|
As predicted the whole George Carlin thing was a mechanical turk https://arstechnica.com/ai/2024/01/george-carlins-heirs-sue-comedy-podcast-over-ai-generated-impression/
|
#
?
Jan 29, 2024 04:55
|
|
|
|
| # ? May 31, 2024 14:11 |
|
|
fez_machine posted:As predicted the whole George Carlin thing was a mechanical turk legally this would be george carlin fanfiction
|
|
#
?
Jan 29, 2024 06:24
|
|
|
I am fairly sure that it is not legally fan fiction, given that they used Carlin's voice.
|
|
#
?
Jan 29, 2024 17:10
|
|
|
|
|
#
?
Jan 29, 2024 17:52
|
|
|
null_pointer posted:I am fairly sure that it is not legally fan fiction, given that they used Carlin's voice. Did they, though? It doesn't sound anything like him. Also, I ran the transcript of the video through https://gptzero.me/ and it did come back as 0% AI written, 100% human written, so it'll be interesting to see how this plays out. I'm not terribly interested in being lied to by liars though, so I'll be avoiding anything the Dudesy crew get up to from here on forward, including Will Sasso, in perpetuity.
|
|
#
?
Jan 29, 2024 18:03
|
|
|
Fair point. The question is whether or not the lawsuits that are moving forward will find any purchase. I'm not sure if the chuckleheads who did this are going to be able to say "any reasonable person would never believe that this was George Carlin, it doesn't sound anything like him" and have it be an effective defense.
|
|
#
?
Jan 29, 2024 18:32
|
|
|
XYZAB posted:Also, I ran the transcript of the video through https://gptzero.me/ and it did come back as 0% AI written, 100% human written Those GPT detectors reliably spit out the truth as much as ChatGPT does itself.
|
|
#
?
Jan 29, 2024 20:16
|
|
|
You doing these in Bing? I similar VHS corruption sets I've done have been in Bing, currently trying SDXL and it's not liking making corruption or artifacts. But loving with the tiler on img2img might be fun in new ways Maybe I should look into making a LORA, but as far as I can tell you need 16GB VRAM to make them for XL and I have a 3060 mobile 8GB. It's a good little guy I don't like to make it render too too much being a laptop so I try to use online resources when possible. "I'm telling ya Marge was right behind me!"  Tiler with incorrect settings  e: loving around on Bing later in the day trying to make it make a guitar of Death Metal Jesus playing a guitar to emits smoke that forms visualizations as you play. (In my head this is winamp)    Bing is fun to make overly complicated requests at I'm surprised at how much it's able to pull in especially trying SD again recently (but man is SD amazing at everything else) pixaal fucked around with this message at 21:57 on Jan 29, 2024 |
|
#
?
Jan 29, 2024 20:19
|
|
|
If anyone is using ComfyUI but hasn't installed some script extensions, I found out about those when I went looking for how we're making these images with the workflow baked in it: I have no idea what I'm doing, any ideas with nodes?  So first I made these textures with Blender and Automatic1111 respectively, no particular reason.    I was getting into prompting in regions and the results with the example and base SDXL were pretty neat right away. I would need more time to fiddle with it though. These are candidly the first images prompted.    (you can really tell these three are others' prompts except I added geese or fruit, also I guess the resolution was wrong) This one I did a 3d model for the room in ComfyUI then used Automatic's inpainting to add the goose/pond.   (meanwhile some guy: "George Carlen is my birthright") moist banana bread fucked around with this message at 08:01 on Jan 30, 2024 |
|
#
?
Jan 30, 2024 07:30
|
|
|
pixaal posted:You doing these in Bing? I similar VHS corruption sets I've done have been in Bing, currently trying SDXL and it's not liking making corruption or artifacts. Yee, all Bing. It definitely nails -wave and -core aesthetics pretty well from my experience.     Stuff's happening in the lab hypercolor trippy cartoon Simpsons [interdimensional physics/abyssal science/cosmological science/genetics] lab
|
|
#
?
Jan 30, 2024 17:09
|
|
|
the dalle 2 version of Bing had a really cool concept of "made with sharpies and highlighters"   unfortunately I used bings "save" for most of them and now only the thumbnails are hosted they deleted the large copies because clicking them causes image not found.  Dalle 3 does not really understand this concept.
|
|
#
?
Jan 30, 2024 17:21
|
|
|
yeah, if you haven't downloaded your Bing images, do so now, because they wipe the images after about 2 months.
|
|
#
?
Jan 30, 2024 17:24
|
|
|
Not sure SDXL totally understood and it really wanted to include the markers and highlighters but got some neat characters.          
|
|
#
?
Jan 30, 2024 17:40
|
|
|
|
|
#
?
Jan 30, 2024 18:09
|
|
|
poo poo is so creatively bankrupt already in hollywood I'm not really mad about this happening. maybe live theater will become a bigger thing again.
|
|
#
?
Jan 30, 2024 18:13
|
|
|
I know what I want to use a TV / movie AI for and it's not to replace my normal viewing it's to make insane cross over episodes you'd never ever create to see what it comes up with. Let's see we're going to combine House, Dexter, and The Expanse. There is a murder on a space station and it is a race between House and Dexter to unmask the killer! Oh it will be so terrible but hopefully funny in the dumbest way as it slams in other existing unprompted characters from the shows.
|
|
#
?
Jan 30, 2024 18:22
|
|
|
pixaal posted:I know what I want to use a TV / movie AI for and it's not to replace my normal viewing it's to make insane cross over episodes you'd never ever create to see what it comes up with. mainstream llm AI is quickly moving in the direction where you'll need to "jailbreak" it if you want to use or reference any kind of intellectual property owned by various studios and media conglomerates and it will be considered illegal "piracy" to have them write crossover material without proper licensing from the brands involved, which they will happily sell you via various subscription packages. i've even seen dall-e balk at drawing pictures of sherlock holmes, famously now in public domain, because of its innate restrictions on known and established fictional characters. you can get it to draw holmes of course as well as pretty much any other character just by writing the prompt to get around those restrictions, but such restrictions are a lot more common and broad than they were like.. 6 months ago. Earwicker fucked around with this message at 18:40 on Jan 30, 2024 |
|
#
?
Jan 30, 2024 18:36
|
|
|
Earwicker posted:mainstream llm AI is quickly moving in the direction where you'll need to "jailbreak" it if you want to use or reference any kind of intellectual property owned by various studios and media conglomerates and it will be considered illegal "piracy" to have them write crossover material without proper licensing from the brands involved, which they will happily sell you via various subscription packages. I'm sure, but do you really think there wont be an open source version without all that?
|
|
#
?
Jan 30, 2024 18:39
|
|
|
pixaal posted:I'm sure, but do you really think there wont be an open source version without all that? there already are right now, mixtral for example is a pretty good one and competitive with gpt, but as to whether that situation will last and how depends on whether certain huge media companies try to shut it down.
|
|
#
?
Jan 30, 2024 18:42
|
|
|
LifeSunDeath posted:poo poo is so creatively bankrupt already in hollywood I'm not really mad about this happening. maybe live theater will become a bigger thing again. Yeah I mean I really don't see a downside, at all, to the current ~Hollywood~ ceasing to exist. We already get the most creatively void recycling of nostalgia (Did you guys see the new new new Ghostbusters preview? lmao) that having AI do it all seems like we'd get exactly the same thing we have now but with 100% less sexual abuse. Can't take advantage of a person sexually if they don't exist in the first place! Earwicker posted:there already are right now, mixtral for example is a pretty good one and competitive with gpt, but as to whether that situation will last and how depends on whether certain huge media companies try to shut it down. I don't think Disney can dictate what France or China or Russia is going to do in the AI space, nor do I think they even can keep up let alone understand what they are trying to prevent, ESPECIALLY since AI runs directly into their money saving, union-smashing capitalist hellscape mode of operation, they see dollar signs in using AI. NVIDIA wants to sell hardware and they honestly don't care to who, which also runs counter to somehow Disney (or similar traditional media conglomerate) getting in their way. The techniques to AI at all are known now, all you need is some hardware which incedently is plentiful and cheap and already in existence. ChatGPT has an edge and they know it isn't going to last forever which is why the are trying to thread the needle on how they somehow know how to do AI safely and really nobody else should be allowed to by the way they work with the Federal Government and Military now. All the information you train these systems on and all of their output is analog, images, audio, video, text. You can't DRM any of it, no matter how hard you try. "Is this AI?" is already a dead question because it's saturated everything already and most people don't even notice.
|
|
#
?
Jan 30, 2024 19:20
|
|
|
big drop of random junk I've been prompting for Lobstermoose    In a way, wouldn't nuclear war be exciting?        Series of Retrofuturistic Hippie Dustbowl with Subtle Anachronistic Technology One-offs:             misc. art             Retrofuturistic spaceship interiors     A series of Black Mage characters:            Attempting cel shaded black and white outdoor shots:    ALIENS / Iridescent Environments and Skellymans                                  Helmet Lady  
|
|
#
?
Jan 30, 2024 19:40
|
|
|
Soulhunter posted:
I love you, helmet lady
|
|
#
?
Jan 30, 2024 20:36
|
|
|
             
|
|
#
?
Jan 30, 2024 22:38
|
|
|
Soulhunter posted:
    
|
|
#
?
Jan 30, 2024 23:06
|
|
|
    
|
|
#
?
Jan 30, 2024 23:26
|
|
|
pixaal posted:the dalle 2 version of Bing had a really cool concept of "made with sharpies and highlighters" Dall-E 3 Bing, on the other hand, is too good at fantasy art, and even if I tried getting it to be a bit more impressionistic and/or rough it couldn't manage it. That particular look seemed to be intrinsic to Dall-E 2, and I kind of miss it.
|
|
#
?
Jan 30, 2024 23:42
|
|
|
                      
|
|
#
?
Jan 31, 2024 00:49
|
|
|
doom guy is jesus, that checks out
|
|
#
?
Jan 31, 2024 00:56
|
|
|
“And out of his mouth goeth a sharp sword, that with it he should smite the nations: and he shall rule them with a rod of iron: and he treadeth the winepress of the fierceness and wrath of Almighty God.”
|
|
#
?
Jan 31, 2024 02:47
|
|
|
I don't think anyone can engineer a button that makes "into the spiderverse" pop out in our lifetime.
|
|
#
?
Jan 31, 2024 04:59
|
|
|
I can't really see AI, at best, being able to generate anything better than the blandest Hallmark movie nonsense. I still think writers with unique voices and especially who are funny will be able to rise above in the new order. Like everything else AI is going to touch, it's the people who were churning out low effort slop just for the paycheck who are going to be first to go
|
|
#
?
Jan 31, 2024 06:35
|
|
|
I'm incredibly annoyed tonight as someone accused me of using AI to write a short nonsense satire piece, leading to community dog-piling and a ban from said community on false accusations. Even with previous writing samples from before widespread LLM availability, they were convinced they found a witch, so I had to be tossed on the pyre, as it were. This left me pondering the question of what, if anything, can be done to counter someone baselessly claiming your writing is AI generated? If it's not a writing piece you'd need to draft out or spend more than five minutes on, then short of recording yourself typing the material out, is there anything one can do to prove their work is their own writing? As far as I'm aware, "AI detector" software is utterly useless (not that the accuser even bothered to go so far as that due diligence). Tossing my writing into a couple free apps (ZeroGPT, etc.) gives no indication of AI usage, which is a small mercy at least and clears me of being paranoid that my writing is too mechanical. I feel like AI models are a useful tool, but I strictly do not utilize them when doing creative writing, so it's discouraging to be accused of using them for this purpose. Is it even possible to break into creative writing for profit right now with the spectre of being accused of using AI tools hanging over your head? In return for suffering through my rambling, here are four nerds getting mad at AI art on their tiny computers.    
|
|
#
?
Jan 31, 2024 07:53
|
|
|
There's no way to reliably watermark text and even if you have end-to-end signing of documents you could simply read ChatGPT off your phone and type it into Microsoft Word. The AI detectors are a grift. You probably won't be able to convince them, even if you came up with a reasonable argument. Everyone is too angry.
|
|
#
?
Jan 31, 2024 09:16
|
|
|
Soulhunter posted:Is it even possible to break into creative writing for profit right now with the spectre of being accused of using AI tools hanging over your head?  Any actual real professional has already adopted AI and has been using it since it's become available. In the professional space nobody gives a gently caress about whatever tools you use. This whole anti-AI thing is the realm of online internet people that somehow think their struggles with automation are the first time this has come up, completely ignoring every single other time this exact same thing has happened, be it photography or drawing tables or photoshop or word processors or whatever else. The anti-AI thing is entirely an online social phenomenon. People still watch movies made by Hollywood no matter who has been crushed, molested, exploited or straight up murdered to create whatever film. People still buy clothes made by actual real anti semites. People go to concerts performed by actual plagiarists. The public does not care, never has and never will, about AI. There is nothing ethical about anything to do with profit, AI doesn't even make the list.
|
|
#
?
Jan 31, 2024 09:54
|
|
|
KakerMix posted:The anti-AI thing is entirely an online social phenomenon. I wish this were the case, I've been working on adoption of AI professionally as part of an R&D rollout to staff for a large organization, and about ~60% of our software developers are also infected with this "AI BAD" mindset. The other chunk of staff (technical and non-technical) seems to think current tools are magic and come to use them with the expectation that it will do all their work for them perfectly without human intervention. You can probably separate the two groups cleanly based on whether they interpret "AI is going to take my job" as a good thing or a bad thing. (It's almost always a good thing in my opinion, less brain-killing menial labor tasks is objectively good.) No matter how we try to show staff that these current generative models are just tools to use and do not replace human effort in a reliable way (yet), there is a significant offline contingent that actually does rage about AI, which is leading to some tech adoption issues internally for us. In the academic world, there's a lot of accusations of AI use in writing similar to the one I dealt with yesterday being tossed around, which also creeps beyond the online world. Since there's no way to detect AI use in writing currently, it's just disappointing and frustrating that the best we can do is to tell accusers to gently caress off or attempt to provide proof of a drafting process that one may not have used. The issue compounds when the ones making accusations come from a position of trusted authority over a young writer - tenured professors, community leaders, editors, and so on - as any governing admin over the authority figure will trust the authority figure over the junior writer based on experience regardless of their ignorance to what AI can actually do at the moment, resulting in negative grades, rejected content, etc. There are tangible and sometimes significant professional consequences to baseless accusations, which makes them all the more insidious and frustrating to deal with, especially with limited means to prove your innocence beyond a reasonable doubt. Consumers will consume, so I think you're correct that the public ultimately won't give a gently caress about AI use in entertainment, with the caveat being that that's only truly the case when the public can't tell AI was used. Slap a "made with AI" logo on a product and I think there's still significant negative sentiment that impacts the bottom line. Unfortunately, the problem comes in where anyone can essentially slap the "this was made with AI!!" accusation on your work, and you have limited or no functional means of proving otherwise. Essentially, you just have to have your reputation eat poo poo and assert the "nuh- uh" defense, I guess, hoping that public opinion doesn't side with the accuser. SCheeseman posted:Everyone is too angry. Ultimately this is the case, and I can't wait for AI adoption to become more widespread so people understand generative AI for writing in all its current commercially available forms is just a tool, and isn't generally "creative" in its own right, much less in a way a human writer can be. Gen AI for audio / image generation seems to be much more capable of humanlike content generation at the moment, but still isn't a magic bullet. Human input is still necessary, whether for tweaking code, punching up a script, splicing audio, or adjusting images in photoshop. AI tools just get you ~75% of the way to a finished product most of the time unless you're extremely persistent or lucky with your prompting. 
Soulhunter fucked around with this message at 14:54 on Jan 31, 2024 |
|
#
?
Jan 31, 2024 14:43
|
|
|
For LLMs in particular, you only get anything good if you put something good in, otherwise you'll only ever get the most generic, bland poo poo imaginable. This goes for most generative AI tbh, and I rarely have much interest in their raw output as much as how they can help assist in existing workflows. For AI generated music I've been avoiding LLMs for lyric writing because of this, as well as exploring ways to build upon what is generated in ways that allow a greater exercise in control. At the moment I'm making food icons for a GTA RP restaurant for a friend's wife, it's also extraordinarily useful for things like that. Even then I'm doing work; lots of adjustments made to each icon to keep them consistent.
|
|
#
?
Jan 31, 2024 14:55
|
|
|
I don't generally have a goal when making AI art I feel more like I'm exploring and seeing what is there. Invoking artist names and passing them off or likenesses of real people it's a line and stuff like that should probably be marked but what are you gonna do law wont catch up, there's a lot of stigma around it. At the same time have you seen how much money senators have invested in companies involved in AI like Nvidia? The laws are going to likely side with AI in a way that doesn't ban it.
|
|
#
?
Jan 31, 2024 14:58
|
|
|
pixaal posted:I don't generally have a goal when making AI art I feel more like I'm exploring and seeing what is there. Invoking artist names and passing them off or likenesses of real people it's a line and stuff like that should probably be marked but what are you gonna do law wont catch up, there's a lot of stigma around it. I've never felt a sense of ownership when doing this. With the AI music stuff, I kinda almost maybe sorta do, mostly because of the lyrics? I don't know where that line is drawn, though.
|
|
#
?
Jan 31, 2024 15:00
|
|
|
Soulhunter posted:is there anything one can do to prove their work is their own writing? Not sure how academics is going to handle it. 20 years ago I had a few classes that made us hand write our essays live, we had to prepare for 5 topics but would only be asked to write on 3 of them at random, ensuring you prepared and mostly knew what you were talking about. I think that's a good system. For researchers, I'm assuming it's going to become an unholy mess of misinformation. Too many people out there trying to fake research to get jobs or grants or claim that their country discovered some bullshit...all of which was already going on but now it's going to be stupid easy to do the writing. In terms of your forum drama, you'll never win that fight so just move on bro.
|
|
#
?
Jan 31, 2024 20:02
|
|
|
|
| # ? May 31, 2024 14:11 |
|
|
KakerMix posted:
i know several people who write for a living and who are definitely not using AI. this is not at all universally true, not even close.
|
|
#
?
Jan 31, 2024 20:52
|
|