You don't own your AI-generated content
Вставка
- Опубліковано 4 жов 2024
- AI-generated content is still not copyrightable. I discuss the status of AI and Copyright and try to explain why it works this way.
Thanks for all of your support! Please consider supporting on Patreon.com/ljfrench, through UA-cam membership, or Floatplane subscription.
April supporters:
"Counselor" Level:
Eevi, Ugly Grill, torpedan, PureMagma, Eric Tams, Gutbrodj, TechTechPotato, King Aires, Kyle Siefring
"Advocate" Level:
Arron Washington, Keith Marrocco, Dustin Rodriguez, Josué Vicioso, Brian Flowers, tdis8629, amelius, Meaningless, Aethero Toland, John Swanson, matthew beller, Priscilla Astling, Schi-Schi, EvoGamer, Kreachers, admalledd, khagerou, Dawn Massaro Guy, wrozmo, ChessWithADHD, robfrawley, Medeka, David Zaslavsky, Christian Ullrich, Brian Roush, Rudolph Bescherer Jr, Firstname Mclastname, JH, Cindy Campbell, Tron Bårdgård, Mark Randall, Sarah Gerweck, Matthew East, Mario Bonales, Ian McDonald, Marcus Agehall, Jonathan Robillard, Amanda Gillies, Justin Waddell, Nathaniel, Tony Cruickshank, Kasaryth, Oisin Creaner, Stefan Persson, CombatZAK, Alexander Sihn, Andrew Reid, Gregory Ford, Brody Eastwood, sehro, Dzyan, Chris Lindsay, Rico Robbins, Kai Raphahn, Andrew "FastLizard4" Adams, Kean, Jessica Pearson, LbxAni, Daniel Ducharme, Ph.D., majikthise, foonix, TheEuphoGuy, Brian, Matthew Bertrand, Jack Draak, Eye_Make_Stuff, Scott F. Comstock, James Melanson, Kat Willhite, itheoryon, DreamerDon, Alan Nise, anton.molyboha, The Disturbed Angel, Travus, Eric Woodley, Raindrop Works, Logan Stromberg, scj643, eyebrows360, Chris Connett, Haplo, Michael Ciesielski, Chris Hilliard, Greg, Q Squared, Jason Glaesemann, Eric Barker, Gef the Mongoose, keeping up with the Devilment on weekends, toadbear, DrJKL, Lawrence Groupe, Michael, DrakeDT, Katrina Middleton, Andrew Venier, Bill Tonnies, Rabid Ronin, Thomas Dinsdale-Young, Zendane, varia, jag1110, Bill Somerville, Mike Dunford, AnnoyedRook, cwestpha, Peter Huston, James Oxford, rcmaehl, Deleted, Evan Davis, Michael Stokes-Byrne, c, Jorge Vittes, Nate Gray, Tony Webster, ChaoQueen, Ross, Dustin Bosveld, Tymorus, Bryan Lubeck, Rodney Nelson, jak_ub, Thedougler303, WarrantyVoid, PinkFluffyTeddyBear, Rovert09, Ivan Chepurnyi, Rob Voisey, Kris Hetzel, Negligibly Negative, Maarten Wisman, RunsWhenChased, ricin, Christina B, Robert Z, Fubar, sean murphy, OsculatingPlane, Landon Noll, Justin Bassett-Green, Ryan Schott, RipeSloth, CostlyFiddle, Jennifer and David Brock, void, RenyPetty, Blueberry Hill, Andrew, Aaron Harvey, Learethak, Emily, Dave Meyers, jsteelkw, TheAmazingDave, Zoonist, XPEric, balayanr, zensalad, Wolfendragoon, Nova_, MeDeOgre, Kantorock, Ava-, Deleted, jSON, Jabberwoky69, Deleted
Welcome back to UA-cam Leonard! We've missed you!
Very much so
Seconded! thirded! and so on!
Fourth-ed
Glad to see this channel making a come back
Making a comeback? Did something happen?
@@lermanct4486 A multi-month hiatus, apparently.
@LermanCT hasn't posted on UA-cam for a long time. I thought the last video was just an update on why he hasn't posted due to being in a legal battle.
He was still active on twitch though. I highly recommend watching the Sunday live stream.
@@daveubermensch I've seen other youtube creators say their revenue was destroyed again and again for no discernable reason, some left for Twitch
An analogy I like is "ordering a pizza on the phone does not make you a chef. It does not matter how specific your directions on toppings are, you can not claim you made the pizza."
What about large corporations who hire artists but hold copyright to the works. This is OK because the artists grant the company copyright. But the a.i could also grant the user of the a.i copyright. This isn't fair and it's the same thing. So basically it's only OK for fat cat corporations to do it but not the people.
@@martymcfly6411 You kind of answered your own point.
Large corporations HIRED artists. The artists were paid.
A.I doesn't pay the artists it copies from and neither do the people who use A.I.
I generally do favour this decision, as it makes a lot more practical sense, imagine if the creator of a voronoi noise texture generator owned any output it produced, just about all procedural texturing would become a legal minefield.
voronoy died over 150 years ago, thats why.
Could you imagine if every map generated by civ6 was covered copyright?
It's actually dumb. It's a tool. in this case an computer, directed by a human, with outputs selected by a human. In what way is this different than a person using Photoshop, a computer directed by a human, with outputs selected by a human? Or a paint brush for that matter? Whether something can copywriten is now dependant on a perceived level of effort? That 'artist' using photoshop didn't build the computer, write the software for the operating system, write the software for Photoshop, build the mouse or keyboard or monitor, but that's different than using a slightly different piece of software they didn't write!
@@petenell5807I think you may have things backwards, it's not that an artist can't create and licence something using AI, its that the result of the AI can't be claimed by the creator of the AI (or the AI itself) as their copyright.
That means the produced output of the AI is, by itself, without any licencing restrictions and can't _be_ licenced unless said artist then goes and does something with it.
Copyright law has always required at least some evidence of creative effort on the artist's part. If the artist does anything more than simply copy-pasting the AI's output, such as compositing, framing, alteration etc, then they cwn claim copyright over the finsl work eiyhout issue.
@@petenell5807 That is a very compelling argument and i just wanna say thank you, because not many people think this way!
That being said, i think what most have problem with is that when you are building a model, you have to feed it data, and that data is usually other's people work...*without consent*
The consent part is where it lies the problem
Copyright with AI (neural networks) is where it is at, because data has been taken from places it didn't had authority from.
Yes, one could argue that the data was public, and it is in fact no different than looking at an image for reference, and the same as such if you were to copy, paste... then edit to a high extend someone's work.
You could argue that if your work is so undistinguishable from someone else it is not their work anymore.
Same with AI, but apparently not to the eyes of everyone...
I remember when i was really young, and would never paint using reference. but would recall from my mind, after years, realized my mind is also reference, and it is no different than straight up copying it to some degree.
There is a huge gray area that isn't spoken nor agreed and AI has shoot an arrow straight to it.
To add a wrinkle in this, there are AI tools that are being integrated into drawing software. So you draw the image, and the texture gets elaborated to a certain extent. This process goes back and froth as you continue to draw and refine the detail, to the point where neither the AI or the artist alone could be said to have created the image.
That has human-user input involved, which would fall under copyrightable (probably). There's no ability to distill that image down to "prompt + seed" the way raw generated images are (and for sufficiently detailed prompts, you could probably copyright the prompt).
@@majorjohnson8001 Let's explore the minimum influence.
In Stable Diffusion, there are things called ControlNets. These allow you to enter another image, drawing, or for this example, a scribble to direct and control the AI.
By using a simple child scribble as a ControlNet, and that being elaborated into a complex piece, is that enough human input?
@@nilok7 still pretty much lottery machine, that would be similar commission with reference sheets and direction. 3D artists who do exo-skeletons for their character does not receive © from those skeleton, only the manual character design. Basically, commission with extra steps
Consider CAD apps that know all the local or regional UCC/builder Codes and tidy up the drawings for the architect or DIY designer. Would US LIBCONG say that AI-driven cleanup of rough sketches on berms/backfill, the piers/footings, soffits/fasciae, foundations, roofs, studs, and more no longer get copyright protection?
Ponder FEA/FEM/CFD ship design and structural planning software. At some point, principals will opt for AI-based rapid ship design once viable, and junior designers and CAD specialists may lose work. Will govs politically refuse to allow design rights for firms that accumulate cash but that only hire minimally?
For over 20 years now, and especially consider recent Orca, Grasshopper, and Nemo, and other plugins for Rhino, it's been possible for a Naval Architect to tell the software "show me options for a destroyer/cruiser hull of 120-160m length, 17-23m beam, 6-7.6 m draft, 17-31 m depth at planned draft, total hull resistance of 30-80 MW, displacement of 9700-14200 tons combat load, and fuel economy of 6,000-10,000 nm, carrying a crew of 195-339, with a food/spares/potable water capacity/ endurance of 2 months, and these sorts of weapons and sensors..."
Surely, LIBCONG would be out of its mind to say NAs can't obtain copyright rights and due protections just fot incorporating AI intor their workflows. It would be sheer madness.
@@davidsyes5970 You would still own the copyrights to your original sketches. The work that YOU do is YOURS. The work that the AI does is not yours. If the AI's work is transformative, then you do not have copyrights. If the AI's work is derivative, then the AI's work is burdened by your original copyrights (just the same as if a human had made a derivative work based off of your own original work).
started watching ever since the Fine Bros. wanted to trademark the word React .
so crazy seeing you on my feed again talking about copyright topics ! hope to see some more informative videos . great to see you much more active .
Really good analogy, because when you replace the sketch artist with a robot doing the sketching, it really sinks in that you still aren't the one doing the creation. Although I could see it becoming more blurry when like Adobe Photoshop will be giving more artistic control, so it starts to become more of an assistant tool to create your vision than something doing everything for you.
Yes, but if you built the robot, in order to paint images for you, then you should be able to copyright those images. What about the people that wrote the AI algorithms and built the AI??
I'm not bothered by no copyright by AI software. What would bother me is if the owner of the AI software is the copyright owner of anything people generate with it.
Wouldn't be the same (kinda?) an engine for a game for example?. The creator of the engine owns the copyright of the engine, I don't know if it owns the copyright of everything created with it. O.O
@@ubiergo1978 Game engines are probably more contract law over copyright. From what I've seen more like buying a guitar to play on a song, or licensing assets to use in a game where the fee you paid is what gives you permission to use it.
The software creators usually retain copyright of all outputs. It's in the terms of service for Midjourney and Stable Diffusion.
Would love to see you cover the Nexon vs Ironmace lawsuit over Dark & Darker.
Some youtubers are supporting Ironmace to the point of seeding their Beta Test Torrent while others like Kira think Ironmace might be in big trouble.
There is A LOT of misinformation about the lawsuit and a LOT of misunderstanding about copyright, too. Kira had some decent points despite not being a lawyer himself, and while I love Dark and Darker, a lot of the fandom is siding with them simply because they like the game, but are missing a lot of the forest for the trees. Many thought the DMCA takedown request WAS the lawsuit itself, and were arguing it was dumb, when in reality it was just red tape to begin legal action.
Leonard's input would be much appreciated to give a fresh perspective.
@@digitalarsenal5234 the problem is nexon is claiming stolen content that was purchased from unreal marketplace... so little of what their claim is has any validity
@@brykanst9071 They never claimed that. If you read the claims of the suit, it is arguing evidence around the fact that the work as a whole is infringing, not individual assets. That is one of the HUGE pieces of misinformation many content creators are spreading
I see it like this:
AI makes composite pictures
AI does not, and cannot create art.
I saw 'AI' in the title, and came running. This is something I've seen discussed all over Twitter and your take on it was interesting.
One aspect that everyone misses is that the underlying generative model was created by going through a curated set of images that were categorized. What is then generated through use of prompts is very often a merge of features of the seed images.
So, if the model just merged several images, that may or may not be copyrighted, into a new one, then how can you claim copyright for the result. And this will be a bigger issue when people start to notice lack of variability in the generated images. And it becomes easier with time to track the seed images that contributed to the outcome.
Also... When I see the generated images it seems very often like looking into the best examples on DeviantArt mashed together. I suspect that these models's learning datasets ripoff a lot of anonymous artists that are on DeviantArt, Pinterest, Flickr and others. So, claiming copyright on generated content might be a much bigger can of worms than you might think.
AI generation isn't a collage. It's a common interpretation that folks like to push but it isn't. It sounds like you've mostly looked at a curated sample of AI generated content and not at the breadth of content it can generate. It's like going to the Las Vegas eiffel tower experience and saying it's a derivative of Paris. That was the point. Same thing with the examples you've looked at. They were made to look like that because that's the style they wanted. Style isn't copyrightable.
@@Trahloc No merging is not the same thing as collage, like reproduction is not the same thing as cloning.
Definitions matter in life, and in baiting.
@@calexico66 Very well, do you see this technology as just a more advanced version of the morphing tech that was popular in the 90s such as Michael Jackson's Black and White? If I continue to misunderstand your usage of the word merge, can you point to an example of the technique you believe the AI is doing that isn't from another AI?
@@Trahloc to create an image of Will Smith you need to add that information to the model, to make him slap someone you will need to add images of people being slapped in in several positions that can be matched within a vector space with Mr Smith images that are on a similar vector space. For the model to extrapolate that prompt it needs to have been fed images that are properly categorized, unless it doesn't have and simply matches whatever vectors cluster together with the words. But if you want accurate results then the reality is that you will have a degree of over fitting, and that means that a percentage of the original image will stand out. This "AI" does not have imagination, it only extrapolates by merging images that are on categories clustering on vector space derived from a learning dataset. And stop baiting.
@@calexico66 Folks have mixed different vector space sizes with success. Breaking down the training process isn't a slam dunk support of your position. All you did was prove it needs to be trained to be able to do what you specifically want it to do with ease not that it can't do it at all without that training. You didn't prove it can't do things outside it's dataset. TI's are basically advanced prompts, they don't alter the model itself which proves that, in theory, you can prompt a photo of Will Smith eating a pineapple without ever having a photo of Will Smith or a pineapple if you know what you're doing well enough. This is why I asked you to explain what methodology you believe the AI is doing *without* using AI otherwise we enter a circular argument.
Always happy to see new LJF videos!
Welcome back. Happy to have your videos show up on my notification again xD
Wonderful and so glad to see you back once again...
Welcome back! Would love to see you comment on the Dungeons and Dragons OGL debacle that happened a while back while you weren't posting!
Good to see your videos in my feed again!
The Naruto case makes this quite clear and simple - if you are assisted by an entity not recognised by copyright, you only get copyright over the portion of the work created by you.
Wow, this is very timely. Right now, there's a strong trend among You Tube content creators to use automation to generate their content, using A.I. to write a script, and then a good text to speech synthesizer to read it, and just slap that over some creative commons video. This is potential powder keg.
I cannot see how this is a "powder keg" in any way.
@@chaosmonkey1595 There's a huge volume of content already, that could be copied frame for frame. This technique is also being used for certain categories of self published books on Amazon. Amazon could say their automation has determined that no copyright can exist on those works, and stop paying royalties.
Oh snap. Haven't had me some LeoFrench for a hot minute! Glad to see you're still crackin son
One of the tangles that comes into copyright is that it is covered by international agreements, most famously the Berne Convention. There are items in the Public Domain in the USA which are in copyright elsewhere.
That sort of problem was mostly eliminated by changes in US law, but there are some duration differences. Changing the law to protect AI-generated works could get very awkward. I have a feeling that too many of the Americans involved in tech issues have a flawed appreciation of International matters.
For most practical purposes it doesn't matter except for international commercial use. In the US, only US copyright law applies even to works originating elsewhere (this is how the Berne Convention is set up). The US treats all works as the would be treated if copyrighted in the US (a few sound recording treaties separate from the Berne Convention excepted). So unless you are using works internationally, following the US laws are fine. Even then, unless you are doing business in a country where the copyright is significantly different, its difficult for their courts to get jurisdiction over someone located in the US (as it would be the other way around). The internet has made this easier to stumble over, but the courts have general required some evidence that the person charged with infringement had some intention for the potentially infringing work to become available in the country to gain jurisdiction, as well as have some interest in the country (money, frequent travel, etc). The easiest way to bring the case is to go to the country of the infringer (or another country where there is substantial interest) to sue. This make learning the ins and outs of every countries copyright law not very useful to the average American. Due to the nature of UA-cam, Facebook, etc.. being officially located in the US most users of these platforms have interest in the US and should at least make sure they are aware of the relevant US copyrights, and I would say the same of anyone in the US using similar products located in other countries (I don't use any so don't have examples, maybe TikTok, that one has always been unclear) and learn their laws
Is there some kind of penalty for not disclosing that an image was created by AI? What stops someone from saying "Hey I'm doing open commissions!" and then just generating the image as requested in DALL-E or some other similar AI? Just because something isn't copyright-able doesn't mean it can't be sold, as you said in this video, but at the same time every image generated by the AI should be unique, so the control of the image would belong only to the person who generated it until it's sold. Maybe some trouble could arise if the purchaser then tries to use the image in a copyrighted work like that graphic novel, but if it's never disclosed that the image is AI generated, how would anyone even know? Just seems like a totally unenforceable ruling IMO.
As time goes on, it will definitely be commonplace for a commissioner to ask whether or not an image is AI generated before they purchase it. Uninformed people purchasing AI generated images pretty much only happened because they were unaware AI images were even a thing, or that AI images are treated differently from human created ones. In terms of commission work. A costumer will likely ask for confirmation that the work isn't created by AI, so not disclosing anything will likely not even work in the future.
With the power that media corporations have I suspect that THEIR ownership of AI generated works will come to pass. AI generated scripts, characters, maybe even voice work - not having to pay ANYBODY for their time and talent - HAS to be enticing.
I have been using AI for writing and I hate to say but they are much better than hiring someone.
theres also the route of proving the acusation if they dont disclose its ai generated how would one know?
Welcome back!! Glad to see you making content again(I just watched the video about LIE-Bold-Weaks trying a DMCA).
I'd love to Leonard talk about the Dark & Darker craziness.
Second that! Especially with the US federal lawsuit just filed by Nexon against Ironmace, and two of its principals.
This is why I always, ALWAYS, utilize my own creative faculties in every AI artwork I create. I keep copies of every step in the process, and I don't rely 100% on the prompt systems, I also utilize photoshop and other tools in every piece. This guarantees that A) nobody could ever re-create my exact image on their own (with the AI), and B) I have inseparable portions of the image that are human created, forcing the image to be taken as a whole as copyrightable.
This is the way.
It’s still made by a computer ^_^ expedites all the work us photographers take hours of time doing.
@@Pixelkip Yup. Things are going to change. Everyone better embrace it. New revolutionary tech always changes things fundamentally. Lots of photography related jobs are going to disappear.
@@user-lv6rn9cf8m lol nah.. eventually people will appreciate artists more than ever. Ai can’t feel and for that reason there’s no reason to give it any artistic value
Nah I'll still steal it if I find out it was A.I. generated I'm taking it nothing anyone can do since it can't be copyrighted another one for the pile.
Welcome back, Leonard!
You're my favorite copyright attorney!
Confirmed, Leonard wants people to break the speed limit!
Good to see you back!
2:30 - "Of course you can drive your car over the speed limit, and you can do it safely to a point."
As a German, i have to agree.
So happy to see you back
Cameras would be an argument against the “needing to draw it yourself” argument.
Please do a video on Nexon vs IRONMACE over Dark and Darker.. There is so much mixed information going on right now and you're the best at getting to the important stuff
IIRC midjourney's EULA states the user retains the rights to the images they generate. Edit: premium users can use their images commercially
lmao that doesnt mean shit, thats just their own end user agreement they can just say whatever they want
@@WeaponsGradeNeurofunk In the absence of any previous legal pronouncements, it was a perfectly reasonable thing is them to put in their EULA, but yeah, this basically means that they're EULA isn't worth anything as far as rights go now. And I don't feel at all sad about that!
Anyone who pays Midjourney for the service retains the rights, if you are an employee of a $1m revenue service, you must buy a pro license
Good to see ya back Leo
2:30 Everywhere I've lived, 90-99.9% of the speed limits have been essentially random. In my current city, if I drive 5-10 mph over the limit, I pass no one except people going well _under_ the limit (old people and work trucks, that sort of thing) and get passed by half the traffic I see, yet accidents are still rare.
This is because speed limits in most jurisdictions are not based on careful calculations of the road design, they're based on blanket laws, fiat, or petitions by busybodies, and almost never get raised, only lowered.
I saw this happen when, as a teenage Boy Scout, I attended a city council meeting. A woman stepped up to the podium and requested that a stretch of road have its speed limit lowered from 40 to 35. The council didn't even question it, they just unanimously voted to do so.
The most egregious example is a stretch of road outside my current city. The state highway it's part of has a speed limit of 55, though no one goes less than 65 and many go 75 (plus the occasional maniac going even faster). However, for a stretch of about 2 miles, the limit drops suddenly to 35 mph. The road is no different, curvier, or more dangerous in any way. Visibility is actually better on that stretch than most of the surrounding area. There is no visible reason to slow down by 20 mph, let alone the 30-40 over the limit that most people drive, so no one does. I drove that road to and from work for a year and never saw a single accident or sign of one, like broken glass, on it. I never saw police speed traps, either, though I saw many police _cars_ as they passed me while I was going 65.
The real kicker is that the speed limit only changes when the road crosses into another municipality, where it goes past neighborhoods, through multiple traffic lights, and becomes very hilly, whereupon the speed limit goes _up_ to 45. And to further prove my point that speed limits are random, if you go any faster than 35 in this 45 zone, there is a dip that _will_ bottom out your vehicle unless it has a _really_ high clearance, like a lifted truck.
Sometimes copyright makes my head hurt
I asked Bard "Who owns the copyright to a story created by Bard?" Here is its answer....
Google AI owns the copyright to any story created by Bard. This is because Bard is a machine learning model that is trained on a massive dataset of text and code. The text and code that Bard is trained on is owned by Google AI, and therefore, Google AI owns the copyright to any creative content that Bard generates.
However, Google AI has a policy of allowing users to share and use creative content that is generated by Bard, as long as the user gives credit to Google AI. This means that you are free to share stories that you create with Bard, as long as you include a credit to Google AI.
For example, you could share a story that you created with Bard on social media, and include a credit to Google AI in the post. You could also include a link to the Google AI website in the post.
By sharing stories that you create with Bard, you can help to promote the development of this powerful new technology. You can also help to inspire others to use Bard to create their own creative content.
Its good that the current ruling is that you can't copyright AI generated works, because it will take the monetary incentives away from trying to use AI works, which should go a long way in preventing human artists from being replaced by AI programs and those who seek to profit from their use.
We shall see.
far far from it. they will simply tie the works to their trademark or other things making the AI generated thing theirs in a legal manner. just watch disney, it will happen.
@@pulsefel9210 I hope you're wrong in the long run, but unfortunately you probably foretell exactly what will happen . Never put it past the greedy to find a workaround for their malignant behavior
@@interstellarsurfer Indeed
@@MaxterandKiwiKing we can pray
The court saying something is right or wrong doesn't change the 'Law', the law is written by the legislative branch. So the copyright office, the executive, could actually follow the law regardless of what the courts say.
Glad you are back. Your mic is tracking very hummy noises when you talk. Maybe you need to use it more often ;)
Do bear in mind that if you sufficiently modify the image after, or use the image as part of another image, the resulting creation can still be protected by copyright.
The recent release was also rather nebulous on how much human input was needed in the final product. Would ScribbleDiffusion require enough human input, would Ebsynth's Style Transfer be too little, etc.?
As far as I understand the law, giving input in general is just not enough to be called the author. Copyright generally does not protect ideas, but instead execution.
Modifying an image would add some protections to an image, but it does not make the unprotected AI generated bits now protectable.
It would depend on the edits but would probably be looked at cutting up public domain photos to make a collage.
It's all very much a case by case basis thing.
No copyright for AI "art": Good.
I find this great news. Eventually we'll have to wonder copyright as AI content becomes more indistinguishable from human content.
At the least, usually human creation has a paper trail. AI only ever gives you a final PNG. Where a human artist can show the art program file, or concept sketches, or just film themselves making it from scratch.
While most of the media attention on this subject involves visual art, what are the legal effects for more practical things like the circuit boards used in electronic products? A circuit board holds electronic components and has lines drawn in copper that connect the pins of those components according to a human-created specification called a netlist. In the old days of the 1980's I would hire a specially-trained artist called a circuit board designer to draw the circuit lines out on a clear plastic sheet using narrow lines of tape, with the condition that I owned the copyright to the artwork. This was black-and-white (clear?) copyright law. As technology advanced, the plastic-and-tape was replaced by a computer drafting program, in the same way that Photoshop and Illustrator allowed artists to simulate brushes and paints with a computer. Still, all of the marks were placed in the artwork under direct human control, so no copyright issues.
As electronic circuits have advanced, the size and complexity of these circuit boards has grown to the point that it is very difficult and time-consuming to create a design that meets all of the requirements. There are now multiple companies working on AI-based circuit board design tools. If I design a board using one of these tools, is it copyrightable, and if so to what extent?
I doubt it was ever challenged in court, but it is likely there was no copyright to be had in circuit board layouts. Circuit boards are functional and with little exception the layouts would be chosen for a functional purpose (excluding any aesthetic portion like branding which would be copyrightable), and would not fall under copyright. They would fall under both trade secrets and patent law. Going to a simpler example, you couldn't copyright a novel gear tooth shape, but you could patent it. Software gets over this problem due to a specific act of congress expanding copyright protection to otherwise un-copyrightable functional code.
What I'm curious about isn't the copyright of raw outputs of an AI but composites, img2img, and combinations of the two.
How much work do you have to put into generating hundreds of outputs, curating the outputs, taking those and editing them together to fit your vision, possibly running that through img2img to blend them, and repeat before it becomes your work.
How much work? No one knows. The legal world is going to fall off the buggy worse than ever. They will not be able to catch up to the AI technology regardless of how much they try. Courts and laws take years or a decade to sort these things out and there just won't be that kind of time available in the immediate future. It will be a wild next 10 years. A lot of laws are going to become toothless.
I hope this issue makes more people realize that "content" is not property, nor is it "ownable". Those are simply metaphors that simplify the intricate (even arcane, even cryptic) rules that have evolved about who is allowed to learn and share what.
The term content creator always felt hollow to me, even I get folk were looking for a catch-all term.
More accurately theyre supposed to be either a vlogger, podcaster, show host, filmmaker, essayist, animator, performer etc.
Please cover Nexon vs. Ironmace - we would all love to see your take on it
2:45 - It's been demonstrated through statistics/practical studies that _without_ "speeders and slow-pokes" everyone going the speed limit would cause _horrendous_ bunching/traffic jams... _(it also had something to do with improving how data packets are grouped and sent/received)_
I'm confused, so if I create a Art database of Art that I created myself and then ran it through an AI, would the resulting image not be protected even though all elements in the generated image was derived from all of my own artwork?
He answers that question in the Disney hypothetical. 5:44
I guess the big question I have is when does it go from a tool used to bring out the creative expression of the human to something created by the human.
If I go to an AI and instruct it to create "an image of a red sports car flying through the air after going off a ramp while a blue motorcycle is in the air above it" there is clearly a significant portion of the creative effort in the product is being done by me rather than by the AI generator.
How about animations where you provide the key frames and software fills in the intervening frames. Most of the animated movie is now not copyrightable as being generated content.
I can see them arguing that the key frames are part of the work, where as providing an AI with a prompt & guideline - those do not make up any part of the final work. So we just need to provide part of the prompt & guideline in the same medium as the AI will output - text for a book, a brushstroke for artwork, and if the same rules for animation applies, then we can copyright AI assisted works.
Simple solution which I use:
Use the output as inspiration. Change it and make it your own. Use it like a brush or tool. Dont just say “i own this line from this brush” but use it to paint a picture that wouldnt exist without you!
hoooray!! you're back
I wonder how long it will be before IP infringement cases come up against prompted AI generated artwork. It seems that where copyright laws don’t apply, IP laws might be another way for creators to win against their original creations being used without their consent. Al they have to do is prove they are the originator of the concept prior to the event of the infringement. It seems almost ethereal.
I think it's going to take Congress actually creating new legislation before we see any significant action on this, one way or another... The existing laws simply didn't consider non-human creative expression with the scope of modern AI.
Another boulder in the machine is other countries. If the US have some laws about this, what happens if the server with the AI is in another country that recognize the AI as an individual, like Saudi Arabia did with an Android, or India with Dolphins ? Or the other way around, the server is in China or Russia, and all products belong to the state. ( Yes, i am hypeboling here, but it is for a not yet possible scenario. )
@@roberthpilesund384 Not super relevant to U.S. enforcement, really... If there's no relevant law a court *MIGHT* look to another country to see what they are doing... But my understanding is that it has no real binding force within U.S. courts unless it's been formally agreed to between the governments, and that's how current international copyright enforcement works. (In theory, at least; once you involve enough money and/or politics, theory often ceases to matter.) Somehow, I doubt many U.S. diplomats have been foreward-thinking enough to have already brought it up and developed agreements, lol.
"Well, imagine my shock?" - Paul Joseph Watson
Driving above the speed limit is not in and of itself dangerous. The source of speed-related danger while driving is the delta between speeds.
Each road has its own natural speed limit. This is defined by how fast the majority of drivers would go if not given a limit. This is affected by how wide, straight, and level the road is, as well as what's along the sides of the road.
The most danger is actually when the speed limit is set below this natural limit. Setting a speed limit above the natural limit doesn't affect much - drivers still go the natural limit, on the whole.
These are the conclusions of many studies on traffic safety.
So the ChatGPT models are listed as having in the range of 117 million to 1.5 billion entries in them as source material.
We have roughly 4.3 million frames of MCU movies.
We have roughly 23,025,600 frames of Simpsons content.
Disney lists 61 films as being of Disney animation studios, very rough ballpark of 120 minutes per movie @24fps, for 10,540,800 frames.
If we take that combined mess of roughly 38 million images, make a model from that, then have that model generate a model of say 20 million images for a model of another AI, without a human ever seeing the images generated from the first AI.
Then the second model is 100% Disney owned content, but filtered through a bot first. Does Disney have any claim over the resulting content? Or even the model file?
If I use a painball gun to put paint onto a canvas and allow it to drip down the canvas, who created the image?
Me? The paintball gun? Gravity?
Isn't this the same as using an AI? I provide the AI with the paint, and basic guidelines on where to apply the paint and then let the AI to "fill in the gaps" - same as gravity dripping the paint down the canvas to increase coverage on the canvas.
Been Waiting for a Lawful Masses video on this topic since many seem to think AI will own everything. It's just a smiling primate with a borrowed camera
I'll be interested to see how this pans out practically. Ignoring glitchy hands, it's often not obvious whether an image is AI-generated. And barring watermarks, the image can be scrubbed of any provenance metadata suggesting AI. So what's to stop folks from just lying and claiming human authorship? Or hiring an artist to "trace" the AI work; then good luck determining who copied who. The "tracing" differences might be sufficient to attract a copyright regardless.
The real questions will come as those errors get fixed and the programs are improved to the point where we really CAN'T tell without investigative scanning software. Society and the law have an awful lot to prepare for.
That doesn't really matter in the long run (I think that's what the analogy of speeding was about). The copyrightability of a work generally only matters in a copyright lawsuit. People lie about creating works made by others and register them all the time. The only practical difference will be in a lawsuit one side will have a new argument that the purported copyright owner had the work generated with an AI, but they would still need to find and bring evidence. Otherwise this is just letting honest people know where they stand legally with the copyright office and AI generated work.
I'd be curious, if AI were to face swap a stunt double with that of an actor. Is that still copyrightable? What about AI touchups to images in movies, ads, magazines?
It seems to me that we are adopting a double standard for what is copyrightable when it comes to AI generation. It's all good if you are a major studio with millions to throw away, but a little webtoonist can't have the same protection.
Dude I've been waiting for you to weigh in on AI content. I wonder how you stand on content that is live generated by someone drawing that is then transformed in real time through AI and they alter their drawing in response to that output to adjust to their desired output. How is the AI not acting as simply an advanced paintbrush at that point?
Edit: this isn't theory, someone has already done this I just can't remember the video title to link it.
Follow-up question: Does this mean that all AI generated images are public domain? (if something can't be copyrighted, does it become public domain?)
Yes, but that doesn't really mean much, because there's still going to be someone else's creativity in the model. If the model can draw Iron Man, for example, then telling the model to draw Iron Man does not give you a public domain image.
There's been cases in the past, before the US moved to ownership granted upon creation, where TV companies forgot to register, say, season 3 of a TV show, making it public domain. Problem is, seasons 1 and 2 are still under copyright, and you cannot use season 3 without infringing on seasons 1 and 2. So it being public domain is meaningless.
This is also the whole motivating factor behind the "Sherlock Holmes' emotions are copyrighted" argument used to harass people making reuse of the public domain Sherlock stories.
The underlying principle is that copyright is *separable*, and each part of the work needs to be public domain before all of it can be. The safest kind of public domain reuse is that of works past the global cutoff date, since copyright currently respects notions of causality (so far, at least).
Not really, the terms of service for AIs often say that the images created are the property of the company. If you use the images without permission, it could be considered a TOS violation (contract law).
It will be interesting to see how this argument stands up when self-driving cars get pulled over for traffic infractions
@@darrylroederer2680 using the analogy he used, if you ask a taxi driver to drive somewhere and the driver speeds, the driver gets the ticket. If you ask a robot to drive somewhere, by this analogy, it should get the ticket. A machine can't get a ticket. Does that mean noone pays?
It's interesting to see how there are wide differences in how countries are handling AI, a month or two back i contacted the UK IPO about the status of ML and procedurally generated works and they responded with the following:
Thank you for your query.
Please be aware that the Copyright Enquiries service is only able to provide general advice regarding current UK copyright law and cannot provide legal advice regarding how the law should be interpreted in specific cases.
Copyright protects original literary, dramatic, musical and artistic works as well as films, sound recordings and broadcasts. In copyright terms an original work is one that can be said to be the author’s ‘own intellectual creation’. Where there is any doubt this is something only a court would have the authority to determine.
When a creator uses AI as a tool in creating a work, if the work still shows the author’s own intellectual creation, this would still be protected as an original work, regardless of whether AI assisted them.
In the UK, there is also protection for computer-generated works without a human creator. Section 9(3) of the Copyright, Designs and Patents Act 1988 states that “the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken”.
You may find the following general guidance on copyright useful: www.gov.uk/copyright.
We hope this response has been of some assistance.
Best regards,
Copyright Enquiries
I'm curious what this means for the copyright of videos that are taken by an automated system? If I write a program to automatically detect meteors in my sky camera and leave it running all night, wake up the next day and find I caught a nice video of a meteor, do I own the copyright on that? Or is it not subject to copyright since the camera was automatically triggered, even if by my own software?
So hypothetically if a crowd sourced learning based AI algorithm was to create some kind of Medication off of a simple prompt we wouldn’t be able to copyright that? What if some one else were to make that medicine they wouldn’t be able to copyright it? Do we just sort of create a public domain medication purely because AI generated it?
Or does this only apply to artistic works?
This is one of the rare cases where I 100% agree with the legal rational on a novel copy right case.
You can create whatever you want with A.I generation, but you do not have the exclusive rights to commercialize that content.
Issue I see is what if Disney uses Mickey Images that are in public domain to train the AI to draw and uses a plot to get all the frames drawn in the movie. It creates an issue in that Disney controlled the process from beginning to end but they don't own any specific copyright as the input images are in public domain even if the output images resemble little to do with the input images, except the character itself.
Wouldn't the Dutch cat fall under Dutch/EU law rather than US copyright law?
Erm Leonard, have you driven recently, noobody follows the speed limit. There are places where matching traffic is 20+mph OVER the speed limit hell my short drive to work increasingly resembles training for nascar!
I can’t figure out how they are going to draw the line here between giving instructions to advanced photoshop filters which are themselves generative and giving instructions to an AI model.
I bet if we dropped a computer with photoshop 100 years ago, they would see it as AI and not allow copyrights.
And I think in the future we will eventually allow copyright for certain AI generated content where the user iteratively creates the work by reviewing the output and adjusting the input in real time.
I hope you can do a video about the guy wanting to copyright every meledoy and it's relation/non relationship to AI
I hope you do a video on the Internet Archive V. Hachette ruling.
Your last example came close to what I was wondering. What if I had AI generate a disturbing image of something I wanted done, then imported that image into Photoshop and made some changes--how much would I need to change in order to make it copyright-able?
Change your example to 'I took a royalty free stock photo and tossed it into Photoshop and made changes' as it would work like that.
There is no minimum bar of 'you own this now' and you'd only own the changes you made. You can only protect what you made so a modified AI image is only going to have a partial copyright, IF your changes had at least some creativity or protectable elements.
It's like how The Animals have a copyright on their specific version of House Of The Rising Sun, but they only own the unique elements to it. the slight change to the lyrics, the specific recording, and some parts of their arrangement. But they can't stop all other covers of House Of The Rising Sun.
Yay, more Frenchy goodness
Aspiraing patent lawyer here...Wouldn't it be possible to argue a copyright over a compilation of images? That seems the best route for copyright over AI generated images. The individual images lack protection, but the compilation put together by a human would.
Hypothetically, you could be limited by your use of an AI image through terms of service with the company that generated the image. BUT, if someone else comes along and takes that image from you and uses it in commercial things, wouldn't they be fine since they aren't bound by the Company's Terms of Use or Copyright since the AI image is copyrightable?
In that case, the only person limited in their use of an AI generated image is someone who commissions one from someone with a Terms of Use
It will be really interesting to see where the line ends up in terms of can you copyright generative content. If a novel text prediction model is trained by and entirely on the work of a single, prolific living author, would the output of that model be subject to copyright since it is published by the author and owner of the model, training data, and output? Alternatively if you algorithmically generated novel 3D sculptures from parts based on your own models, then 3D printed them, would the end result be a copyrightable work?
The way I see it, even if it's only been trained by a single author from only their own copywrited material, anything output from the model would still be uncopywritable, as it's the act of fixing the concept onto a medium that is copywritable, and that the model is doing the fixing, not the prompter or trainer of the model.
The text file and the 3D models would be the medium, and 3D printing the model would be like printing a digital photograph.
That being said, I'm not even american, so if I'm wrong I'd like to be corrected.
@@LuxurioMusic I wonder how smaller scale ai/algorithms play into this then because if I follow the logic that anything that an algorithm changes is uncopyrightable then does that mean if I use an algo to reformat and add details to my story that I've lost the ability to copywrite my story? I don't think that the "you own the input but not the output" argument holds water if you can't first pick out clearly what the ai did vs the original author of the work that was trained on. I just don't see how it is enforceable I guess.
The answer is no, with caveats. As stated in the Disney example, only the portions of the final product that are made/altered by a person can be copyrighted. It doesn’t matter that the inputs are yours, or even that the AI is yours. The issue is that you didn’t create the final product. And even in cases where you edit the product afterwards, you only own the edits. The base files are uncopyrightable.
AI generation is useful as an idea machine. But the final product must be made by you AFTER being run through the AI
@@bioxx666 I totally agree that enforcability will be a huge mess. With the example you give, I think anything that would be classified as a creative decision that's done by a model would be uncopyrightable, but I'm unsure how that extends to reformating. I imagine if details are added by the model, the details would be uncopyrightable. I think that this would be the case regardless of whether you can distinguish between the human and model, just that being able to enforce it properly could be impossible to figure out.
@@Megumin_Random say I am a digital artist and I create a digital image. I then write a script that divides the image into 16 tiles and shuffles their order and outputs 32 variants. Would the output of a simple algorithm, written by the author of the original work and acting on a wholly original work be uncopyrightable?
At what level of complexity does the mechanical or digital processing of a work render it transformed beyond the threshold of copyright?
How many human interaction is necessary to make AI art copyrightable? It seems like no one knows, beyond a vaguely defined "substantial alteration".
as a question of semantics and modernising the thought experiment . If you take the "infinite aI's in a sealed disconnected serverroom one of them will reproduce some copyrighted work of fiction and you can guarantee that that work was not in the ai training data" if that happened what would the legal status of that work be? It was produced completely independently with no knowledge of the original but isnt a product of a person? what does that mean for the status of the original work?
Wasn't Getty also sending demand letters for images in the public domain?
Any thoughts on the Wizards of the Coast debacle of January?
Now eventually, the interesting question becomes if we have a series of AI generated stories that are only partially AI generated: if altered by humans how much alteration makes it copyright worthy, or if only the human generated parts are, what the cutoff is for being considered as indiscernible for being able to apply copyright laws to the work?
Since AI are pretty much considered a third party since anything they create is not copyrightable by either the prompter (the one feeding the prompts to the AI) nor the creator of the AI (so not in a collaboration or in a work for hire situation), if an IP holder were to willingly feed their IP into an AI wouldn't that dilute there hold over it since AI can't sign an acknowledgment of said IP holder.
Those Mickey/Donald abominations are nightmarefuel. 👀
My question with this ruling is the following: Let's say one of the programmers of midjourney tries to copyright something generated with AI. Does that count for the copyright? He made the program and imputed everything, so the resulting image is something born from his human creation and let's say a couple of years in the future if someone creates something like midjourney or dalle on its own, do creations of that program created by that person are eligible if the answer is no then I would like to know if pendulum paintings are eligible for copyright because in those the human only has control of some inputs but not the result
Interestingly, I asked chatgpt and it got the answer wrong. The monkey case made it pretty to clear to me what the answer would actually be.
This is great. AI does not really know what it is doing in any case. AlphaGo does not understand go strategy, chat GPT does not understand what information it is giving, Midjourny does not understand what it is drawing. It only draws from data it has saved. It took a .jpg and saved it in a format it understands, same as if you converted it to a .bmp. There are artist out there who's images are being used every day in Midjourny. It is cool, but you can't commercialize it without infringing on someone's copyright.
What I'm wondering is the legality of the AI generating art program itself. Because the AI is trained off of copyrighted material and then sometimes sold, would selling the training data that scraped from copyrighted material constitute a violation?
that is a seperate issue, you can train the model on non-copyrighted art or art you own the copyright to.
Will have to wait to see when it gets to the Supreme Court!
PETA should get right on this one! 😁
I'm curious how granular this ends up getting. For instance: Say I use an AI to generate a 3d character model. Then I also use an AI to separately generate a line of clothing and accessory models that can be used in conjunction with that character model. So far nothing is copyrightable. I then import those assets into another software suite, select which combination of clothing and accessories to attach to the model, and then output a still image of that combination.
The question then is if the creative decision making involved in assembling several uncopyrightable AI generated elements together in a combination that hasn't previously existed before sufficient to allow me to claim ownership of that specific combination? Would someone using that same combination of elements need to seek my permission?
Then there's the question of whether or not I can take an AI generated cartoon character, register it as a trademark for print and visual media, then enforce that trademark on anyone that produces a book or video in which that character features prominently on the cover, packaging, or in promotional materials. The obvious argument being that the character appearing prominently in those circumstances may cause the public to confuse that product as being one that my company manufactured.
Well for the second, on a basic level you don't use Trademark for cartoon characters, for the most part. The specific Mickey Mouse 3 circle logo is trademarked, but Mickey Mouse the character is not. Trademarks are for logos and symbols. You can make a bootleg Mickey shirt and be liable for copyright infringement, but not trademark infringement.
And for the first example, I'd swap the AI clothes with just real clothing and see how the example reads. Like, what would you protect? You don't own the idea of wearing clothes, or the concept of a shirt having colors, or being short or long sleeved. Could someone ever claim ownership of an outfit? Has Nintendo ever stopped anyone else from wearing a red shirt and blue overalls? Would an arrangement of any set of objects ever have strict protection? At the least it prob really depends on the specifics.
@@larryinc64 Disney started creating trademarks that incorporate their iconic character designs as an element. The 3 circle logo you speak of is only one of many different Disney trademarks
To be specific about what I'm getting at; the "Steamboat Willie" film will enter public domain in 2024. Disney Animation Studios incorporated the Steamboat Willie character design into their trademark in 2007. You can see it, and part of the actual film itself, incorporated into that trademark splash screen at the beginning of any Disney animated feature created since then.
They're not shy about describing the character, including the Steamboat Willie design, as being directly associated in the eyes of the consumer with the Disney brand and authentic Disney products. Nor have they ever been shy about their intention to act on anything they estimate as causing consumer confusion.
It's more likely than not that in the next few years they, being as litigious as they are, will attempt to enforce a trademark in lieu of being able to enforce a copyright. For instance by arguing that a print on a t-shirt of Mickey from Steamboat Willie holding a ship's wheel and whistling is substantially similar enough to the Disney Animation Studios trademark as to cause the consumer to confuse it for being an authentic Disney brand product.
As for your response to the first part. "Has Nintendo ever stopped anyone else from wearing a red shirt and blue overalls?" Uhh... sort of but not exactly? Look up Nintendo v. Marika Co.
At least some arrangements of objects can be protected. In your Mario example you mention a red shirt and overalls which are only some of the elements than are contained in Mario's outfit.
Rather than trying to apply it to a cosplayer, which is what I interpreted you bringing up Mario in relation to real clothes was intended to make me think of, let's keep it in the realm of the holistic character design fixed in some form of media that I was asking about.
Put a long sleeved red t-shirt and blue overalls on a male character of a video game that's neither intended to be a parody of, nor a criticism of, Mario Bros. Then add to that outfit 2 oversized gold buttons for the straps and put a bright red newsboy cap on his head.
I might be crossing the line there. Maybe I'd get away with it and not get sent legal threats because the newsboy cap doesn't have a white circle in the front that contains a red M that looks like an upside down W.
It is Nintendo though and they're pretty hamfisted about protecting their copyright. Even in circumstances which do very obviously meet every criteria for being fair use.
When you do a work for hire the one doing the hiring doesn't immediately own the image. The artist owns the copyright to it and then transfers that ownership to the person who hired them.
AI can't do work for hire because the ai can't own copyright to begin with thus cannot transfer copyright to whoever 'hired' them
Can you explain the legal logic used to differentiate the copyrightability of "AI"-generated content versus literally any other computer-generated content? For example, the copyrightability of computer generated graphics (CGI) is well established. While it is okay to copyright the final rendered video track of a Disney, Pixar, or other animation studio movie, humans didn't literally "create" the rendered frames of that movie, they created instructions (or inputs) for the software to follow while it uses a combination of geometry, textures, lighting models, physics, and other simulations (water, sand, etc). The computer actually created the final image(s).
Another example is MIDI-triggered sounds to compose songs (a relatively common occurance). While those triggered sounds *could* be recorded by a human, lets make the argument even more clear by saying the triggered sounds are entirely computer generated noises. Given this, all the user did was write out a tablature of their song (just like sheet music). It was the computer itself that "played" back the notes and rhythms; these human-created instructions are simply an input into the application that then itself plays back the instructions and actually records it into a "fixed medium."
In all these situations, a human created a piece of software, a human provided inputs to that software, and then the computer generated an output using a combination of that software and that user input. Current AI isn't doing anything "unexplainable," in the sense that humans built and configured every level of the system, from the electricity to power it, to the silicon to run it, to the software to control it, to the inputs to train (or "program") it. How is this process fundamentally different than my examples?
While I don't necessarily think this would be the best outcome, taking the law literally, I'd asser that all of my examples would not be copyrightable, as a "human" didn't "fix" the final output into a "tangible medium," the software did. So, what "logic" is the court using to determine that *this specific type of output* isn't copyrightable (I,e, things labeled "AI"), while countless other types of computer generated output is?
Original thoughts in Twitter thread at twitter.com/robfrawley/status/1652892074876760064
My issue is that AI is not actually intelligence. Its a statistical predictive model. Its no different than a Jackson Pollock, only that the noise has been transformed into something you recognize more clearly.
Here's the problem with this whole thing: How will they know? You can strip metadata. You can take the image you had a prompt design and hand it to an actual artist to commission to make a copy of. Hell, you could draw it yourself and cut out the middle man. There is no way on gods green earth that there's enough memory and storage to store every single image an AI has ever created. When people say yes you can, they're naïve and they need to stop eating paste. With the amount of images that are being generated per MINUET I doubt there's enough storage to keep that stuff forever. Even with hashs, there would still be someone that can outright bypass the bullshit legal overhang and claim AI generated shit as their own. These people are demented.
I always thought there might be a very grey middle area. My art skills are still very lacking, and Ive been using ai to help practice, essentially. The ai generates the image, but I trace over and correct anything wrong with it. Does that count as human or ai?
I don't really understand how this decision is at all sustainable.
For consideration, look at Stable Diffusion's img-to-img function. I draw a doodle (or collage it, or take a photo, whatever), and feed it in as the "image" part. I also provide a text which mixes some characteristics shared with the doodle (e.g. colors, features, etc), as well as some original changes I want made; this is the text prompt. There is a little slider called "prompt strength" which determines how much the AI pays attention to the original image, versus my text prompt.
If I have a 0 strength prompt, the resulting image is just the original image. If you have a 1 strength prompt, the original image has effectively nothing to do with the final product. Obviously some uncontrolled "random" or "computer-driven" output does not result in an image becoming uncopyrightable, or using any non-deterministic tool in Photoshop or paint.net would void your copyright, so at what point does the result become uncopyrightable?
The img-to-img thing would be more like a derivative work. You wrote a song and gave it to another band to cover. Their recording is based on your IP but you don't own fully what they made. The riff or lyrics they took from you would be yours, but any changed lyrics or an added solo would not be.
As said in the video, you own the parts you made. An image can be partially protected, it's not all or nothing.
Okay. Again. At what point can I sue somebody for stealing the resulting pic and reusing it without permission in a manner that is not fair use?