deep dish peat moss fucked around with this message at 21:47 on Mar 15, 2023 |
|
#
?
Mar 15, 2023 21:42
|
|
|
I caved and subbed to the Chat GPT+ thing to mess with GTP-4 and boy howdy. If you play pen and paper RPGs and like me, think Shadowrun rules, well this thing has you covered. If you don't know or care about Shadowrun then this might not mean much, but I am wholly impressed. Again. My Prompt to GPT-4 posted:I would like you to write me a backstory for a 40 year old Street Samurai (he is a Hawaiian dwarf) in the Shadowrun Universe. In the story the year is currently 2058. This 40 year old Dwarf had a father he was close with that was uncomfortably close with the corporate world and specifically the dragon Dunkelzahn. GPT-4 posted:Name: Kainalu "Tidal Wave" Kahananui I am by no means some Shadowrun lore expert, but I'm reading this and going  This is wild, wild poo poo to me, for it to understand context like it does completely in universe. Feed it a few details and it goes hog-wild, well above and beyond anything I'd write for my own character, let alone an NPC. It understands that metahumans typically have it tough, it understands the whole corporate world thing, but beyond all that it gave the dude a nickname and then at the very end used the theme of his nickname to end the text as a story hook tying it all together. It named his father, tied in where he was from with Hawaiian lore, sprinkled in a lot of hooks for other ways to enrich the story. A whole lot of role players are about to get way better stories in their games for sure. EDIT lmao gently caress My Continued Prompt to GPT-4 posted:Can you describe Kainalu, physically? GPT-4 posted:Kainalu "Tidal Wave" Kahananui is a distinctive figure, embodying both the strength of his dwarf heritage and the grace of his Hawaiian roots. Standing at 4 feet 2 inches tall, he possesses a solid, muscular build that belies his small stature. His broad shoulders and powerful limbs are a testament to his years of rigorous training in the martial art of Lua and the grueling life of a street samurai. My Continued Prompt to GPT-4 posted:Can you describe Kainalu's childhood friend? Make them human, but otherwise the details can be anything. It's also up to you to decide if they are still friends now that they are adults, and how this friend fits into Kainalu's life. GPT-4 posted:Name: Leilani "Whirlwind" Kealoha My Continued Prompt to GPT-4 posted:Can you describe Leilani physically? GPT-4 posted:Leilani "Whirlwind" Kealoha is a striking figure who embodies the grace and power of the Hawaiian islands. Standing at 5 feet 8 inches tall, she possesses a lean, athletic build, honed through years of rigorous training in the martial art of Kapu Kuialua. Her toned muscles and agile movements speak to her dedication and expertise in her chosen discipline. this is loving nuts Just spin up fully formed NPCs or hell, even main characters in minutes. EDIT x2 I won't keep posting blocks of text because boring as poo poo no matter the context, BUT I asked it to make the Tidal Wave guys dad, and it made a mistake. It made him a dwarf too, but born in 1990. I told GTP4 this fact, and asked it to change the father into a human but also take into account UGE (Unexplained Genetic Expression, when elves and dwarves were born to human parents). It gave me the same story with the details changed and a section on him dealing with his UGE baby. KakerMix fucked around with this message at 22:24 on Mar 15, 2023 |
|
#
?
Mar 15, 2023 21:46
|
|
|
So I tried new v5 of Midjourney. I am not impressed at all. the results are comparable to v4, and are sometimes even worse. Maybe its because the prompts changed, but I really don't think its worth checking out at the moment. Ugh, and apparently now the images display as loving webp on the website. That file format is a blight on the internet. IShallRiseAgain fucked around with this message at 23:15 on Mar 15, 2023 |
|
#
?
Mar 15, 2023 23:13
|
|
|
IShallRiseAgain posted:So I tried new v5 of Midjourney. I am not impressed at all. the results are comparable to v4, and are sometimes even worse. Maybe its because the prompts changed, but I really don't think its worth checking out at the moment. When v4 came out I felt the same way, but at this point having tested v4 vs v3, v4 is way better. In sure it will be the same with v5. Here's a compare and contrast with exactly identical prompts: V4  V5 
|
|
#
?
Mar 15, 2023 23:27
|
|
|
Paying for GPT4 has made me realize I am too stupid to even come up with creative things to do with it. Back to making weird logos with stable diffusion I guess 🤷♂️
|
|
#
?
Mar 15, 2023 23:29
|
|
|
IShallRiseAgain posted:So I tried new v5 of Midjourney. I am not impressed at all. the results are comparable to v4, and are sometimes even worse. Maybe its because the prompts changed, but I really don't think its worth checking out at the moment. It's a lot better at certain things, namely that the images are much higher resolution and details are a lot cleaner. You no longer need to upscale, it generates a high-res version right from the get-go. Pressing the Upscale version instantly sends you the full-sized .png version This is just some experiments with re-running old prompts:      It's not immediately obvious without seeing the old generations but the small details are so much cleaner now, it's astounding. deep dish peat moss fucked around with this message at 23:44 on Mar 15, 2023 |
|
#
?
Mar 15, 2023 23:30
|
|
|
cr0y posted:Paying for GPT4 has made me realize I am too stupid to even come up with creative things to do with it. ask gpt4 to suggest creative things you could do with it
|
|
#
?
Mar 15, 2023 23:30
|
|
|
MJ v4 and up are supposed to have deterministic seeds so here's the exact same prompt + seed in v4 and v5 for comparison. v4:  v5: Edit: Actually these are both v5 whoops  I don't have a non-upscaled version of this v4 generation. The first one is a v4 generation that got upscaled in v5Here's another using a more unique style: v4:  This was actually upscaled at the time of v4 v5:  
deep dish peat moss fucked around with this message at 00:34 on Mar 16, 2023 |
|
#
?
Mar 15, 2023 23:36
|
|
|
With just a couple tests on my end so far, v5 seems leaps and bounds better at making hands. Most of my tests had the right amount of fingers!
|
|
#
?
Mar 15, 2023 23:47
|
|
|
Also worth noting is that v5 brings back a lot of old v3 features that were missing from v4: --iw for image weight, --tile for seamless tiling, and aspect ratios greater than 2:1
|
|
#
?
Mar 15, 2023 23:57
|
|
|
Megazver posted:There is 'exciting' AI art debate happening right now in the TG as an Industry thread in Tradgames, you can see what the normies think. There's already been a mod warning. It sucks, I get it because a lot of the details are pretty hard to understand but I wish people wouldn't so confidently read things like a study on overfitting, be completely certain they understand exactly what's going on and argue about that with a bunch of other people who also know very little about it. My masters is in psychology, so I'm very, very used to that. Psych is one of those fields that research in gets wildly misinterpreted by people with no background in it and weaponized to make some point. Also the argument about tracing using a bunch of img2img examples. That's where AI really is the most like Photoshop, and people can do some very theft like things with it. But people see that and are like "this is just what the AI does this is all of it" cuz, again, they don't really know how it actually works. But I don't know the answer, you cant just be like "you can't have an opinion until you get your doctorate in machine learning!" but it is a very easily misunderstood thing.
|
|
#
?
Mar 16, 2023 00:54
|
|
|
People thought 486 computers were gods. People will learn by experiencing them. Computers in the workplace met resistance as well.
|
|
#
?
Mar 16, 2023 01:04
|
|
|
BrainDance posted:It sucks, I get it because a lot of the details are pretty hard to understand but I wish people wouldn't so confidently read things like a study on overfitting, be completely certain they understand exactly what's going on and argue about that with a bunch of other people who also know very little about it. I've yet to meet anyone who is both Anti-AI and has spent a significant enough amount of time with it to even understand what it does or what it's capable of, usually they maybe used the MJ free trial to generate a few pictures of Spiderman or whatever and then declared "This is a plagiarism machine!" (because they literally used it to plagiarize pictures of Spiderman, of course). Like, I get that someone who is fundamentally against AI is unlikely to sign up and spend time with it, but it's pretty telling that no one who signs up and spends significant time with it comes away going "Yeah, this is a theft machine that's going to obsolete artists." - they pretty universally come away from it saying "Holy poo poo, there's so much that artists can do with this!" But it does sort of redefine the term "Artist" because "someone who can output a visual product in technical detail" is now no more of an "artist" than "someone who has good ideas and can coalesce them together into a neat package" and I guess that's why it rubs some types of traditional artists the wrong way - the "artistry" of using AI is mostly conceptual, compared to the "artistry" of drawing or painting which is mostly technical. I got into traditional art with the express purpose of coalescing a bunch of things together into a neat package, so I've been on board with AI the entire way. MJ5 is a big leap toward cutting that technical kind of artist out of the process, and I think there's a large and vocal enough anti-AI crowd that "purist" traditional human art will always have about the same cultural relevance it does now (that is: not much), but within a few years the commercial space will be dominated by artists who are using AI for most of the technical work and focusing on the conceptual work themselves. I think hybrid art is going to be the way of things - e.g. a human artist hand-drawing expressive characters, but using AI for the less-meaningful backgrounds and nameless unimportant characters. There will naturally be a whole lot of garbage-tier pure-AI low-effort bullshit, too, but there's already a whole lot of low-effort bullshit with market share in the art space. deep dish peat moss fucked around with this message at 01:27 on Mar 16, 2023 |
|
#
?
Mar 16, 2023 01:14
|
|
|
 I think it's a little better at the hands, yeah.
|
|
#
?
Mar 16, 2023 01:28
|
|
|
Megazver posted:I think it's a little better at the hands, yeah.
|
|
#
?
Mar 16, 2023 01:56
|
|
|
v5 pretty much has guns down, too. HEre are a few generations of Space Ghost in Warhammer 40k. The trick is that you have to be pretty specific about what the gun is doing in the scene. Not mentioning the gun at all, just expecting it based on the figure being a soldier:  Not mentioning the gun but specifying that the character is aiming at someone off-camera:  (Note: the inclusion of the phrase 'camera' made all four generations look like photos of WH40k miniatures, presumably being interpreted as camera photos of wh40k things, which would be minis - this may not be true if you are using subjects other than WH40k) Not mentioning the gun but specifying that the character is aiming at someone off-screen:  Specifying that the character has a sniper rifle:  Specifying that the character is a sniper:  Specifying that the character is aiming down the scope of a sniper rifle:  Specifying that the character is "aiming at someone off-screen aiming at the viewer"  "aiming at someone off-screen, aiming at the 4th camera"  Specifying that the character is "aiming at the screen" (i think it's interpreted as them like, aiming at a computer screen up above them) 
deep dish peat moss fucked around with this message at 02:29 on Mar 16, 2023 |
|
#
?
Mar 16, 2023 02:07
|
|
|
deep dish peat moss posted:v5 pretty much has guns down, too.
|
|
#
?
Mar 16, 2023 02:56
|
|
|
I'm so impressed with V5 that I don't even know what to use it for. I remember having this problem with v4 at first too but it's like a whole new world of possibilities. My biggest complaint about V4 was that it was never good at mixing styles - if a certain part of the prompt was heavily associated with a specific visual style, then any prompts using those words would use the style they were associated with and ignore any other style cues in the prompt. My example in the early V4 days was Warhammer 40k. You could generate a photo of Shaquille O'Neal as a Wh40k space marine easily, but there was only one art style it would generate in (the standard WH40k character profile style) no matter what else you added to the prompt (style cues like "oil painting" or "illustration") unless it was something equally-distinct and universal, like Pixel Art V5 still does a little bit of that (like my WH40k + camera example in the last post - any WH40k image with the word 'camera' in the prompt is almost guaranteed to look like a photo of WH40k miniatures) but it's much, much more relaxed and it at least makes the attempt to blend styles together, like V3 did. Here's Shaquille O'Neal as a WH40k space marine, in the style of a pixelart screenshot from a SNES game from 1994.  Does it look like Shaq? No. Does it look like WH40k? No. But it does look like a screenshot from a 1994 game featuring a large-statured man who dresses vaguely like a WH40k space marine, which is something V4 can't do. Here's the same prompt run in V4:  Which looks like all of the other "Shaq as a WH40k space marine" images from V4 (4a in particular in this case) :  with a bad pixel art filter applied, and it doesn't bother trying to look like a screenshot from a SNES game from 1994. The V5 image example is arguably a "lower-quality image" here but my point here isn't about quality, but about flexibility of V5 to get away from the "standard universal style" of v4 e: Another takeaway of this post and my last post: MJ doesn't really understand objects being "on-camera" as opposed to "off-camera" - like it doesn't consider the image it is providing you as a "screen" that is viewing what a "camera" shows you. It doesn't really understand "in the image" as opposed to "out-of-image" either because it tries often tries to insert an image into the image. I'm stumped on what word to use to refer to that concept in a way that's unambiguous to the AI, but "off-screen" works as a half-measure for now - it's just difficult to then finesse it into understanding off-screen in what direction (like toward you, the viewer - or to the left, or toward the background - how do I refer to something that's off-screen because it's behind the camera/viewpoint/person who is viewing the picture?). I think it might understand "the 4th wall" but other times it inserts a wall into the image (or four walls). "The 4th camera" kind of works too (it's a sitcom thing - camera 3/4 are usually the cameras the character on the left of the screen can look dead into) deep dish peat moss fucked around with this message at 04:27 on Mar 16, 2023 |
|
#
?
Mar 16, 2023 04:13
|
|
|
|
| # ? Jun 5, 2024 01:38 |
|
|
Someone should give gpt4 a scan of the voynich manuscript and see if it figures it out.
|
|
#
?
Mar 16, 2023 06:20
|
|