[HN Gopher] No elephants: Breakthroughs in image generation
___________________________________________________________________
No elephants: Breakthroughs in image generation
Author : Kerrick
Score : 337 points
Date : 2025-04-05 03:51 UTC (3 days ago)
(HTM) web link (www.oneusefulthing.org)
(TXT) w3m dump (www.oneusefulthing.org)
| roenxi wrote:
| That proper "no elephants" first image is hilarious. Another key
| point here is the generative AI's meme game is getting rather
| strong.
|
| Which isn't a small thing, humour is an advanced soft skill.
| YurgenJurgensen wrote:
| Having thousands of copies of that image in your training set
| isn't a skill at all.
| hnbad wrote:
| AI's "meme game is going strong" by the same metric I would
| use to try to argue that Elon Musk's is.
|
| I wouldn't call it a good metric, though.
| loudmax wrote:
| The "Game Design Otter" action figure seems to come with a pair
| of flashlights. I bet that's a residue from the previous prompt
| about illuminating the tablet with a flashlight.
| aenvoker wrote:
| Having used lots of different image generators, so far the only
| one with a sense of humor has been Dall-e3.
|
| https://www.reddit.com/r/dalle2/s/khb5XuNFdl
|
| There's probably some sort of connection to ChatGPT in there.
| But, I don't know enough about how it works.
| x187463 wrote:
| This is a before/after moment for image generation. A simple
| example is the background images on a ton of (mediocre) music
| youtube channels. They almost all use AI generated images that
| are full of nonsense the closer you look. Jazz channels will
| feature coffee shops with garbled text on the menu and furniture
| blending together. I bet all of that disappears over the next few
| months.
|
| On another note, and perhaps others are feeling similarly, but I
| am finding myself surprised at how little use I have for this
| stuff, LLMs included. If, ten years ago, you told me I would have
| access to tools like this, I'm sure I would have responded with a
| never ending stream of ideas and excitement. But now that they're
| here, I just sort of poke at it for a minute and carry on with my
| day.
|
| Maybe it's the unreliability on all fronts, I don't know. I ask a
| lot of programming questions and appreciate _some_ of the
| autocomplete in vscode, but I know I 'm not anywhere close to
| taking full advantage of what these systems can do.
| Retr0id wrote:
| I've never used a stock photo site before, so I suppose it's no
| surprise I have no real use for "generate any image on demand".
| YurgenJurgensen wrote:
| Their main application appears to be taking blog posts and
| internal memos and making them three times longer and use ten
| times the bandwidth to convey no more information. So exactly
| the application AI is 'good' at.
| wongarsu wrote:
| If anything stock image websites are even worse at this
| then AI. With AI you come up with an image idea, then try
| to make the AI produce something close to it. With stock
| images you come up with an image idea, then hope some
| photographer had a similar idea and uploaded it to a stock
| website.
| esperent wrote:
| I've used stock photo sites occasionally but I use vector art
| and icon sites multiple times a week. Even today, I used an
| few different sites while designing some stuff on Canva.
|
| The reason I don't use AI is because it gives me far less
| reliable and impossible to specify results than just
| searching through the limited lists of human made art.
|
| Today, for undisclosed reasons, I needed vector art of
| peanuts. I found imperfect but usable human made art within
| seconds from a search engine. I then spent around 15 - 25
| minutes trying to get something closer to my vision using
| ChatGPT, and using the imperfect art I'd found as a style
| guide. I got lots of "huh that's cool what AI can do" but
| nothing _useful_. Nothing closer to my vision than what I
| started with.
|
| By coincidence it's the first time I'vr tried making art with
| AI in about a year, but back then I bought a Midjourney
| account and spent a month making loads of art, then installed
| SD on my laptop and spent another couple of weeks playing
| around with that. So it's not like I'm lacking experience.
| What I've found so far is that AI art generators are great
| for generating articles like this one. And they do make some
| genuinely cool pictures, it blows my mind that computers can
| do this now.
|
| It's just when I sit down with a real world task that has
| specific, concrete requirements... I find them useless.
| card_zero wrote:
| I was wondering yesterday how AI is coming along for tweening
| animation frames. I just did a quick search and apparently last
| year the state of the art was garbage:
|
| https://yosefk.com/blog/the-state-of-ai-for-hand-drawn-anima...
|
| Maybe this multimodal thing can fix that?
| GaggiX wrote:
| That blog post is a year old.
|
| There has been a lot of progress since then:
| https://doubiiu.github.io/projects/ToonCrafter/
| kridsdale3 wrote:
| Very impressive. This is going to result in an explosion of
| content creation by pro studios, just as CG with cel-
| shading renderers did. I greatly prefer the hand-drawn + AI
| tweened look to the current low-budget CG 3D models look.
| __loam wrote:
| Most of the professionals in this industry actively
| despise this technology.
| satvikpendem wrote:
| That's true, but it will likely be newer studios with
| younger professionals who are going to be using it, much
| as Miyazaki doesn't like CGI either yet it's widely used
| now in anime. The young drive the advances while the
| older eschew them, that's generally how human progress
| has been.
| GaggiX wrote:
| I don't even think it will be newer studios and that's
| it, probably well known studios will adopt it after a
| production ready solution is presented and battle tested,
| I doubt many will complain about not making in between
| frames. Probably newer studios with a smaller budget will
| just test it first.
| fc417fc802 wrote:
| I doubt they will have a choice given the price
| difference.
|
| This is arguably a good thing because if production cost
| drops it should mean either higher quality or more
| content.
| GaggiX wrote:
| Yeah it will be much better than the low budget 3D models
| in anime, hopefully there will be a production ready
| product that works at a high enough resolution and
| studios will probably adopt it instead of using cheap
| labor.
| loudmax wrote:
| That is a very interesting point about how little use of AI
| most of us making day to day, despite the potential utility
| that seems to be lurking. I think it just takes time for people
| and economies to adapt to new technology.
|
| Even if technological progress on AI were to stop today, and
| the best models that exist in 2030 are the same models we have
| now, there would still be years of social and economic change
| as people and companies figure out how to make use of novel
| technology.
| milanove wrote:
| Unless I'm doing something simple like writing out some basic
| shell script or python program, it's often easier to just do
| something myself than take the time to explain what I want to
| an LLM. There's something to be said about taking the time to
| formulate your plan in clear steps ahead of time, but for
| many problems it just doesn't feel like it's worth the time
| to write it all out.
| danielbln wrote:
| I find that if a problem doesn't require planning it's
| probably simple enough that the LLM can handle it with
| little input. if it does require planning, I might as well
| dump it into an LLM as another evaluator and then to drive
| the implementation.
| skybrian wrote:
| Image generation is still very slow. If it generated many
| images instantly like Google's image search, it would be a lot
| more fun to use, and we would learn to use it more effectively
| with practice.
| nyarlathotep_ wrote:
| > They almost all use AI generated images that are full of
| nonsense the closer you look. Jazz channels will feature coffee
| shops with garbled text on the menu and furniture blending
| together.
|
| Noticed tht.
|
| Maybe it's my algorithm but YouTube is seemingly filled with
| these videos now.
| rasz wrote:
| Youtube Studio is has build in AI thumbnail functionality.
| Google actively encourages use of AI to clickbait and to
| generate automatic AI replies to comments ala onlyfaps giving
| your viewers that feeling of interaction without reading
| their comments.
| UncleEntity wrote:
| They insist on feeding me AI generated videos about "HOA
| Karens" for some odd reason.
|
| True, I do enjoy watching the LawTubers and sometimes they
| talk about HOAs but that is a far stretch from someone taking
| a reddit post and laundering it through the robots.
| _DeadFred_ wrote:
| All my music cover images are AI generated. At the same time
| I refuse to listen to AI music. We're all going to sink alone
| on this one.
|
| What's frustrating me is if I tell the Youtube algo 'don't
| recommend' to AI music video channels it stops giving me any
| music video channels. That's not what I want, I just don't
| want the AI. They need to seperate the two. But of course
| they need to not do that with AI cover images because
| otherwise it would harm me. :)
| satvikpendem wrote:
| Probably is your algorithm as mine is pretty good in not
| showing me those low effort channels. Check out extensions
| like PocketTube, SponsorBlock, and DeArrow to manage your
| YouTube feeds better.
| Gasp0de wrote:
| I love using LLMs to generate pictures. I'd call myself rather
| creative, but absolutely useless in any artistic craft. Now, I
| can just describe any image I can imagine and get 90% accurate
| results, which is good enough for the presentations I hold,
| online pet projects (created a squirrel-themed online math-
| learning game for which I previously would have needed a
| designer to create squirrel highschool themed imagery) and
| memes. For many, many websites this is going to be good enough.
| candiddevmike wrote:
| My problem with finding enjoyment in this is the same problem
| I have when using cheat codes in games: the doing part is the
| fun part, getting to the end or just permutations of the end
| gets really boring.
| williamcotton wrote:
| Trying to draw a squirrel when you have no artistic talents
| or experience is not the fun part.
|
| I've produced my own music recordings in the past and I've
| hired musicians to play the instruments that I cannot.
| Having exasperated recording engineers watch my 5,000th
| take on a drum fill that I absolutely cannot play is not
| the fun part. Sitting behind the glass and watching my
| vision come to life from a really good drummer is
| absolutely the fun part.
| munksbeer wrote:
| >I love using LLMs to generate pictures. I'd call myself
| rather creative, but absolutely useless in any artistic
| craft. Now, I can just describe any image I can imagine and
| get 90% accurate results
|
| May I ask what you use? I'm not yet even a paid subscriber to
| any of the models, because my company offer a corporate
| internal subscription chatbot and code integration that works
| well enough for what I've been doing so far but has no image
| generation.
|
| I have tried image generation on the free tier but run out of
| free use before I get anyway pleasing.
|
| What do you pay for?
| wincy wrote:
| I was generating pictures to use for a little game I made
| with my six and ten year old kids. They were so excited to
| see us go from idea to execution so quickly, they were
| laughing and we had a ton of fun. The only thing that
| disappointed me was I got throttled. We'd need to pay for
| API image gen to get it even faster.
|
| I made a logo for an internal product that wouldn't have
| had a logo otherwise at our company. I also make a lot of
| shitpost memes to my friends to trash talk in the long
| running turn based war game we've all been playing, like
| "make a cartoony image of a dog man and a Greek giant
| beating up a devil" and the picture it gave was just
| hilarious and perfect, like an old timey Popeye cartoon.
|
| Two years ago I was spending three hours using local models
| like Stable Diffusion to get exactly what I wanted. I had
| to inpaint and generate 100 variations which would have
| been insanely expensive if I wasn't powering it with my own
| hardware.
|
| Now I get something good in minutes, it's crazy really.
| genewitch wrote:
| Feels like you didn't answer the question. i know you
| weren't who was asked, but still.
| munksbeer wrote:
| Thanks. Which service do you use please? I'm wanting to
| try a paid service, just want to know which ones people
| recommend.
| HelloMcFly wrote:
| I've used Midjourney and chatGPT. Midjourney is better
| for rapid iteration, cycling through options faster, and
| to a large extent getting "weirder". It's easier to tweak
| using parameters.
|
| ChatGPT is far, far superior (especially now) when you
| want something more specific that you've already
| imagined. But it's slower, and unlike Midjourney you
| don't get four versions to choose to build and iterate
| on, you get a single image that takes longer to load.
| __loam wrote:
| If you use this technology, you're actively harming creative
| labor.
| Empact wrote:
| Creative labor is not entitled to the work parent comment
| is describing. We employ labor because it is beneficial to
| us, not merely because it exists as an option. Creative
| labor's responsibility is to adapt to a changing world and
| find roles where their labor is not simply produced /
| exceeded by a computer system.
|
| Practically speaking, the work described would most likely
| never have been done, rather than been done by an artist if
| that were the only option - it's uncommon to employ artists
| to help with incidental tasks relative to side projects,
| etc.
| genewitch wrote:
| Whatever. I wrote and co-wrote ten albums and my total take
| was $3.
|
| The market is saturated and the way it works means ten get
| rich for every million artists. I feel as though this has
| been pretty constant throughout history.
|
| Of course there's a lot of talent out there, "wasted", but
| I think that's always been the case. How many William
| Shakesmans did we lose with all the war, famine, disease?
|
| I actually decided I'd probably never write music again
| after 1-shot making a song about the south Korea coup
| attempt several months ago. I had the song done before the
| news really even hit the US. Why would I destroy my own
| hearing writing music anymore when I can prompt an AI to do
| it for me, with the same net result - no one cares.
|
| here's the 3-shot remix, the triangle cracks me up so much
| that i had to upload it
| https://soundcloud.com/djoutcold/coup-detat-symphony-remix
|
| the "original" "1-shot" is on my soundcloud page as well.
| https://soundcloud.com/djoutcold/i-aint-even-writing-
| music-a...
|
| it's in lojban. That's why you can't understand it. Yes.
| Lojban. Brings a tear to my eye every time i hear it. fkin
| AI
|
| [0] more my style - hold music for our PBX
| https://soundcloud.com/djoutcold/bew-hold-music also all my
| stuff is CC licensed, mostly CC0 at this point.
| Garlef wrote:
| > How many William Shakesmans did we lose with all the
| war, famine, disease?
|
| (Just a small comment out of context of the remaining
| discussion:)
|
| Maybe not many? It could be that "cultural attention" is
| limited and there's not much space at the top anyways. In
| other words: It might be that there's always a few famous
| artists that get remembered and the rest is forgotten.
| Same as winning the world cup: There's always a team that
| wins and it says nothing about the quality in a universal
| way. At best it says something about quality relative to
| the competition.
|
| (Not sure I'd fully get behind the argument i composed
| here. But I found it interesting.)
| ldoughty wrote:
| Can you elaborate how there's no possible way to use this
| technology without actively harming artists?
|
| If a classroom of 14 year olds are making a game in their
| computer science class, and they use AI to make placeholder
| images... Was a real artist harmed?
|
| The teacher certainly cant afford to pay artists to provide
| content for all the students games, and most students can't
| afford to hire an artist either.. they perhaps can't even
| legally do it, if the artist requires a contract... they
| are underage in most countries to sign a contract.
|
| This technology gives the kids a lot more freedom than a
| pre-packaged asset library, and can encourage more
| engagement with the course content, leading to more people
| interested in creative-employing pursuits.
|
| So, I think this technology can create a new generation of
| creative individuals, and statements about the blanket harm
| need to be qualified.
| __loam wrote:
| You cannot ethically use a tool that was produced by
| appropriating the labor of millions of people without
| consent. You are a bad person if you use it.
|
| This is a nice strawman you've put up, but students and
| teachers are not entitled to nice assets. As you
| mentioned, there's plenty of free and open source content
| on the internet, like open game art or kenney, that can
| be used ethically, or the 14 year olds can make their own
| assets and understand how much artistic labor is involved
| in making games. Having them use the theft engine would
| stifle their creativity, not enhance it.
|
| I kind of think you're also showing an implicit bias in
| your comment that bothers me a lot in the tech industry.
| If using AI is okay for the creative labor, why shouldn't
| the students also use it for the programming too? Why do
| anything at all?
|
| It feels too me like a lot of people in tech feel that
| their labor is implicitly more valuable than artistic
| labor. You think it's fine to use AI for the art because
| you think that isn't the important part. The importance
| for you is in the programming only.
|
| Games are massively technical, multi disciplinary works
| of art that require a wider variety of skills than just
| programming, yet programmers always assume their
| contributions are the most valuable, and the most worthy
| of being taught and passed down to the next generation.
| The aesthetic value of the art is just an afterthought.
|
| It's kind of why I'm always so disgusted by this
| conversation. These image generators wouldn't exist
| without artistic labor, and that labor is arguably the
| more important contribution compared to the technical
| labor that produced the training algorithm, but rather
| than revere that contribution, the members of this
| community seem almost annoyed that artists are upset with
| them. It's just so ungrateful. And then you invent these
| scenarios to justify that theft like this, telling us an
| artist wouldn't get paid anyway so why are you so mad?
|
| I'm mad because you're telling me this is going to lead
| to a new generation of creatives while you smother the
| economies and information ecosystems that support the
| training of creative skills, and you're too ignorant,
| deluded, or malicious to stop.
| fc417fc802 wrote:
| The straw man is yours. No claim of entitlement was made.
| A scenario was provided that appears to refute your
| unconditional assertion that using this technology
| actively harms creative labor.
|
| You've presented all sorts of wild assumptions and
| generalizations about the people who don't share your
| vehement opposition to the use of this technology. I
| don't think it's the person you're responding to with the
| implicit bias.
|
| You've conflated theft with piracy (all too common) and
| assumed a priori that training a model on publicly
| available data constitutes such. Do you really expect
| people to blindly adopt your ideological views if you
| just state them forcefully enough?
|
| > If using AI is okay for the creative labor, why
| shouldn't the students also use it for the programming
| too?
|
| They absolutely should! At least provided it does the job
| well enough.
|
| Unless they are taking a class whose point is to learn to
| program yourself (ie the game is just a means to an end).
| Similar to how you might be forbidden to use certain
| advanced calculator features in a math class. If you
| enroll in an art class and then just prompt GPT that
| likely defeats the purpose.
| immibis wrote:
| Only if there was ever any chance you would have hired some
| for that task.
| Animats wrote:
| Creative labor is going the way of manual labor.
| becquerel wrote:
| All labor is bad.
| satvikpendem wrote:
| Interesting philosophy, what is this predicated on? Do
| you mean that people should not have to work for a
| living, ie labor versus play?
| fc417fc802 wrote:
| Is that not self evident? When people engage in labor for
| the task itself (as opposed to a heavily abstracted
| version of not wanting to starve) we generally refer to
| that as a hobby.
|
| So stating that people shouldn't need to worry about
| starving (metaphorically or otherwise) would be roughly
| equivalent.
| satvikpendem wrote:
| It is not always evident especially when it comes to a
| site all about capital accumulation like HN, more due to
| its association with a venture capital firm.
| fc417fc802 wrote:
| Statistically, Jane Street probably employs at least a
| few communists.
| ahmeneeroe-v2 wrote:
| Agreed. I am firmly on the side of Capital.
| nitwit005 wrote:
| > For many, many websites this is going to be good enough.
|
| It was largely a solved problem though. Companies did not
| seem to have an issue with using stock photos. My current
| company's website is full of them.
|
| For business use cases, those galleries were already so
| extensive before AI image generation, that what you wanted
| was almost always there. They seemingly looked at people's
| search queries, and added images to match previously failed
| queries. Even things you wouldn't think would have a photo
| like "man in business suit jump kicking a guy while
| screaming", have plenty of results.
| dylan604 wrote:
| Really? What stock service would have a selection of
| squirrels in a high school setting doing various math or
| other subject related things?
|
| To think any/all combined stock services would be the end
| all is just unrealistic. Sure, someone one might have
| settled on something just because they got tired of
| scrolling (much like streaming video services), that does
| not mean they are happy with their selection. Just happy to
| be done.
|
| Now, with generativeAI, they can have squirrels doing
| anything in any setting they can describe. If they don't
| like it, they can just tweak the description until they are
| happy. It's an obvious plus for them.
|
| I never drank the kool-aid to be all gung-ho on this
| boom/fad, but I'm not going to be so obstinate that I
| refuse to accept some people find it quite useful. May
| someone make all the squirrel attending highschool
| generative art they want, but you can't tell me some stock
| place is good 'nuff for everything.
| Der_Einzige wrote:
| That feeling of not knowing what to do with it is an example of
| humans being stupid. We are all victims of being "Johnny" in
| this paper:
|
| https://dl.acm.org/doi/full/10.1145/3544548.3581388
| genewitch wrote:
| I have the Gemini app on my phone and you can interact with it
| with voice only and I was like oh this is really cool I can use
| it while I'm driving instead of listening to music.
|
| I can never think of anything to talk to an AI about. I run LM
| local, as well
| JFingleton wrote:
| Have it interview you (as-in a job interview) on your
| specialisation. Works your interviewer skills.
|
| Ask it to teach you a language.
|
| DnD works really well (the LLM being the game-master).
| mycall wrote:
| > I am finding myself surprised at how little use I have for
| this stuff
|
| I think this will change as more practical use cases begin to
| emerge as this is all brand new. For example, the photos you
| take with your smartphone can tell a story or be annotated so
| you can see things in the photos you didn't think about but
| your profile thinks you might. Things will get more
| sophisticated soon.
| dontlaugh wrote:
| The unreliability and inability to debug are why I think these
| tools are actually a liability for any serious work.
| Retr0id wrote:
| I had a reasonable intuition for how the "old" method works, but
| I still don't grok this new approach.
|
| "in multimodal image generation, images are created in the same
| way that LLMs create text, a token at a time"
|
| Is there some way to visualise these "image tokens", in the same
| way I can view tokenized text?
| fxtentacle wrote:
| Imagine you cut the image into 32x32 pixel blocks. And then for
| each block, you can chose 1 out of 128,000 variations. And then
| a post-processing step smoothes out the borders between blocks
| and adjusts small details. That's basically how a transformer
| image generation model works.
|
| As such, the process is remarkably similar to old fixed-font
| ASCII art. It's just that modern AIs have a larger alphabet
| and, thus, more character shapes to choose from.
| rwmj wrote:
| I don't get how this would produce consistent images. In the
| article, the text could be on a grid, but the window and
| doorway and sofa don't seem to be grid-aligned. (Or maybe the
| text is overlaid?)
| danielbln wrote:
| The model looks ahead, just like LLMs look ahead. An LLM
| outputs token by token but can still output a fully
| coherent and consistent story for example. This new crop of
| auto-regressive image models does the same.
| WhiteNoiz3 wrote:
| I haven't see any details on how OpenAI's model works, but the
| tokens it generates aren't directly translated into pixels -
| those tokens are probably fed into a diffusion process which
| generates the actual image.. The tokens are the latent space or
| conditioning for the actual image generation process.
| bonoboTP wrote:
| > I haven't see any details on how OpenAI's model works
|
| Exactly. People just confidently make things up. There are
| many possible ways, and without details, "native generation"
| is just a marketing buzzword without clear definition. It's a
| proprietary system, there is no code release, there is no
| publication. We simply don't know how exactly it's done.
| og_kalu wrote:
| Open AI have both said it's native image generation _and_
| autoregressive. It has the signs of it too.
|
| It's probably an implementation of VAR
| (https://arxiv.org/abs/2404.02905) - autoregressive image
| generation with a small twist. Rather than predict every
| token at the target resolution directly, start with
| predicting it at a small resolution, cranking it higher and
| higher until the desired resolution.
| card_zero wrote:
| Looking at the example where the coffee table is swapped, I
| notice every time the image is reprocessed it mutates, based on
| the previous iteration, and objects become more bizarre each
| time, like chinese whispers.
|
| * The weird-ass basket decoration on the table originally has
| some big chain links (maybe anchor chain, to keep the theme with
| the beach painting). By the third version, they're leathery and
| are merging with the basket.
|
| * The candelabra light on the wall, with branch decorations,
| turns into a sort of skinny minimalist gold stag head, and then
| just a branch.
|
| * The small table in the background gradually loses one of its
| three legs, and ends up defying gravity.
|
| * The freaky green lamps in the window become at first more
| regular, then turn into topiary.
|
| * Making the carpet less faded turns up the saturation on
| everything else, too, including the wood the table is made from.
| bla3 wrote:
| The pictures on the wall change too.
| card_zero wrote:
| Yes, first a still life and something impressionist, then a
| blob and a blob, then a smear and a smear. And what about the
| reflections and transparency of the glass table top? It gets
| very indistinct. Keep working at the same image and it looks
| like you'll end up with some Deep Dream weirdness.
|
| I think the fireplace might be turning into some tiny stairs
| leading down. :)
| YurgenJurgensen wrote:
| Only sailors know how to leave.
| rob74 wrote:
| Actually, _almost everything_ changes slightly - the number,
| shape and pattern of the chairs, the number and pattern of
| the pillows, the pattern of the curtains, the scene outside
| the window, the wooden part of the table, the pattern of the
| carpet... The blue couch stays largely the same, it just
| loses some detail...
| M4v3R wrote:
| Yeah, this is in my opinion the biggest limitation of the
| current gen GPT 4o image generation: it is incapable of editing
| only parts of an image. I assume what it does every time is
| tokenizing the source image, then transforming it according to
| the prompt and then giving you the final result. For some use
| cases that's fine but if you really just want a small edit
| while keeping the rest of the image intact you're out of luck.
| danielbln wrote:
| It just means that you comp it together manually. That's
| still much better than having to set up some inpainting
| pipeline or whatever.
| echelon wrote:
| 100%. Multimodal images surpass ComfyUI and inpainting (for
| now). It's a step function improvement in image generation.
|
| I'm hoping we see an open weights or open source model with
| these capabilities soon, because good tools need open
| models.
|
| As has happened in the past, once an open implementation of
| DallE or whatever comes out, the open source community
| pushes the capabilities much further by writing lots of
| training, extensions, and pipelines. The results look
| significantly better than closed SaaS models.
| wavemode wrote:
| Is manually comping actually going to be easier (let alone,
| give better results) than inpainting? I can imagine it
| working in simple cases, but for anything involving 3D
| geometry you'll likely run into issues of things not quite
| lining up between the first and second image.
| atommclain wrote:
| I thought the selection tool allows you to limit the area of
| the image that a revision will make changes to, but I tested
| it and I still see changes outside of the selected area which
| is good to know.
|
| As an example the tape spindles, among other changes, are
| different: https://chatgpt.com/share/67f53965-9480-800a-a166-
| a6c1faa87c...
|
| https://help.openai.com/en/articles/9055440-editing-your-
| ima...
| qingcharles wrote:
| Yeah, I'm not sure what the selection brush actually
| _does_. Is it just a hint to the LLM?
| og_kalu wrote:
| It's kind of clear that for every request, it generates a new
| image entirely. Some people are speculating a diffusion decoder
| but i think it's more likely an implementation of VAR -
| https://arxiv.org/abs/2404.02905.
|
| So rather than predicting each patch at the target resolution
| right away, it starts with the image (as patches) at a very
| small resolution and increasingly scales up. I guess that could
| make it hard for the model to learn to just copy and paste
| image tokens for editing like it might for text.
| flkiwi wrote:
| BUT it's doing a stunningly better job replicating previous
| scenes than it did before. I asked it just now for a selfie
| of two biker buddies on a Nevada highway, but one is a quokka
| and one is a hyrax. It did it. Then I asked for the same
| photo with late afternoon lighting, and it did a pretty
| amazing job of preserving the context where just a few months
| ago it would have had no idea what it had done before.
|
| Also, sweet jesus, after more than a year of hilarious
| frustration, it now knows that a flying squirrel is a real
| animal and not just a tree squirrel with butterfly wings.
| og_kalu wrote:
| I agree. I'm not saying it's a different model generating
| the images. 4o is clearly generating the images itself
| rather than sending a prompt to some other model. I'm
| speculating about the mechanism for generation in the model
| itself.
| flkiwi wrote:
| Oh, no, I wasn't taking issue with what you said, just
| reacting that, yes, it's not editing the same image but
| redrawing from scratch every time, BUT it's doing a much
| better job of that, with some understanding of the
| context of the previous image so that it can tweak it,
| even if it's never bit for bit identical.
| empath75 wrote:
| The vast majority of people wouldn't notice any of that in most
| contexts in which such an image would be used.
| NitpickLawyer wrote:
| > The results are not as good as a professional designer could
| create but are an impressive first prototype.
|
| I like to look at how far we've come since the early days of
| Stable Diffusion. It was fascinating to play with it back then,
| but it quickly became apparent that it was "generic" and not
| suited for "real work" because it lacked consistency, text
| capabilities, fingers! and so on... Looking at these results now,
| I'm amazed at the quality, consistency and ease of use. Gone are
| the days of doing alchemy on words and adding a bunch of "in the
| style of Rutkovsky, golden hour, hd, 4k, pretty please ..." at
| the end of prompts.
| probably_wrong wrote:
| > _Is it okay to reproduce the hard-won style of other artists
| using AI? Who owns the resulting art? Who profits from it? Which
| artists are in the training data for AI, and what is the legal
| and ethical status of using copyrighted work for training? These
| were important questions before multimodal AI, but now developing
| answers to them is increasingly urgent._
|
| I have to disagree with the conclusion. This was an important
| discussion to have two to three years ago, then we had it online,
| and then we more or less agreed that it's unfair for artists to
| have their works sucked up with no recourse.
|
| What the post should say is "we know that this is unfair to
| artists, but the tech companies are making too much money from
| them and we have no way to force them to change".
| shkkmo wrote:
| > This was an important discussion to have two to three years
| ago, then we had it online, and then we more or less agreed
| that it's unfair for artists to have their works sucked up with
| no recourse.
|
| Speak for yourself, there was no consensus online. There are
| plenty of us that think that dramatically expanding the power
| of copyright would be a huge mistake that would primarily
| benefit larger companies and do little to protect or fund small
| artists.
| OtherShrezzing wrote:
| >There are plenty of us that think that dramatically
| expanding the power of copyright would be a huge mistake that
| would primarily benefit larger companies and do little to
| protect or fund small artists.
|
| The status quo also primarily benefits larger companies, and
| does little (exactly nothing, if we're being earnest) to
| protect or fund small artists.
|
| It's reasonable to hold both opinions that: 1) artists aren't
| being compensated, even though their work is being used by
| these tools, and 2) massive expansion of copyright isn't the
| appropriate response to 1).
| eadmund wrote:
| > it's unfair for artists to have their works sucked up
|
| I never thought it was unfair to artists for others to look at
| their work and imitate it. That seems to me to be what artists
| have been doing since the second caveman looked at a hand
| painting on a cave wall and thought, 'huh, that's pretty neat!
| I'd like to try my hand at that!'
| BriggyDwiggs42 wrote:
| Right the difference is that it's a large company looking at
| it then copying it and reselling it without credit, which
| basically everyone would understand as bad without the
| indirection of a model.
|
| Edit: the key words here are "company" and "reselling"
| eadmund wrote:
| But it's not copying and reselling -- it's imitation.
|
| Copying is controlled by copyrights. And imitation isn't
| controlled by anything.
|
| As for a company: a company is just a group of people
| acting together.
| BriggyDwiggs42 wrote:
| #1, it's extremely easy to coerce direct copies out of
| models that artists could be pursued for infringement if
| they drew, but companies reselling said copyrighted
| artwork face no penalty
|
| #2, yes, it's a group of people who came together to
| build an algorithm that learns to extract features
| learned from images made by other people in order to
| generate images somewhere between these images in a high
| dimensional space. They sell these images and give no
| credit or cash to the images being "interpolated"
| between. Notice this doesn't extend to open source, it's
| the commercial aspect that represents theft.
|
| The reality is that laws are meant to be interpreted not
| by their letter but their spirit. The AI can't exist
| without the hard work its trained on, and the outputs
| often resemble the inputs in a manner that approaches
| copying, so selling those outputs without compensation
| for the artists in the training set should be illegal. It
| won't be, but it should.
| RandallBrown wrote:
| I guess I don't understand what you mean when you say
| companies reselling said copyrighted artwork face no
| penalty. Why wouldn't they? If I was to make a copy of a
| Studio Ghibli movie and sell it, I would absolutely face
| a penalty if I was caught.
| BriggyDwiggs42 wrote:
| A common joke is to type in a description of some
| corporate IP and have ChatGPT generate it without ever
| saying it directly. Plenty of people have paid a
| subscription to do that, generate corporate IP that an
| artist could be sued over, but I don't believe OpenAI has
| faced any legal issues if I'm correct, just as an
| example.
| AnthonyMouse wrote:
| > it's extremely easy to coerce direct copies out of
| models that artists could be pursued for infringement if
| they drew, but companies reselling said copyrighted
| artwork face no penalty
|
| The purpose of the model isn't to make exact
| reproductions. It's like saying you can use the internet
| for copyright infringement. You can, but it's the user
| who chooses the use, so is that on AT&T and Microsoft or
| is it on the users doing the infringement?
|
| > They sell these images and give no credit or cash to
| the images being "interpolated" between.
|
| A big part of the problem is that machines aren't
| qualified to be judges.
|
| Suppose the image you request is Gollum but instead of
| the One Ring he wants PewDiePie. Obviously this is using
| a character from the LOTR films by Warner Bros. If you're
| PewDiePie and you want this image to use in an ad for
| your channel, you might be in trouble.
|
| But Warner Bros. got into a scandal for paying YouTubers
| to promote Promote Middle Earth: Shadow of Mordor without
| disclosing the payments. If you're creating the image to
| criticize the company's behavior, it's likely fair use.
|
| The service has no way to tell why you want the image, so
| what is it supposed to do? A law that requires them to
| deny you in the second case is restricting a right of the
| public. But it's the same image.
|
| Meanwhile in the first case you don't really need the
| company generating the image to do anything because
| Warner Bros. could then go after PewDiePie for using the
| character in commercial advertising without permission.
|
| > Notice this doesn't extend to open source, it's the
| commercial aspect that represents theft.
|
| It's also not really clear how this works. For example,
| Stable Diffusion is published. You can run it locally. If
| you buy a GPU from Nvidia or AMD in order to do that, is
| that now commercial use? Is the GPU manufacturer in
| trouble? What if you pay a cloud provider like AWS to use
| one of their GPUs to do it? You can also pay for the
| cloud service from Stability AI, the makers of Stable
| Diffusion. Is it different in that case than the others?
| How?
| __loam wrote:
| You need to copy the work to use it for AI training.
| BriggyDwiggs42 wrote:
| I'd argue that would be fine for non-commercial use; it's
| once the AI outputs are sold that the problem arises.
| SirMaster wrote:
| You don't see a massive difference in the shear number of
| images that the AI can look at and the speed at which it can
| imitate it as a fundamental difference between AI and a human
| copying works or styles?
|
| For a human it took a lot of practice and a lot of time and
| effort. But now it takes practically no time or effort at
| all.
| Workaccount2 wrote:
| Well yeah, but copyright infringement isn't a function of
| how quickly you can view and create works.
|
| Copyright is meant to secure distribution of works you
| create. It's not a tool to stop people from creating art
| because it looks like your art. That has been a thing for
| centuries, we even categorize art by it's style. Imagine
| anime was had to adhere to a copyright interpretation of
| "it's my style!".
| SirMaster wrote:
| Current copyright yes.
|
| But do you not for a second think that the current way
| the laws and rules are set are _because of_ how hard and
| time consuming it was to replicate work?
|
| Just because "that's how it's always been" doesn't mean
| it's acceptable to keep it that way when the means to
| perform the action have so drastically changed.
| fc417fc802 wrote:
| I don't think the rules ever existed for the benefit of
| the individual, but rather the collective. If skilled
| artists couldn't sustain themselves from their work they
| wouldn't exist. Historically there was no alternative.
|
| When a machine can do something there is not generally a
| (collectively beneficial) reason to protect the
| individual that competes with it. Backhoes weren't
| regulated in order to protect ditch diggers.
| eadmund wrote:
| It took a truly colossal amount of human time and effort to
| build AI systems. It takes significant amount of energy to
| run those AI systems.
|
| I don't see any meaningful difference at all between the
| system of a human, a computer and a corpus of images
| producing _new_ images, and the system of a human, a
| paintbrush, an easel, a canvas and a corpus of images
| producing _new_ images. Emphasis on the new -- copying is
| still copying, and still controlled by copyrights.
| SirMaster wrote:
| >It took a truly colossal amount of human time and effort
| to build AI systems. It takes significant amount of
| energy to run those AI systems.
|
| Those people and effort aren't at all tied to the people
| who are making and using the art.
|
| In the past every individual person would have to
| individually study art and some style and practice for
| years of their life to be able to replicate it really
| well. And for each piece of artwork it could take them
| days to make 1 single piece.
|
| I would argue that this is why it wasn't really
| problematic to copy someone's work or style. Because the
| individual time and effort per person to even do that was
| so high.
|
| But now that time and effort for an individual is next to
| nothing.
| munksbeer wrote:
| I do wonder what the outcome would be for a model trained
| only on truly non copyright work, and derivatives from
| there. I'm no AI expert, but from what I understand they
| use some models to generate data with which to train
| further models. I'd be interested in the output, whether
| it would eventually just match what we have now anyway,
| so the copyright question may end up moot. I wonder how
| the argument would shift at that point?
|
| I think in reality, it is probably too late for that,
| because the internet is now polluted with AI generated
| images which would be consumed by any "ethical" model
| anyway.
| fc417fc802 wrote:
| I expect it would require significantly more human labor
| to train (ie no longer fully unsupervised). I imagine
| that this constraint would lead to significant additional
| research to improve the efficiency of the training
| process, and that novel approaches would be developed.
|
| In other words I think it would suck up a lot of money
| over a few years and then we would arrive back pretty
| much where we are now.
| SirMaster wrote:
| You don't see a difference between a person spending
| years learning techniques to create art by hand, and
| spending months or years studying and practicing some
| famous artists style, and then spending days manually
| crafting drawing a single piece of artwork in the style
| and quality of the originals.
|
| The difference between that, and a person just entering a
| prompt to create some drawing in some style.
|
| The model looked at orders of magnitude more examples of
| artwork than a single human could look at and study in a
| lifetime.
|
| To me there is a clear difference here.
|
| I am merely saying that perhaps the rules should change
| due to the drastic change in time and effort required to
| do the work.
| butlike wrote:
| The rules do change, but as a meritocracy as society
| simply decides to move on or not. There will be no cabal
| of artists who define how the rules will change. It will
| be organic. Like moving on from cave paintings to
| impressionism.
| pixl97 wrote:
| Therefore we should give up all heavy equipment and all
| ditches should be dug with a spoon.
|
| Sometimes technology changes and what was nearly
| impossible in the past becomes trivial.
| Taek wrote:
| I don't think there's consensus around that idea. Lots of
| people (myself included) feel that copyright is already vastly
| overreaching, and that AI represents forward progress for the
| proliferation of art in society (its crap today, but digital
| cameras were crap in 2007 and look where they are now).
|
| Its also not clear for example that Studio Ghibli lost by
| having their art style plastered all over the internet. I went
| home and watched a Ghibli film that week, as I'm sure many
| others did as well. Their revenue is probably up quite a bit
| right now?
|
| "How can we monetize art" remains an open question for society,
| but I certainly don't think that AI without restrictions is
| going to lead to fewer people with art jobs.
| hnbad wrote:
| Copyright is a logical consequence of property rights. I'd
| agree that property rights hold back industry and trade but
| if you want to abolish property rights, you first have to
| decommodify the essentials like food, housing, public
| infrastructure and healthcare, because unleashing the market
| when it has control over all of these is going to have some
| very undesirable consequences.
| AnthonyMouse wrote:
| Copyright isn't a property right. Property rights are
| rivalrous. If you own a sandwich and a thousand other
| people want to eat your sandwich, only one person can, so
| property rights exist to define who gets to choose who gets
| to eat the sandwich. Writings and discoveries are non-
| rivalrous. To quote the first head of the US Patent Office:
|
| > He who receives an idea from me, receives instruction
| himself without lessening mine; as he who lights his taper
| at mine, receives light without darkening me.
|
| The term "intellectual property" is an attempt to conflate
| these things, to justify net-destructive money grabs like
| retroactive copyright term extensions, because traditional
| property rights don't expire but copyrights explicitly and
| intentionally do.
| immibis wrote:
| The definition of capitalism is a system in which all
| sorts of things that are not property are artificially
| made into property and given artificial property rights
| which can be traded.
| AnthonyMouse wrote:
| That is the definition of capitalism people use when they
| want to apply the term _capitalism_ to something that
| sucks.
|
| Allowing people to own physical items as business
| inventory or production equipment and compete with each
| other for the customer's dollar is entirely possible
| without the existence of copyright or patents. You would
| then be relying on some combination of open source,
| charitable contributions and patronage, industry joint
| ventures, personal itch scratching, etc. to create
| writings and inventions, but books and the wheel were
| created before patents and copyrights were.
| fc417fc802 wrote:
| > You would then be relying on ...
|
| More likely trade secrets, NDAs, non-competes, and
| increasingly invasive DRM. In addition to the direct
| financial incentive, part of the logic behind IP law is
| to foster a more open market because that should be to
| the benefit of society at large in multiple ways.
|
| Patents, for example, ensure that at least some minimal
| description of the process gets published for others to
| take inspiration from.
| AnthonyMouse wrote:
| > More likely trade secrets, NDAs, non-competes, and
| increasingly invasive DRM.
|
| These are all also creatures of the law. If there was no
| copyright there would be no Digital Millennium Copyright
| Act. In many cases they _can 't_ work, e.g. because of
| the analog hole or because the mechanism of operation is
| observable to anyone who buys the product.
|
| The incentives to uncover those things are also much
| stronger in modern day because of the connectedness of
| the world. If there were two wheelwrights in your town
| and one of them had a secret process, no one but the
| other would have any use for it and if they found out
| they wouldn't even have any else to tell it to.
|
| If someone had a secret video encoding strategy today,
| some hobbyists would reverse engineer it and post it on
| the internet.
|
| > Patents, for example, ensure that at least some minimal
| description of the process gets published for others to
| take inspiration from.
|
| Have you read a modern patent? They're inscrutable, and
| to the fullest extent allowable attempt to claim the
| overall concept of doing something rather than describing
| a specific implementation.
| DeathArrow wrote:
| >Its also not clear for example that Studio Ghibli lost by
| having their art style plastered all over the internet.
|
| Maybe Studio Ghibli is much more than merely a style. Maybe
| people aren't looking at their production just for the style.
|
| Most people dislike wearing fake clothes and the dislike
| wearing fake watches or fake jewelry. Because it isn't just
| about the style.
| pixl97 wrote:
| >Most people dislike wearing fake clothes and the dislike
| wearing fake watches or fake jewelry
|
| I'd disagree. Most people don't like buying something
| 'real' then finding out it's fake. Far more people don't
| mind an actual fake if it's either high quality or is very
| low priced.
| thot_experiment wrote:
| Yeah ngl I have some fake designer stuff I got as a gift
| and I love it, and I especially love that it's fake. It
| feel like I'm pulling one over on the tryhards that care
| about that stuff being real, but I still get to enjoy the
| wild LV coat I have and ain't nobody checking the
| stitching on the lining to make sure it's the real thing.
| I could see myself buying more fakes in the future, but
| I'd never ever buy the real thing.
| wavemode wrote:
| Companies like Studio Ghibli are not being harmed by AI,
| small freelance artists are.
| masswerk wrote:
| I think, Studio Ghibli will be affected, as well, since
| their "trademark style" (as we used to say), formerly a
| welcome sight and indicative for a certain type of story
| telling, will be devaluated as an indicator for slop. (Much
| like there are certain traits of an image, which we
| associate with soap operas and assume to be indicative of a
| low-value production.)
| butlike wrote:
| I doubt that. "Which movie does the slop belong to?" "Oh
| none of them? Ok" Is a pretty easy search term
| masswerk wrote:
| I doubt that, when confronted with an image that you've
| learned to associate with a plethora of low-quality /
| low-effort productions, you'd search for the possible
| origin, in the first place.
|
| (After all, it's yet another ephemeral image in "that AI
| style", with no apparent thought having gone into it,
| just some name dropping, at best. Or some generated,
| senseless story, you would be glad, the algorithm hadn't
| pointed your kids at. Why should you?)
| kelseyfrog wrote:
| I'd take it farther to say that copyright and intellectual
| property is a legal fiction that ultimately benefits the
| wealthy[those who can pay to legally enforce it] over small
| artists.
|
| Small artists get paid to create the art; corporations
| benefit from exclusivity.
| thwarted wrote:
| > _Its also not clear for example that Studio Ghibli lost by
| having their art style plastered all over the internet. I
| went home and watched a Ghibli film that week, as I 'm sure
| many others did as well. Their revenue is probably up quite a
| bit right now?_
|
| This sounds like a rewording of "You won't get paid, but this
| is a great opportunity for you because you'll get exposure".
| Taek wrote:
| Exposure has value! The meme around trying to pay artists
| with exposure is because some people think their "exposure"
| has meaningful value when they are offering to expose the
| artist to 100 people, 99 of whom aren't likely even target
| customers.
|
| Studio Ghibli on the other hand had exposure to millions of
| people (maybe hundreds of millions), and probably >5% of
| those were potential customers.
|
| So yes, being paid in exposure makes sense, if the exposure
| is actually worth what the art is worth. But most people
| offering to pay in exposure are overvaluing their exposure
| by 100x or more.
| __loam wrote:
| Nearly every artist I've spoken to or have seen talk about
| this technology says it's evil, so at least among the victims
| of this corporate abuse of the creative community, there's
| wide consensus that it's bad.
|
| > but I certainly don't think that AI without restrictions is
| going to lead to fewer people with art jobs.
|
| It's great that you think that but in reality a lot of
| artists are saying they're getting less work these days.
| Maybe that's the result of a shitty economy but I find it
| very difficult to believe this technology isn't actively
| stealing work from people.
| mycall wrote:
| > "How can we monetize art" remains an open question for
| society
|
| Yet much of the best art imho is in the wild to the element
| while being at home at some random place. Or perhaps in
| someone's collection forgot and displaced. Art's worth will
| always be an open question.
| mrdependable wrote:
| Studio Ghibli might not have been affected yet, but only
| because the technology is not there yet. What's going to
| happen when someone can make a competing movie in their style
| with just a prompt? Should we all just be okay with it
| because it's been decided that Studio Ghibli has made enough
| money?
|
| If the effort required to create that can just be ingested by
| a machine and replicated without consequence, how would it be
| viable for someone to justify that kind of investment? Where
| would the next evolution of the art form come from? Even if
| some company put in the time to create something amazing
| using AI that does require an investment, the precedent is
| that it can just be ingested and copied without consequence.
|
| I think aside from what is legal, we need to think about what
| kind of world we want to live in. We can already plainly see
| what social media has done to the world. What do you honestly
| think the world will look like once this plays out?
| hnbad wrote:
| > What the post should say is "we know that this is unfair to
| artists, but the tech companies are making too much money from
| them and we have no way to force them to change".
|
| It seemed a fact of life that companies will just abuse your
| personal data to their liking and can do what they want with
| information they collect about you because "if it's free,
| you're the product" (and even if you paid for it, "you should
| know better" etc). Then GDPR and its international derivatives
| came along and changed that.
|
| It seemd a fact of life that companies that _technically_ don
| 't have an _actual_ market monopoly can do whatever they want
| within their vertically integrated walled gardens because
| competitors can just create their own vertically integrated
| walled gardens to compete with them and the rules for markets
| don 't apply to walled gardens. Then the DSA and DMA came along
| and changed that.
|
| I don't see why legislation can't change this, too. Of course
| just with the GDPR, DSA and DMA we'll hear from libertarians,
| megacorps and astroturf movements how unfair it all is to mom &
| pop startups and how it's going to ruin the economy but I think
| given the angle grider the US is currently taking to its own
| economy (and by extension the global economy because we're all
| connected), I think that's no longer a valid argument in
| politics.
| DeathArrow wrote:
| >> it's unfair for artists to have their works sucked up
|
| What framework can we use to decide if something is fair or
| not?
|
| Style is not something that should be copyrighted. I can pain
| in the style of X painter, I can write in the style of Y
| writer, I can compose music in the style of Z composer.
|
| Everything has a style. Dressing yourself has a style. Speaking
| has a style. Even writing mathematical proofs can have a style.
|
| Copying another person's style might reflect poor judgement,
| bad taste and lack of originality but it shouldn't be illegal.
|
| And anyone in the business of art should have much more than a
| style. He should have original ideas, a vision a way to tell
| stories, a way to make people ask themselves questions.
|
| A style is merely a tool. If all someone has is a style, then
| good luck!
| gosub100 wrote:
| In music, someone can sing the same style as another, but if
| they imitate it to the point that there is brand confusion,
| where the consumer believes the product came from X when it
| actually came from Y, that's clearly crossing the line.
| ChadNauseam wrote:
| Is that actually crossing a line? I'm sure some consumers
| have thought that rocket league was associated with FIFA,
| or that studio Ghibli movies were made by Disney. But these
| aren't widespread issues because we have a robust system of
| trademarks that draws a clear line: you can't use
| trademarked names or iconography in a way that causes
| confusion. But if some people heading Olivia Rodrigo's
| "good 4 you" think they're listening to Paramore because
| they have a similar style, that has never been illegal
| gosub100 wrote:
| vanilla ice vs queen case doesn't support your claim.
| ChadNauseam wrote:
| vanilla ice completely copied queen's baseline, not just
| the queen style. (not to mention that it never went to
| court)
| yencabulator wrote:
| It's already gone quite a bit further than "style".
| https://www.404media.co/listen-to-the-ai-generated-ripoff-
| so...
| wat10000 wrote:
| "Fair" doesn't matter. The only consensus that matters is what
| is legal and profitable. The former seems to be pretty much
| decided in favor of AI, with some open question about whether
| large media companies enjoy protections that smaller artists
| don't. (The legal battle when some AI company finally decides
| to let their model imitate Disney stuff is going to be epic.)
| Profitable remains to be seen, but doesn't matter much while
| investors' money is so plentiful.
| __loam wrote:
| > The former seems to be pretty much decided in favor of AI
|
| None of the cases against AI companies have been decided
| afaik. There's a ton of ongoing litigation.
|
| > but doesn't matter much while investors' money is so
| plentiful.
|
| More and more people are realizing how wasteful this stuff is
| every day.
| rkharsan64 wrote:
| Are there any local models that use this new approach to
| generating images?
| GaggiX wrote:
| GPT-4o is the only model that seems to work well in the text-
| image joint space to this degree, even Gemini Flash 2.0 with
| native image support is not nearly as good so it will probably
| be a while for a good open source alternative to pop up (a
| while in the context of AI development).
| gerash wrote:
| depends on the use case.
|
| I used GPT-4o for some image editing (adding or removing
| things) to an image of a person and they distort the look of
| the people after each edit but (Gemini Flash + image out) did
| much better.
|
| The main problem is there is little control. For example I
| asked to add a helicopter to an image in a ski resort but
| then it seems cumbersome for me to have to write a full
| paragraph to describe where exactly I want this helicopter to
| be rather than if I could just do it by dragging things with
| a mouse.
| DeathArrow wrote:
| Yes, there's HiDream which yields even better results than 01.
|
| https://github.com/HiDream-ai/HiDream-I1
| GaggiX wrote:
| This is just a diffusion text-to-image model like many
| others, completely different than a LLM with a native image
| support.
| orbital-decay wrote:
| 4o still exhibits the "pink elephant effect", it's just...
| subtler, and tends to reveal itself on a complex or confusing
| prompt. Negations are also still not handled properly, they tend
| to slightly confuse the model and decrease the accuracy of the
| answer or the generated picture. The same is true for any other
| LLM. Moreover, the author is asking the model to rationalize the
| decision he already made ("tell me why there can't be any
| elephants"), which could work as an equivalent to a CoT step.
|
| It's "just" a much bigger and much better trained model. Which is
| a quality on its own, absolutely no doubt about that.
| Fundamentally the issue is still there though, just less
| prominent. Which kind of makes sense - imagine the prompt "not
| green", what even is that? It's likely slightly out of
| distribution and requires representing a more complex
| abstraction, so the accuracy will necessarily be worse than
| stating the range of colors directly. The result might be
| accurate, until the model is confused/misdirected by something
| else, and suddenly it's not.
|
| I think in the end none of the architectural differences will
| matter beyond the scaling. What will matter a lot more is data
| diversity and training quality.
| danielbln wrote:
| But it's literally a different architecture (auto-regressive,
| presumably sequence based vs diffusion). In my experiments it
| is significantly, overwhelmingly better at consistency,
| coherence and prompt adherence. Things I needed control nets
| before it just... does it. and even zooming into fine details,
| they make sense.
|
| Here is an example with a bunch of negations:
| https://i.imgur.com/P8G5ICs.png
| wat10000 wrote:
| It's interesting that out of all the aquatic animals it could
| have used, it chose one that perhaps looks the most like an
| elephant.
| danielbln wrote:
| That's just luck of the draw, in a new session with the
| same prompt it outputted a bird:
| https://i.imgur.com/SLT8cYe.png
| wat10000 wrote:
| I'm not convinced. I tried it and it showed me a swimming
| hippopotamus, which is even more elephant-like than the
| turtle. I tried again and it gave me a pelican, which is
| not generally very elephantish, but this particular one
| has a gray body with a texture that looks a lot like
| elephant skin.
| orbital-decay wrote:
| Of course, it's a tiny specialized model vs a big generalist
| model. They're absolutely incomparable in size/quality
| though, especially of the text part. How much of this happens
| because of the poor encoder and worse training in other
| models, and how much due to the architectural differences?
| I'm not saying it's not better than the existing image gen
| models somehow, but it's pretty hard to separate the two
| since both are present. All current SotA LLMs including 4o
| itself have negation inaccuracy in text (you need a really
| complex prompt, or a long one with thousands of tokens, not a
| toy one), and I don't see why this one should behave
| differently in similar conditions. Especially considering
| that it also suffers from pretty much the same artifacts as
| other image models, just much less (fingers, extra limbs,
| perspective/lighting issues, overfitting, struggles with out-
| of-distribution generation etc.)
| d4rkp4ttern wrote:
| Diagrams are still a big unsolved problem. Making diagrams for a
| talk or paper is an extremely tedious process and I am still
| waiting for a good multimodal LLM solution for this. It should
| take a sketch and/or text description of what you want and in a
| few iterations you should get what you want. GPT4o tries hard but
| results are still bad.
| chthonicdaemon wrote:
| I've had the best luck at getting it to produce diagrams-as-
| code like mermaid or plantuml.
| d4rkp4ttern wrote:
| I know but those diagrams often don't adequately capture what
| I want. Think of diagrams in nice technical talks or papers.
| I've even tried having the LLM (Claude) generate SVGs. They
| all fall short.
| shubhamjain wrote:
| The Ghibli trend completely missed the real breakthrough -- and
| it's this. The ability to closely follow text, understand the
| input image, and maintain context of what's already there is a
| massive leap in image generation. While Midjourney delivered
| visually stunning results, I constantly struggled to get anything
| specific out of it, making it pretty much useless for actual
| workflows.
|
| 4o is the first image generation model that feels genuinely
| useful not just for pretty things. It can produce comics, app
| designs, UI mockups, storyboards, marketing assets, and so on. I
| saw someone make a multi-panel comic with it with consistent
| characters. Obviously, it's not perfect. But just getting there
| 90% is a game changer.
| empath75 wrote:
| I had chatgpt generate a flow chart with mermaid js for
| something at work and then write a scott mccloud style comic
| book explaining it in detail and it looked so convincing, even
| though it got some of the details a bit wrong. It's _so close_
| to making completely usable graphics out of the box.
| freeamz wrote:
| Hmmm isn't stable diffusion already doing that?
| aenvoker wrote:
| SD has a very primitive conceptual model. Basically "bag of
| words nudging pixels around for a while". Words near each other
| influence each other. But, there's nearly no understanding of
| grammar.
|
| Midjourney is similar with text prompts. But, with image
| prompts it is able to understand content separately from style.
| You can give it a photo of two people and it can return many
| images of recognizable approximations of those people in
| different poses.
|
| SD can only start from pixels, blur and deblur those pixels in
| place.
|
| MJ image prompts probably works via image-to-tokens added on to
| your text-to-tokens-to-image.
| thrance wrote:
| Each generation follows the prompt a little bit better than the
| last, but I don't see any revolutions. Fingers are still messed
| up, eyes are wonky and legs sometimes still fork into two.
| Fundamentally it's still the same diffusion technique, with the
| same limitations.
| gcanyon wrote:
| It's interesting to hear people side with the artists when in
| previous discussions on this forum I've gotten significant
| approval/agreement arguing that copyright is far too long.
|
| As I've argued in the past, I think copyright should last maybe
| five years: in this modern era, monetizing your work doesn't
| (usually) have to take more than a short time. I'd happily
| concede to some sort of renewal process to extend that period,
| especially if some monetization method is in process. Or some
| sort of mechanical rights process to replace the "public domain"
| phase early on. Or something -- I haven't thought about it _that_
| deeply.
|
| So thinking about that in this process: everyone is "ghiblifying"
| things. Studio Ghibli has been around for very nearly 40 years,
| and their "style" was well established over 35 years ago. To me,
| that (should) make(s) it fair game.
|
| The underlying assumption, I think, is that all the "starving"
| artists are being ripped off, but are they? Let's consider the
| numbers -- there are a handful of large-scale artists whose work
| is obviously replicable: Ghibli, the Simpsons, Pixar, etc. None
| of them is going hungry because a machine model can render a prom
| pic in their style. Then you get the other 99.999% of artists,
| _all_ of whose work went into the model. They _will_ be hurt, but
| not specifically because _their_ style has been ingested and
| people want to replicate _their_ style.
|
| Rather, they will be hurt because no one knows their style, nor
| cares about it; people just want to be able to say e.g. "Make a
| charcoal illustration of me in this photo, but make me sitting on
| a horse in the mountains."
|
| It's very much like the arguments about piracy in the past:
| 99.99% of people were never going to pay an artist to create that
| charcoal sketch. The 0.01% who might are arguably causing harm to
| the artist(s) by not using them to create that thing, but the
| rest were never going to pay for it in the first place.
|
| All to say it's complicated, and obviously things are changing
| dramatically, but it's difficult to make the argument that
| "artists need to be compensated for their work being used to
| train the model" without both a reasonable plan for how that
| might be done, and a better-supported argument for why.
| Taek wrote:
| I find it unlikely that someone who was willing to pay an
| artist for a charcoal sketch would be satisfied with an AI
| alternative.
|
| You don't just buy art for the aethstetic, you buy it for a lot
| of reasons and AI doesn't give any of the same satisfaction.
| zwnow wrote:
| I'm all for paying artists for their work. Unfortunately,
| same as tattoo artists, some just heavily overcharge for
| mediocre results (been tattooing myself AND I know a few
| things about art). Like, sorry, but if you want to earn money
| doing art, please be good at it...
| RhysU wrote:
| > Like, sorry, but if you want to earn money doing art,
| please be good at it...
|
| By definition, almost half of all $ARTISTS are worse than
| the median. Should that half not get paid for their time?
| zwnow wrote:
| Yes. They should look for a job that actually covers
| their lifestyle instead of crying about AI taking their
| jobs.
|
| They can always put in more hours and become better. I
| cant imagine they have a lot of paying customers anyway.
| WindyMiller wrote:
| What about the half of the remaining artists that are
| below the new median?
| zwnow wrote:
| Should be good enough to alrdy have established some
| customers.
| flkiwi wrote:
| The problem with that is that people aren't asking for AI
| generated images in the style of Raven from Topeka with
| an Etsy shop. They're asking for Ghibli. So the people
| whose livelihoods are most directly impacted are
| (assuming they're not centuries dead) the famous,
| talented, and trend-making artists, not the lower tier
| making bad Precious Moments knockoffs. Society's problem
| is understanding that not wanting to pay for bad Precious
| Moments knockoffs is rational, while not wanting to pay,
| say, a Studio Ghibli for quality, professional creativity
| is insane.
| nasmorn wrote:
| Couldn't Ghibli zeitgeist moment lead to them making out
| hugely with a new release or just a cinema screening of
| Totoro right now?
| flkiwi wrote:
| That is "artists should be grateful to work for exposure"
| on a grand scale.
| nasmorn wrote:
| Except they didn't do any work for the exposure. If a
| marketing agency had come up and executed the Ghiblify
| everything model as a PR stunt we would call it the most
| genius creative campaign of the decade
| flkiwi wrote:
| They did though. The studio engaged in tremendous amounts
| of work and created good will, in addition to their
| specific creative works. Their visual style is tied up in
| that good will. Use of the visual style _for profit_
| without consent is, at least ethically, misappropriation
| of another 's value. And "You should be pleased I used
| your creative work because now more people will know
| about you and you will make a lot of money from this!" is
| one of the oldest defenses to misappropriation of
| creativity.
|
| I'm not even mad. We do a terrible job in our society of
| valuing artists and creative people generally and in
| explaining the value of intangible things, especially
| something like good will. People have been
| misappropriating fonts and clipart and screenshots in
| presentations and posters and whatnot, duplicating clever
| branding ideas and the creative efforts of others, and so
| on for _decades_ if not longer, all without ill intent.
| It's something we need to fix and never will. But when
| that becomes a channel for another to directly profit, it
| begins to venture out of harmlessness.
| zwnow wrote:
| As if the Ghibli trend wasnt just a short trend people
| will have forgotten about in 4 weeks... Also I couldnt
| care less about big studios, they print money anyway.
| RhysU wrote:
| I keep waiting for physical objects to become important
| again. AI isn't coming for the ceramic folks.
| zwnow wrote:
| I think that would be great. Traditional art market is
| not nearly as big as the digital space. I'd love if
| people valued traditional art again as thats the only
| stuff I do.
| ane wrote:
| Depends on the niche. Original physical art for trading
| card games or comics is a significant chunk of the income
| of your typical artist. Digital art in those niches does
| not have this source of income. But then again digital
| art has other niches where the actual commission rates
| are high enough to not make this a problem.
| ben_w wrote:
| > AI isn't coming for the ceramic folks.
|
| *looks at 3D printer on desk that can apparently handle
| ceramic filaments, thinks about all the mass-produced
| ceramics sold in supermarkets*
| amazingamazing wrote:
| > some just heavily overcharge for mediocre results
|
| if people are paying, then they aren't "overcharging"
| zwnow wrote:
| In the tattoo business people have no other places to go.
| Charging 1kEUR+ for half a sleeve is extremely
| overcharged. If people are paying they often simply don't
| have enough alternatives.
| flkiwi wrote:
| Given the time commitment and network of basic
| biological, anatomical, and health knowledge required,
| that doesn't strike me as an insane price, assuming an
| artist who is able to create the requested art.
| zwnow wrote:
| Tbh you barely have to know anything, most important one
| is how deep to go with the needle and sanitizing.
| Everything else is not rlly important.
| flkiwi wrote:
| A friend who is heavily inked has gone on at length to me
| about understanding skin elasticity--particularly how it
| changes over a lifetime--as well as the way joints and
| muscles change and distort visual lines, etc. It sure
| seems like a skilled trade to me.
|
| And, I don't know, depth of penetration of a needle in
| flesh and sanitation don't strike me as minor things to
| get right.
| zwnow wrote:
| People love to make things seem harder than they are. I
| tattoo people, I am aware about skin types, usually thats
| not a big issue unless its heavily scarred. The quality
| of your tattoo machine matters most, as my 70EUR eBay
| makeshift one wasnt nearly as good as a proper one.
| Amount of ink matters, needle depth, skin type, sweat.
| But thats stuff you have figured put after your 20th
| tattoo. Its like knowing datatypes in programming. You
| just know stuff after some practice.
| UncleEntity wrote:
| > But thats stuff you have figured put after your 20th
| tattoo.
|
| So... after spending hundreds, if not thousands, of hours
| learning a skill?
|
| I got a tattoo back in the day and specifically went to
| one the guys in my platoon said was good due to him being
| featured in magazines or whatever. It's kind of an
| important thing to get right on the first, not 20th,
| attempt IMHO.
| zwnow wrote:
| 20 tattoos hundreds of hours? Lol.
| autoexec wrote:
| > People love to make things seem harder than they are.
|
| In my experience people tend to underestimate or downplay
| how difficult something will be or how complex it is.
| This happens in people who know only a little about
| something, but also in people who are highly experienced
| because it becomes normal and easy for them and they can
| quickly evaluate a situation and know which
| considerations don't apply.
| bongodongobob wrote:
| Penetration is practice, sanitation is basically use
| gloves and an autoclave.
| zwnow wrote:
| Also swap needles and use some cream, people really
| overestimate how hard tattooing is
| cannonpr wrote:
| Being someone who has paid a lot of attention to Ghibli, I
| wouldn't say their style was well established 35-40 years
| ago... There is considerable evolution and refinement to their
| style from Naushika, to later works, both in the artistic style
| and the philosophical content it presents.
|
| I think allowing it to be fair game would have destroyed
| something quite beautiful that I've watched evolve across 40
| years and which I was hoping to see the natural conclusion of
| without him being bothered by the AI-fication of his work.
| gcanyon wrote:
| Yeah, of course their style isn't static, but I was taking
| Kiki's Delivery Service (1989) as a point where much of their
| visual style was pretty well-established.
| cannonpr wrote:
| I agree that some of his main elements were relatively set
| by then but others continue to evolve. Personally I feel
| that his work is entirely fair game after his death, by
| which I mean to say after the end of his journey that he
| has poured his entirety into, in terms of value to humanity
| and respect to him for that value I think it's reasonable
| to respect his wishes during his lifespan.
| xrd wrote:
| The Ghibli fight is the same fight that is being fought in the
| NASDAQ. That is to say, there was an established set of rules
| that everyone thought were fixed and now they are being
| radically disrupted. Both the creative industry and the general
| business industry are trying to figure out what is life going
| to be like with a totally different and fluid set of
| regulations, whether it be copyright law or tariffs.
|
| No wonder sama and Trump are so cozy. They both see the same
| legacy.
| Avshalom wrote:
| >It's interesting to hear people side with the artists when in
| previous discussions on this forum I've gotten significant
| approval arguing that copyright is far too long.
|
| Well broadly that's because most arguments about
| copyright(length/scope) are made against corporations attacking
| individual artists and arguments about copyright(AI/scope) are
| made against corporations attacking individual artists.
| amazingamazing wrote:
| > I think copyright should last maybe five years: in this
| modern era, monetizing your work doesn't (usually) have to take
| more than a short time. I
|
| funny how people who say this kind of stuff are never content
| creators (in the monetization sense).
| bko wrote:
| There are a lot of programmers on this platform (myself
| included), and I love that my work has an impact on others.
|
| I have a number of public repos and I have benefitted greatly
| from other public repos. I hope LLMs made some use of my
| code.
|
| I wrote blogs for years without any monetization. I hope my
| ideas influenced someone and would be happy if they made some
| impact on the reasoning of LLMs.
|
| I'm aware of patent trolls and know people with personal
| experience with them.
|
| So I generate a lot more content that the typical person and
| I am still in favor of much looser IP rights as I think they
| have gone overboard and the net benefit for me, a content
| creator, is much greater having access to the work of others
| and being able to use tools like LLMs trained on their work.
| amazingamazing wrote:
| posting stuff for free is different than selling stuff.
| 6510 wrote:
| If you make a blog with nice original long form articles it
| may take much longer to gain traction. Reproducing the
| content in "your own" wording quickly gets fuzzy.
|
| I like the practical angle. Any formula that requires
| monitoring what everyone is doing is unworthy of
| consideration. Appeal to tradition should not apply.
| gcanyon wrote:
| Not that it should impact the validity of my argument, but I
| have sold commercial software in the past, and it is absurd
| that that software will be copyrighted through most of the
| 21st-century.
| rikroots wrote:
| My personal preference is for (say) 15-20 years.
|
| And, as a content creator, I practice what I preach - at
| least when it comes to my poetry:
| https://rikverse2020.rikweb.org.uk/blog/copyrights
| pixl97 wrote:
| I mean, this is about as useful as saying anti-slavery people
| should become slave owners so they understand the hardships
| of making money.
|
| My example is extreme to the absurd, so how about we go with
|
| >It's difficult to get a man to understand something when his
| salary depends on not understanding it.
| AlecSchueler wrote:
| It's one thing to argue that copyright terms should be
| shortened, and another to accept that a handful of corporations
| should be able to forcefully shorten it for certain actors
| entirely on their own terms.
| ben_w wrote:
| Mm.
|
| The arguments about wanting copyright to be life+70 have always
| felt entitled, to me. Making claims about things for their kids
| to inherit, when the median person doesn't have the option to
| build up much of an inheritance anyway, and 70 years isn't just
| the next generation but the next 2.5 generations.
|
| I don't know the exact duration of copyright that makes sense,
| the world changes too much and different media behave
| differently. Feels like nobody should have the right to block
| remakes of C64 games on copyright grounds, but I wouldn't
| necessarily say that about books published in the same year.
|
| From what I've seen about the distribution of specifically book
| sales, where even the top-100 best sellers often don't make
| enough to justify the time involved, I think that one of the
| biggest problems with the economics of the arts is a mixture of
| (1) the low cost of reproduction, and (2) _all the other
| artists_.
|
| For the former: There were political campaigns a century ago,
| warning about the loss of culture when cinemas shifted from
| live bands to recorded music[0]; Today, if I were so inclined,
| I can for a pittance listen to any of (I'm told) 100 million
| musical performances, watch any of 1.24 million movies or TV
| shows. Even before GenAI, there was a seemingly endless
| quantity of graphical art.
|
| For the latter: For every new book by a current living author
| such as Charlie Stross (who is on here sometimes), my limited
| time is also spread between that and the huge back-catalogue of
| old classics like the complete works of Conan Doyle, Larry
| Niven, or Terry Pratchett.
|
| [0] https://www.smithsonianmag.com/history/musicians-wage-war-
| ag...
| jeffreygoesto wrote:
| I thought the US wants to re-industrialize now? Then 5 years is
| laughably short to protect your invest.
| Workaccount2 wrote:
| As you grow older and run through more cycles of general
| opinions, you realize that pretty much everyone is in it for
| themselves, what serves them best, and support what narrative
| aligns with that.
|
| 2007: Copyright is garbage and must be abolished (so I can get
| music/movies free)
|
| 2025: Copyright needs to be strengthened (so my artistic
| abilities retain value)
| UncleEntity wrote:
| You forgot:
|
| 2024: What do you mean I can't copyright AI generated
| artwork?
| Der_Einzige wrote:
| Correct. This is why Stirner is the best Philosopher.
| https://en.wikipedia.org/wiki/The_Ego_and_Its_Own
|
| There is nothing other than Egoism.
| another-dave wrote:
| I'd agree with limiting copyrights but would do it based on
| money earned rather than time, so something like when you make
| $X million, the work becomes public domain.
|
| As a specific example -- _A Game of Thrones_ was released in
| 1996. It picked up awards early on but only became a NYT best
| seller in 2011, just before the TV show aired.
|
| It would feel harsh for an author to loose all their copyright
| because their work was a "slow burn" and 5 years have elapsed
| but they've made little to no money on it.
| pixl97 wrote:
| >o something like when you make $X million, the work becomes
| public domain.
|
| https://en.wikipedia.org/wiki/Hollywood_accounting
|
| No, no metrics that can be gamed.
| gcanyon wrote:
| It's a super-interesting idea, but GoT seems highly cherry
| picked: the vast majority of all works would never leave
| copyright if the requirement was that they clear even $1000.
| qiqitori wrote:
| Wha- wha- what? I tried to generate an image in ChatGPT after the
| announcement a while back and the image wasn't bad, but the text
| on it (numbers) was nonsense. (Analog gauge with nonsense numbers
| instead of e.g. 10, 20, 30, 40, etc.)
|
| Gave it another chance now, explicitly calling out the numbers.
| Well, they are improved but not sure how useful this result is
| (the spacing between numbers is a little off and there's still
| some curious counting going on. Maybe it kind of looks like the
| numbers are pasted in after the fact?
|
| https://chatgpt.com/share/67f4fa33-70dc-8012-8e1e-2dea563d3d...
| nvalis wrote:
| These images are still created with the old model. The share
| link states "Made with the old version of image generation. New
| images coming soon." below the first image.
| haswell wrote:
| > _The question isn 't whether these tools will change visual
| media, but whether we'll be thoughtful enough to shape that
| change intentionally._
|
| Unfortunately I think the answer to this question is a resounding
| "no".
|
| The time for thoughtful shaping was a few years ago. It feels
| like we're hurtling toward a future where instead we'll be left
| picking up the pieces and assessing the damage.
|
| These tools are impressive and will undoubtedly unlock new
| possibilities for existing artists and for people who are
| otherwise unable to create art.
|
| But I think it's going to be a rough ride, and whatever new
| equilibrium we reach will be the result of much turmoil.
|
| Employment for artists won't disappear, but certain segments of
| the market will just use AI because it's faster, cheaper, and
| doesn't require time consuming iterations and communication of
| vision. The results will be "good enough" for many.
|
| I say this as someone who has found these tools incredibly
| helpful for thinking. I have aphantasia, and my ability to
| visualize via AI is pretty remarkable. But I can't bring myself
| to actually publish these visualizations. A growing number of
| blogs and YouTube channels don't share these qualms and every
| time I encounter them in the wild I feel an "ick". It'll be
| interesting to see if more people develop this feeling.
| pixl97 wrote:
| >But I think it's going to be a rough ride, and whatever new
| equilibrium we reach will be the result of much turmoil.
|
| Honestly visual media just seems to be the start. In the past
| two years we've seen about as much robotics progress as the
| last 20. If this momentum keeps up then we're not just talking
| about artists that are going to have issues.
| eapriv wrote:
| It's always fun to read posts like that: they say "look at this
| amazing thing it drew", and the image is utter garbage.
| NiloCK wrote:
| The 'before' image passes the test this time in a "Treachery of
| Images" sort of way.
| morkalork wrote:
| Huh, the coffee table reminds me of all those cheap e-retailers
| who very clearly (and badly) photoshop their clothes on to same 2
| or 3 stock model images. If you thought shopping online sucked
| before, it's just going to get even worse now.
| vunderba wrote:
| 4o, despite OpenAI's practically draconian content policies, is a
| pretty big leap forward. I put together a comparison of some of
| the most competitive generative models (Imagen, 4o, Flux, and
| MJ7) where I prioritized increasingly difficult _prompt
| adherence_. If Imagen 3 had 4o 's multimodal capabilities (being
| able to make constant adjustments against a generated image by
| prompting) I would say its nearly on-par with 4o.
|
| https://genai-showdown.specr.net
| ge96 wrote:
| > we guac you covered
| xnorswap wrote:
| The "How to build a boardgame" infographic looks like half my
| linkedin "feed" now, but a boardgame instead of random basic
| programming / recruimentment / sales topic.
|
| Feed is in quotes because my feed seems to be 90% suggested
| posts.
| tantalor wrote:
| This could help:
|
| https://www.linkedin.com/help/linkedin/answer/a1379064
| hansmayer wrote:
| > " Image generation is likely to be very disruptive in ways we
| don't understand right now. " Is anyone getting tired of these
| formulations ? When a tech is disruptive, we know it immediately.
| Uber was disruptive. AirBnB, Gmail, Amazon, even Facebook at one
| point. You just knew it, nobody was writing long essays trying to
| justify those products. Robots generating statistically median
| images is impressive, but not disruptive at all. If something is
| "likely" to be "disruptive", but in ways "we don't understand
| yet", how can the claim even be made? What is it based on? If we
| do not understand it yet, how can we understand if it is "likely
| to be disruptive".
| Der_Einzige wrote:
| The #1 reason that this technology won't proliferate more quickly
| is that humans are a bunch of COOMERS!
|
| We get Stable Diffusion V1.5 and SDXL and what does the community
| go do with it? Lmao see civit.ai and it's literal hundreds of
| thousands of NSFW loras. The most popular model today on that
| website is the NSFW anime version of SDXL, called "Pony
| Diffusion" (I'm literally not making this up. A bunch of Bronies
| made this model!)
|
| Imagine that an open source image generator which does tokens
| autoregressively like this at this quality is released.
|
| The world is simply not ready for the amount of horny stuff that
| is going to be produced (especially without consent). It appears
| that the male libido really is the reason for most bad things in
| the world. We are truly the "villains of history".
| smusamashah wrote:
| I am waiting for when I could provide these a scene snippet from
| "Hitchhiker's Guide To Galaxy" (or any book) and it could draw
| that for me. The gold planets, the waking up on the beach, total
| perspective vortex etc.
|
| I like the book, but there are quite a few scenes which are quite
| hard to visualize and make sense. An image generator that can
| follow that language and detail will be amazing. Even more
| awesome will be if it remains consistent in follow ups.
| wrboyce wrote:
| I love the idea, but I feel like I have to say that I've got a
| pretty solid idea of what the total perspective vortex would
| look like for someone being subjected to it. When I first read
| the books I immediately had a visual and that has never changed
| when I've read them again (and again...).
|
| I'm not sure what that says about either of us, but I would say
| that your definitive "quite hard to visualise" statement is
| very much subjective.
| smusamashah wrote:
| Vortex may be not so much but there are other hard to
| visualize things. I am on the third book, and I have no idea
| what Beeblebrox's two heads look like. Second head is often
| mentioned in passing. Sometimes its mentioned as its always
| there, other times it feels like its just pops out of
| somewhere, otherwise, it's like it doesn't exist.
|
| There is the scene when they see themselves on the beach on
| first rescue by the ship. That was hard to grasp. Or the
| insides of the ship itself, the bridge, the panels etc. Also
| that black ship they stole.
|
| But may be its just me having a hard time with these
| concepts.
|
| It's not just about scene being difficult to visualize, even
| if I can see them in my head, I want to see them on paper too
| because those thing excite me.
| thwarted wrote:
| We have alt text for images, you want alt images for text.
|
| You can see other people's interpretation of Zaphod's two
| heads by watching the BBC HHGTTG show (Mark Wing-Davey) or
| the movie (Sam Rockwell), among other renditions, which
| offer completely different interpretations, none of them
| canonical (not the least of which is because there was no
| canonical version of HHGTTG according to DA). I'm sure
| there are multitudes of fan art for HHGTTG on deviantart.
| Having AI generate an image doesn't offer any more
| "official" visualization.
|
| Zaphod's second head is mentioned just as much is
| warranted. If a character has a limp or a crazy haircut it
| is not mentioned every time, because it has nothing to do
| with what is going on. And the book mentions that one head
| is often distracted/asleep, so it sounds like you do have a
| good visual of what his two heads are like.
|
| While I understand that people think differently and some
| people are more visual thinkers, a good portion of the
| concepts expressed through writing are meant to be
| mindfucks that are difficult to express visually. A picture
| may be worth a thousand words, but the meat of writing is
| usually not the visual representation of its concepts.
| That's a great thing about writing: you can fill in the
| visuals yourself and it's fodder for fans to discuss.
|
| (BTW, Hotblack Desiato's ship would just be black. Your
| eyes couldn't focus on it. Even the controls were black
| labels on a black background. There is nothing here to
| visualize other than, well, blackness).
| smusamashah wrote:
| I agree that people interpret things differently and
| visualise differently and that is my point. I want to see
| the concept from my head in a solid visual form. Some
| concepts that are not clear, like those that I mentioned
| having trouble with, I want to see any kind of visual
| representation at all. I will probably not like some of
| those, but there these tools can help generate a bunch of
| variations _tailored specifically for me_. I can choose
| one and carry on with that. I can come back and more
| details to that as I read further.
|
| If someone can _show_ me _exactly_ what I am thinking of,
| won 't that be amazing.
| pixl97 wrote:
| >Hotblack Desiato's ship would just be black
|
| These days we have paint that black, though we can't
| reproduce the effect on the monitor.
|
| Those superblacks really do mess with your mind. It's
| like a cutout of the void.
| wrboyce wrote:
| Again, I have to disagree - which I suppose reinforces the
| whole subjectivity angle. I was positive that Zaphod's two
| heads were side-by-side to the extent that it pissed me off
| a fair bit in the most recent movie adaption (among, let's
| face it, plenty of other candidates).
|
| I don't know if the "layout" of the heads is mentioned or
| not in the books - I'd have to go back and check - but it's
| often quite jarring when a book becomes a movie and doesn't
| match my inner vision (and how incredibly unthoughtful of
| them, to boot).
| ARandumGuy wrote:
| I've seen stuff that echoes this sentiment before, and I have
| to say I don't understand this desire at all. Why would I need
| a computer to show me what something in a book looks like? I
| already have an imagination for that!
|
| Books are fundamentally a collaborative artform between the
| author and the reader. The author provides the blueprint, but
| it's up to the reader to construct the scene in their own head.
| And every reader is going to have slightly different
| interpretations based on how they imagine the events of a book.
| This act of imagination and re-interpertation is one of the
| things I love about reading books.
|
| Having a computer do the visualization for you completely
| destroys what makes books engaging and interesting. If you
| don't want to visualize the book yourself, I have to wonder why
| the hell you're reading a book in the first place.
|
| If you need that visual component, just watch a movie or read a
| comic book or something. This isn't a slight against movies or
| comics! They're fantastic mediums and are able to utilize their
| visual element to communicate ideas in ways that books can
| struggle with. And these visuals will form a much more cohesive
| artistic vision then whatever an AI outputs, since they're an
| integrated and intentional part of the work.
| smusamashah wrote:
| You make it sound like an unfair wish, which I would say is
| unfair itself. I like the book, I like visualising it in my
| head, and I fantasise what a scene would look like. AI won't
| generate a _true_ visual. It 's all fantasy anyway and AI can
| actually do it the way I have it in my head. It will solidify
| that thought.
|
| For this book in particular, I read the comic version and I
| didn't like the visuals very much. I have a different idea of
| babel fish. Vogons look different. I would love to see the
| visual that's in my head on paper.
| nowittyusername wrote:
| There is circumstantial evidence out there that 4o image
| manipulation isn't done within the 4o image generator in one shot
| but is a workflow done by an agentic system. Meaning this, user
| inputs prompt "create an image with no elephants in the room" >
| prompt goes to an llm which preprocesses the human prompt >
| outputs a a prompt that it knows works withing this image
| generator well > create an image of a room > and that llm
| processed prompt is sent to the image generator. Same happens
| with edits but a lot more complicated, meaning function calling
| tools are involved with many layers of edits being done behind
| the scenes. Try it yourself, take an image, send it it, and have
| the 4o edit it for you in some way, then ask it to edit again,
| and again, and so on. you will notice noticeable sepia filter
| being applied every edit, and the image ends up more and more
| sepia toned with more edits. This is because in the workflow that
| is one of the steps that is naively applied without consideration
| of multi edit possibility. If this was a one shot solution where
| editing is done within 4o image model by itself, the sepia
| problem wouldn't be there.
| renewiltord wrote:
| The prompt enrichment thing is pretty standard. Everyone does
| that bit, though some make it user-visible. On Grok it used to
| populate to the frontend via the download name on the image.
| The image editing is interesting.
| genewitch wrote:
| All the stable diffusion software I've used names the files
| after some form of the prompt, and probably because SD
| weights the first tokens higher than the last tokens,
| probably as a side effect of the way the CLIP/BLIP works.
|
| I doubt any of these companies have rolled their own
| interface to stable diffusion / transformers. It's copy and
| paste from huggingface all the way down.
|
| I'm still waiting for a confirmed Diffusion Language Model to
| be released as gguf that works with llama.cpp
| danielbln wrote:
| Auto1111 and co are using the prompt in the filename
| because it's convenient, not due to some inherent CLIP
| mechanism.
|
| If you think that companies like OpenAI (for all the
| criticisms they deserve) don't use their own inference
| harness and image models I have a bridge to sell to you.
| genewitch wrote:
| i give less weight to your opinion than my own. I'm not
| sure how you misunderstood what i said about clip/blip,
| as well. I was replying to a comment about "populating
| the front end with the filename" - the _first tokens are
| weighted higher in the resulting image than the later
| tokens_. And therefore, if you prompt correctly, the
| filenames will be a very accurate description of the
| image. Especially danbooru style, you can just split on
| space and use them as tags, for all practical purposes.
|
| I guess the "convenience" just happened to get ported
| over from "Auto1111", or it's a coincidence, or
| diggan wrote:
| > There is circumstantial evidence out there that 4o image
| manipulation isn't done within the 4o image generator in one
| shot
|
| I thought this was obvious? At least from the first time (and
| only time) I used it, you can clearly see that it's not just
| creating one image based on the prompt, but instead it first
| creates a canvas for everything to fit into, then it generates
| piece by piece, with some coordinator deciding the workflow.
|
| Don't think we need evidence either way when it's so obvious
| from using it and what you can see while it generates the
| "collage" of images.
| Voloskaya wrote:
| > This is because in the workflow that is one of the steps that
| is naively applied without consideration of multi edit
| possibility.
|
| Unconvinced by that tbh. This could simply be a bias with the
| encoder/decoder or the model itself, many image generation
| models showed behaviour like this. Also unsure why a sepia
| filter would always be applied if it was a workflow, what's the
| point of this?
|
| Personally, I don't believe this is just an agentic workflow.
| Agentic workflows can't really do anything a human couln't do
| manually, they just make the process much faster. I spent 2
| years working with image models, specifically around
| controllability of the output, and there is just no way of
| getting this kind of edits with a regular diffusion model just
| through smarter prompting or other tricks. So I don't see how
| an agentic workflow would help.
|
| I think you can only get there via a true multimodal model.
| nialv7 wrote:
| None of your observations say anything about how these images
| are generated one way or another.
|
| The only thing we currently have to go off of is OpenAI's own
| words, which claims the images are generated by a single
| multimodal model autoregressively, and I don't think they are
| lying.
| pclmulqdq wrote:
| Generated autoregressively and generated in one shot are not
| the same. There is a possibility that there is a feedback
| loop here. Personally, I wouldn't be surprised if there was a
| small one, but not nearly the complex agentic workflow that
| OP may be thinking of.
| vunderba wrote:
| As somebody who actually tried to build a multimodal stable
| diffusion chat agent about a year back using YOLO to build
| partial masks for adjustments via inpainting, dynamic
| controlnets, and a whole host of other things, I _highly doubt_
| that it 's as simple as an agentic process.
|
| Using the prompt to detect and choose the most appropriate
| model checkpoint and LoRa(s) along with rewriting a prompt to
| most appropriately suit the chosen model has been pretty bog
| standard for a long time now.
| echelon wrote:
| > Using the prompt to detect and choose the most appropriate
| model checkpoint and LoRa(s) along with rewriting a prompt to
| most appropriately suit the chosen model has been pretty bog
| standard for a long time now.
|
| Which players are doing this? I haven't heard of this
| approach at all.
|
| Most artistic interfaces want you to visually select a style
| (LoRA, Midjourney sref, etc.) and will load these under the
| hood. But it's explicit behavior controlled by the user.
| lawlessone wrote:
| huh, i wa thinking myself based on how it looked that it was
| doing layers too. The blurred backgrounds with sharp cartoon
| characters in front are what made me think this is how they do
| it.
| justinator wrote:
| But the annotations are still wrong,
|
| https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_...
|
| (nice URL btw)
|
| The room, the door, the ceiling are all of a scale to fit many
| sizes of elephants.
| lupusreal wrote:
| The image annotated to explain why no elephants are possible is
| very amusing.
|
| To me, this kind of image generation isn't very interesting for
| creating final products, but is _extremely_ useful for
| communicating design intent to other people when collaborating on
| large creative projects. Previously I used crude "ms paint"
| sketches for this, which was much more tedious and less
| effective.
| ziofill wrote:
| I usually agree with most of Gary Marcus' points, but I'd really
| like to hear his take on this. One of his examples is that "the
| system can't generate a horse riding an astronaut" and in fact I
| tried a lot in the past but it would always draw the astronaut on
| top of the horse. Well, here is the result now:
| https://postimg.cc/QFtRjbHM
| jsheard wrote:
| Whenever one of these well known gotcha prompts gets "solved"
| the question is always whether they actually solved the
| underlying reason it used to fail, or did they just have a
| bunch of third-world workers tag pictures of horses and
| astronauts until the model started handling that specific
| example more reliably. As the saying goes, every measure which
| becomes a target becomes a bad measure.
| pixl97 wrote:
| Well, what you do is try examples close to the problem space
| around it. For example a fiddler crab riding a chameleon
|
| https://sora.com/g/gen_01jrbq91wtefjtpb8ceajdh9mt
|
| and then iterate around other combinations to see if it's
| generalized or not.
| DonHopkins wrote:
| Q: How do you know if there's an elephant hiding under your bed?
|
| A: Your face is pressed up against the ceiling!
| lou1306 wrote:
| Putting my Wittgenstein hat on: How can I ever be sure that the
| machine is not generating an incredibly tiny elephant, maybe
| hidden under the sofa?
| cadamsdotcom wrote:
| Anyone stuck claiming AI isn't useful - there are so many useful
| things it can now do. With text that makes sense you can generate
| invitations for your next picnic. That wasn't possible mere weeks
| ago.
|
| Wonderful to be alive for these step changes in human capability.
| klik99 wrote:
| I've seen a few YouTube thumbnail generation examples on Reddit
| (I'm on vacation so not gonna search for a link) that show
| multimodal with inline text giving specific instructions. It's
| impressed me in a way that I haven't been with LLMs for 2 years,
| IE it's not just getting better at what it already does, but a
| totally new and intuitive way of working with generative AI.
|
| My understanding is it's a meta-LLM approach, using multiple
| models and having them interact. I feel like it's also evidence
| that OpenAI is not seriously pursuing AGI (just my opinion, I
| know there's some on here who would aggressively disagree), but
| rather market use cases. It feels like an acceptance that any
| given model, at least now, has its own limitations but can get
| more useful in combination.
| 1970-01-01 wrote:
| For the first 9 years of an elephant's life, it can easily walk
| into that room. I don't find this to be a breakthrough. I find it
| to be clickbait.
| m4thfr34k wrote:
| I am very impressed with the current image generators out there,
| 4o / Leonardo / etc., but I cannot wait until they include some
| step to actually "check their work". Ask it to produce a watch
| with the time of 6:37. It fails every time, because almost all
| watch photos out there are set to a specific time, and seems like
| something an initial "did I do this right" check could confirm.
| The time example is trivial but a general "does this output
| actually make sense considering what the user asked" checked
| would be tremendously valuable.
___________________________________________________________________
(page generated 2025-04-08 23:00 UTC)