i disagree with devon on a few subjects (alright, maybe a lot), but they're just legal disagreements. hermeneutics stuff. let me tell you, you don't know what a breath of fresh air is seeing someone from tech calling an actual law specialist to talk about law and AI. this is A VERY RESPONSIBLE APPROACH and every tech engineer/AI software developer and tech content creator should learn from you guys.
Not a fan of legal eagle, he's obsessed with attacking Trump and has major TDS. Legal eagle is a defender of the establishment and a defender of the democrat party.
If the software needs to steal your work in order to understand what a style of your name means, and your name is a very thing that defines the quality of its output and the output occupies the same market as yourself doesn´t it infringe one´s copyright? If John is an impressionistic oil painter that uses broad strokes and different shades of blue, and you put this description into AI - its one thing, since there are a lot of ways one can define this. It would also need a certain art direction to get the style where one wants it to be. But if you copied John´s works and plugged it into a software and say "in style of John" and you produce 1000 images pro minute and sell these for 10$, people don´t have to pay John for his original work or prints. Than you are disrupting the value of his art misusing his name and copying his works for a purpose he hasn´t allowed
I'm a third of the way through and I'm enjoying it but you guys need a backlight to create a little separation that will stop your hair blending in. Fortunately you both have the same colour hair so you can both be lit the same way. I'm guessing that you decided to wear light coloured t-shirts for much the same reason. If you had black t-shirts on you'd look like a couple of disembodied faces... And you're obviously incredibly handsome guys so you should standout on the screen. Devon's video is so brightly lit, that when it cuts to you it gives the impression of a couple of guys hiding in some bunker doing a secret podcast. Which I don't think is what you're going for but it would still be a dark studio setting with a soft backlight installed. I'm also watching on my phone screen so you probably get lost in the background even more on such a small screen (pixel 7 pro)
Thanks for the amazing feedback Jacob! We will moving to a new studio soon so we will definitely keep some of your tips in mind. Thanks again for checking us out!
30:03 _"…Van Gogh style…"_ Funny that you mentioned Van Gogh. Last year I created 4 videos about Van Gogh's journey through the northern part of The Netherlands: The province of Drenthe. I created the script, as a monologue, hearing his thought. But thought it would be nice to have the narration be in the style of Vincent's letters. So, after many ChatGPT iterations, I got the video's in that style. I was basically the director of what AI was generating. Compilation example of the video: ua-cam.com/video/hiQVv5zdPjk/v-deo.html The video is in Dutch. But the generated English captions are quite spot _(for about 90%)_ on! Just turn them on with the cog wheel in YT Fun facts: The train scenes are shot in old trains with a large green-screen outside. The landscape, to replace the green-screen, is shot with my iPhone, while driving through the landscape where I filmed om that road at 7:49At 7:04 is the room where Vincent actually lived. At 9:47 the music is not from my usual Artlist Stock Library, but created by my girlfriend's gitarist "Gijs Kos". It hints to Don McLean's "Vincent", as you can hear at the end of the track. It kinda has the melody of the musical phrase "starry, starry night". It would take too many prompts to polish that that last track into something usable. AI can be a very handy tool. But it only mimics the art of a real artist!
I think that when displaying a AI generated image, that incorporates a particular artist as part of the prompt... then that artist should be credited / referenced, similar to crediting a photographer / organisation when displaying a copywrited image (eg Getty).
Crazy interesting, the whole legal aspect of it. It's fascinating how it's quite difficult to draw these lines around which end of the ai art is copyright infringing: the data collection part ( arguably it's the same intent for these images like mr LegalEagle said, but on the other hand it's what humans do) or the output part and who's liable for the infringement there
We can't possibly hope to figure this out when we can't even say how human brain works. Free will is most likely illusion, but where do we get ideas from?
WARNING TO VIEWERS: In every legal dispute/issue, one lawyer will be proven wrong. This lawyer is often correct in this video but there are significant errors. Also, there is no word “copyrightable”. Which he erroneously uses repeatedly. The word is not “copywrite”. Many lay people get this wrong. A lawyer shouldn’t, especially one self-proclaimed as an expert in a public video. A copyright is a noun, not a verb. It is a noun like the word “deed”. You cannot say “deedable” (unless you are unschooled). AI will be deemed an “act of nature”. An act of nature cannot copyright or patent. That is the answer he should have given. As to why scraping (the interviewee says “ingesting”) of prior works is not infringement, it is not, as he argues, based on a fair use defense, but because nothing scraped is substantially similar to the AI product, nor compromises the commerciality of the original work claimed to have been copied.
Be irreplaceable? Has Devon not considered what AI, AGI, AGI+ is implying for the future of human economic /existential value? AI IS the information economy, anything that can be converted into information, is going to be the domain of AI executors.
whenever a user enters one or more words to guide the AI towards outputting a picture (that resembles the rough picture the user has already very loosely formed in his mind), ZERO AI generated pictures are 100% AI, therefore , because a human (author) entered those words, and in that order*, ALL AI pics (created partly through incorporating the human entered parameters/prompts/hints/styles) are ALL potential candidates for copyright protection....the courts will , in time, determine how little or great that potential is. My best guess: IF an AI tool does NOT output effectively the SAME picture when an identical set of text prompts are entered by the human user in precisely the same order* then my GUESS is that the courts will allow AI pics to be eligible for copyright protection...same as photographs. After all, what is the difference between choosing to select those particular words and entering them in that order* and selecting from the various light related settings on a camera? very little as far as i can tell. BOTH processes shape the output/picture and both do so through the input of conscious HUMAN decisions/design. * i'm assuming here that the AI algorithm/s/ware prioritizes the user's first entered word over the second, second over the third and so on, but that might not be how the technology functions at all. (but, even if its processing is not that basic/ stringent, i'd be very surprised if its processing took no account of the order of those words at all) Not prioritizing the human entered prompts one over the other (or in some other codified/defined way), would very likely decrease the credit attributed to human 'authorship' by the courts as they seek to legally define 'the spectrum' Devin talked about in the vid.. Point remains the same though, just minus that extra , possibly legally decisive, variable.
Just found this because of the post on Cold Fusion just now. Dude, should have said something a bit earlier about the podcast, or maybe you did and I missed it. Anyways, quite the surprise to see your face, you do not look like what I have imagined. But then again, that does not matter, gonna devour this content for sure!!!
I don't think AI is copyright infringement. How is asking chat gpt to write a new Harry Potter story any different than the person who wrote Harry Potter and and Cursed Child? It's taking existing characters and plot and writing another part to it. Furthermore there are plenty of artists who paint in the style of other artists. Or make work inspired by other artists and it's through the process of critique that we can say "this new artist uses techniques made famous by Van Gogh or Monet" or be even more critical and say "this is a cheap imitation of Van Gogh". Much like we say Cursed Child is a cheap imitation of the original 7 HP books. Whenever something gets popular, someone else tries to ride the wave of popularity. Arguably popular books like HP and twilight and the hunger games prompted countless other books to be written that didn't live up to the popularity of those ones. Films like The Notebook prompted dozens of similar films that fell flat. So granted, AI can create work that is commercially viable but is not going to beat out the original inspiration. Now if the original inspiration starts to fall flat and the imitations become better (the "drake" AI songs by ghostwriter are better than the stuff drake has been putting out lately), then that's kind of a grey area.
Once your nation's written statute books/ law books and a large chunk of its recorded case law (a paltry total of just 57 volumes for the UK) has been 'fed' into its database AI tools will apparently render many/most of the services offered by today's lawyers unnecessary / unprofitable. Given how much intricate knowledge one needs to have to practice law in each of its specialist/particular areas/fields, it seems likely to me that an AI tool (which 'learns' ) will undermine 'the law' as a profession at a shockingly rapid rate. drawing as it will upon a very precise 'memory' of every single law in your nation/state and all of the case law that has effectively interpreted those laws over centuries, it will be able to output to the user a solid legal case to take before the court, no matter which specific area of law is involved . (or simply tell you that on the info the AI received from you, you are not likely to win your case, even giving you a precise % win/lose rate for cases of that nature in your nation/area ) potentially saving you time, money and much anxiety) One can imagine that in relatively straightforward/simple cases (at least initially!) , rather than a lawyer the users, armed with such output. may instead choose to hire the services of an actor/other confident orator to represent them in court...simply to verbally present their case to the court ! (as specified by the AI tools output). AI, armed with all that legal info/precedent could just as readily be imagined as able to assess the merits of cases too, assisting (eventually replacing) the work of magistrates/judges ! ... Eg; identify/remove/avoid any biases* from their decision making processes that may have unwittingly or otherwise , crept in over time. *racial/social/class/religious biases for example.
I think question should not be: is it legal, but: should it be legal? I don't know why people think old law will always be good for today world. Like what is stealing when digital come to world, because you not taking anything, so there is not direct lost. And law still don't take to account that coping something don't mean someone lose something. Because many research shows that people which illegal coping stuff, don't making lose in profit for owners of copyrights, so why court still think they do? When research shows they don't? Because law should change in direction for people, not corporations. Personally I think copyright should be lost when author die, our civilization always base on pat experience, so we only hurting ourself by blocking progress for profit of few. Most discoveries would be done anyway, as it was common in past for few people in same time to discover same thing, because it was base on our understanding, not on they genius only.
To AI content creators : What the heck-a-roony is this Mrs. Black ; It is a candy dish Ned, $90. Well I guess you could put a lot of nice things in there... No its a candy dish Ned $90 !
This is the biggest misconception about Mechine Learning Algorithms, they don't copy when learning they observe and adjust. If you see a picture is online then it's open for observation. That's what you as a user on the internet intended when you published it. The web browser and the internet work by downloading content and showing it to you. No download, no observation, no content. What they do when training this models is they visit whatever page and observing it, the same way human do. I feel like everyone is missing this point because they are not educated of how software or Mechine Learning works. That's a big issue and really scarry that we have people trying to decide on disciplines that they don't understand. I think the problem will continue to build up because the progression of digital age.
I believe a good prompt is like a good google search. It's a skill. You must know what and how to ask the AI in order to create something worth looking at. So I believe the prompter has the copyright. But then, what about an AI that prompts another AI?
33:15 this is funny, because Mickey as way of drawing is as art style, so why Disney can so easily block people for using this style? If they don't name it Mickey and it would not copy other aspects?
Another idea here is: Could AI systems like DALL·E or ChatGPT, stick a copyright license agreement on any content their computers generated. And furthermore, could they oblige all their users to accept a terms of use agreement to that effect, as part of an account application process. Open AI could concievably archive everything they generated and when a person published something, they could say our archives indicate that this was originally produced by us.
If I had to guess from a technical solution: AI-companies will be required by law to submit a digital signature to a block chain-like system to register the work, so people know it's AI-generated. And original artists/authors will end up doing the same to register their works. And thus every new work will end up having a signature attached. Maybe we'll do the same with e-mails, tweets, etc. so we don't end up with scams.
@@autohmae That's interesting. Then what about combining NFTs with digitally created art and text. It sounds like a similar thing. If you were trying to enforce copyright, then that could work.
It makes me quite upset to hear that an AI generated image prompted by me isn't own partially by me. At least you're trying to tell me that somebody can prompt the exact same image in the exact same way as me. I don't think so but this is very new stuff so I can understand the legalities are lacking
12:35 so he missed part that AI generate at last 4 images and human is selecting one which is use? Yeah, this would be great point to defend right to AI generate images, not all, but one selected.
Really appreciated hearing the lawyer perspective on all this. Will say that the generative art AIs are really only mimicking artists. Artists are already a struggling profession that’s hard to break into. If there is no longer a demand for entry to junior level artists because they’re replaced by AI, we will see fewer and fewer “highly skilled” artists who are actually creating original content AIs are trained off of. Worst case, art becomes strictly a hobby or profession for the wealthy, and we enter an era of stagnation when it comes to creativity.
unless you're exceptionally talented/gifted isn't 'art' already a profession that is effectively only an option for the wealthy ? who in the 'west' , typically with a colossal mortgage debt and/or other incessant and inescapable costs to existing, ( student debt and/or ever increasing rent for very impermanent shelter for example) opts to study/work towards becoming a professional artist? only those with the means to cover those inescapable 'western' expenses , and potentially for many years...people in that category are the wealthy, at least i think they are.. how many working/middle class kids currently make enough income to support just themselves from their work as an artist? Rhetorical Q obviously. but my guess is a maximum of only as many as those who displayed an exceptional talent for art from early age and therefore had a more than reasonable expectation (and likelihood) of making a living as an adult artist (should they go on to decide to study/work towards that goal.)
Interesting about the copyright. I know a guy who wrote a digital novel that changes different parts of its content on each download. I think he's still trying yo figure out how to work around copyright protection on it.
@@ThroughTheWeb There's so many I'd love to see... Anton Petrov, Fraser Cain, Marcus House, or Sabine Hossenfelder for any kind of science or space discussions.
I'm glad people have such a good understanding of neural network AI already, makes me positive we won't use it for nefarious means as much as if most people didn't knew how it worked. Just because knowledge makes it harder to exploit people. The AI training issue and fair use make sense to me, but when you ask the AI for very specific images it will basically copy some pretty famous works almost 1:1. You got some great examples in your talk, as well. This might come down to the same laws like in music, where a few notes can't be copyrighted but a theme or whole melody potentially can. Might take seperate AIs to check AIs for infringement before the piece is published.
I'm sorry. They don't understand mechine learning. If you ask a machine learning model to write as Shakespeare you are not copyrighting shakespeare. The words that a models outputs is not copyrighted. You can copyright a book but not the words. The same thing goes for art. You can't copyright a scene, color, style. But you can copyright your art. The ML models don't copy they observe, the same way that those artists did when they learned. Thus making it not a care for copyright infringement. The other issue is I can ask an artist to draw me like Picasso. That's not copyright infringement. That's basically what everyone is doing when asking these models to output a new image. There is no case to be made here. This is just some companies and artists that are really angry. Sure they can be angry but there is no case to be made. Just use the new tools and become better artist, have fun
When all of humanity has a Universal Basic Income - the point of copyright will evaporate (as copyright is strictly a financial concept). Where AI ends up taking us is just utterly beyond our current ability to conceive. Afraid it's rather a "suck it and see" roller coaster ride into the future 😎😎😎
Excellent content! Although style cannot be copyright, artists can actually sue when a song infringes their "personality rights". "Heart On My Sleeve" ft. fake Drake had to be taken down. Tennessee just passed the "Elvis act", and there's actually also the "No Fakes Act". Big steps have happened since this video.
Performers will have an easier time getting restrictions on the use of voice and likeness since that’s more relatable to the general public (and politicians) than for having their data used to train a model whose primary purpose is generating new content.
@@pokepress yup, producers and songwriters' styles will be mimicked freely. But then again, we've been doing that for centuries. We copy bach's style of fugues, to bossas, to blues, to edm...
There is no such thing as “personality rights”. The Elvis Act has been in existence for decades. The fake Drake song was not taken down because of any legal decision. Possibly they were afraid of the cost of legal action or were compensated. We don’t know.
"a work solely created by AI ... cannot be copyrighted" Seems like a misunderstanding of how these tools are used. They cannot create anything "solely" on their own, they must be given a prompt by a human, in order to do anything. Even human commission artists are capable of making art on their own.. lol If I can copyright a descriptively written story, we should be able to copyright works created by using AI generation tools. A detailed description of a scene, aka the prompt, is just as "human produced" as anything else that can be copyrighted. The output is just as unique and no one else is likely to come up with anything that looks the same. Many AI artists write lengthy paragraphs, that describe every little detail of a scene, in such a way that the AI has no choice but to directly create what the user wanted, giving them full the control over the creative decisions. The creation of which can take hundreds of modifications, trial and error, and hours upon hours of work, to create the image the artist originally imagined in their mind. Just as much creative effort as any other form of art. We're seeing a HUGE bias against this, and a refusal to imagine the possibilities of how different people might use this with more effort than others.
As an artist, I appreciate hearing the nuanced and balanced take on the matter coming from Legal Eagle. It's reassuring to know that purely AI-generated content can't be copyrighted and the precedent for that is already pretty strong. But there's still a lot of gray areas that could go either way in court (like where the line should be drawn for partially-AI generated content for example), so it's great to see a detailed breakdown of that from a lawyer.
We will have AI rights activists in 20-30 years, if not sooner. :P The replacement of jobs is happening since the industrialization. People have been replaced by machines for aeons and then manufacturers always found out that there are certain aspects of a production you just can't automate. Tesla still has human workers in assembly because they have an eye and intuition for potential damage and failure, while AI can only deal with it after it already happened. So it's actually cheaper for Tesla to still have some humans working in assembly.
At 22:05 As far as I know there's no "copy" happening in that process. The companies that have dataset collections typically claim their data was collected legally, and the responsibility is on those companies, not the AI researchers. And as far as I was aware, some of these datasets are just links to existing publicly available content, so the data isn't even being stored anywhere other than original locations.. And to top it off, once the training process is complete, the data is essentially discarded, and all that remains is a model file full of mathematical "weight" values between "neuron" connections, no data whatsoever, not even in any compressed form. Otherwise they'd have just discovered the world's greatest form of image compression. The AI can never accurately reproduce images from it's dataset, without significant mutations being present.
@@ThroughTheWeb I don't really have anyone in mind, I think you two do an excellent job together; no 3rd party needed! I appreciate you keeping (partisan & divisive) politics out of the stuff you do, so bringing guests like this who has a pretty heavy political leaning baggage will alienate parts of the audience. I personally can't stand LegalEagle and his politics (I'm probably in the minority, looking at other comments), so others would likely feel the same about someone I would suggest.
![[UA] NAVI проти G2 Esports | Blast Premier Fall Final](http://i.ytimg.com/vi/QWlIm4FGESQ/mqdefault.jpg)
i disagree with devon on a few subjects (alright, maybe a lot), but they're just legal disagreements. hermeneutics stuff. let me tell you, you don't know what a breath of fresh air is seeing someone from tech calling an actual law specialist to talk about law and AI. this is A VERY RESPONSIBLE APPROACH and every tech engineer/AI software developer and tech content creator should learn from you guys.
This is the best crossover of two of my favourite UA-camrs 😁
Let's go! Hope you enjoy it Daniel.
@@ThroughTheWeb it's late at night but totally worth it!
LegalEagle is a hack, we learned that during the Rittenhouse trial.
@@YeeLeeHaw really. What happened?
Not a fan of legal eagle, he's obsessed with attacking Trump and has major TDS. Legal eagle is a defender of the establishment and a defender of the democrat party.
If the software needs to steal your work in order to understand what a style of your name means, and your name is a very thing that defines the quality of its output and the output occupies the same market as yourself doesn´t it infringe one´s copyright?
If John is an impressionistic oil painter that uses broad strokes and different shades of blue, and you put this description into AI - its one thing, since there are a lot of ways one can define this. It would also need a certain art direction to get the style where one wants it to be. But if you copied John´s works and plugged it into a software and say "in style of John" and you produce 1000 images pro minute and sell these for 10$, people don´t have to pay John for his original work or prints. Than you are disrupting the value of his art misusing his name and copying his works for a purpose he hasn´t allowed
A.I. will create a mass effect giving rise to the Geth. 😂
If I could make everyone have Dagogos voice I wouldn't because that'd be confusing but he has the best voice.
🤣🤣🤣
Incredible comment loll
I'm a third of the way through and I'm enjoying it but you guys need a backlight to create a little separation that will stop your hair blending in. Fortunately you both have the same colour hair so you can both be lit the same way. I'm guessing that you decided to wear light coloured t-shirts for much the same reason. If you had black t-shirts on you'd look like a couple of disembodied faces... And you're obviously incredibly handsome guys so you should standout on the screen.
Devon's video is so brightly lit, that when it cuts to you it gives the impression of a couple of guys hiding in some bunker doing a secret podcast. Which I don't think is what you're going for but it would still be a dark studio setting with a soft backlight installed.
I'm also watching on my phone screen so you probably get lost in the background even more on such a small screen (pixel 7 pro)
Thanks for the amazing feedback Jacob! We will moving to a new studio soon so we will definitely keep some of your tips in mind. Thanks again for checking us out!
30:03 _"…Van Gogh style…"_
Funny that you mentioned Van Gogh. Last year I created 4 videos about Van Gogh's journey through the northern part of The Netherlands: The province of Drenthe.
I created the script, as a monologue, hearing his thought. But thought it would be nice to have the narration be in the style of Vincent's letters.
So, after many ChatGPT iterations, I got the video's in that style. I was basically the director of what AI was generating.
Compilation example of the video:
ua-cam.com/video/hiQVv5zdPjk/v-deo.html
The video is in Dutch. But the generated English captions are quite spot _(for about 90%)_ on! Just turn them on with the cog wheel in YT
Fun facts:
The train scenes are shot in old trains with a large green-screen outside. The landscape, to replace the green-screen, is shot with my iPhone, while driving through the landscape where I filmed om that road at 7:49At 7:04 is the room where Vincent actually lived.
At 9:47 the music is not from my usual Artlist Stock Library, but created by my girlfriend's gitarist "Gijs Kos". It hints to Don McLean's "Vincent", as you can hear at the end of the track. It kinda has the melody of the musical phrase "starry, starry night".
It would take too many prompts to polish that that last track into something usable. AI can be a very handy tool. But it only mimics the art of a real artist!
WOW.. I have subbed CF for many many years now and for the first time seeing your face, Dagogo.. mind blowing 🤯
Another fantastic episode - Thank you Dagogo & Tawsif ♥
I think that when displaying a AI generated image, that incorporates a particular artist as part of the prompt... then that artist should be credited / referenced, similar to crediting a photographer / organisation when displaying a copywrited image (eg Getty).
Use A.I image generators to create millions of Mickey mouse pictures, Then watch the A.I get Disney cease & desisted into oblivion.
Crazy interesting, the whole legal aspect of it. It's fascinating how it's quite difficult to draw these lines around which end of the ai art is copyright infringing: the data collection part ( arguably it's the same intent for these images like mr LegalEagle said, but on the other hand it's what humans do) or the output part and who's liable for the infringement there
We can't possibly hope to figure this out when we can't even say how human brain works. Free will is most likely illusion, but where do we get ideas from?
This was super informative! Thank you 🙂. You guys brought up some really interesting talking points.
Awesome crossover! I know its gonna be a good one 😎😎
It’s so weird. I’ve heard your voice so much from ColdFusion, and I’ve seen your face from your posts, but I’ve never seen them together
WARNING TO VIEWERS: In every legal dispute/issue, one lawyer will be proven wrong.
This lawyer is often correct in this video but there are significant errors.
Also, there is no word “copyrightable”. Which he erroneously uses repeatedly.
The word is not “copywrite”. Many lay people get this wrong. A lawyer shouldn’t, especially one self-proclaimed as an expert in a public video.
A copyright is a noun, not a verb. It is a noun like the word “deed”. You cannot say “deedable” (unless you are unschooled).
AI will be deemed an “act of nature”. An act of nature cannot copyright or patent. That is the answer he should have given.
As to why scraping (the interviewee says “ingesting”) of prior works is not infringement, it is not, as he argues, based on a fair use defense, but because nothing scraped is substantially similar to the AI product, nor compromises the commerciality of the original work claimed to have been copied.
Be irreplaceable? Has Devon not considered what AI, AGI, AGI+ is implying for the future of human economic /existential value?
AI IS the information economy, anything that can be converted into information, is going to be the domain of AI executors.
I guess lawyers and judges will be obsolete in the future.
Everything will be open-sourced and in the public domain.
whenever a user enters one or more words to guide the AI towards outputting a picture (that resembles the rough picture the user has already very loosely formed in his mind), ZERO AI generated pictures are 100% AI, therefore , because a human (author) entered those words, and in that order*, ALL AI pics (created partly through incorporating the human entered parameters/prompts/hints/styles) are ALL potential candidates for copyright protection....the courts will , in time, determine how little or great that potential is.
My best guess:
IF an AI tool does NOT output effectively the SAME picture when an identical set of text prompts are entered by the human user in precisely the same order* then my GUESS is that the courts will allow AI pics to be eligible for copyright protection...same as photographs. After all, what is the difference between choosing to select those particular words and entering them in that order* and selecting from the various light related settings on a camera?
very little as far as i can tell. BOTH processes shape the output/picture and both do so through the input of conscious HUMAN decisions/design.
* i'm assuming here that the AI algorithm/s/ware prioritizes the user's first entered word over the second, second over the third and so on, but that might not be how the technology functions at all. (but, even if its processing is not that basic/ stringent, i'd be very surprised if its processing took no account of the order of those words at all)
Not prioritizing the human entered prompts one over the other (or in some other codified/defined way), would very likely decrease the credit attributed to human 'authorship' by the courts as they seek to legally define 'the spectrum' Devin talked about in the vid..
Point remains the same though, just minus that extra , possibly legally decisive, variable.
Just found this because of the post on Cold Fusion just now. Dude, should have said something a bit earlier about the podcast, or maybe you did and I missed it. Anyways, quite the surprise to see your face, you do not look like what I have imagined. But then again, that does not matter, gonna devour this content for sure!!!
I don't think AI is copyright infringement. How is asking chat gpt to write a new Harry Potter story any different than the person who wrote Harry Potter and and Cursed Child? It's taking existing characters and plot and writing another part to it.
Furthermore there are plenty of artists who paint in the style of other artists. Or make work inspired by other artists and it's through the process of critique that we can say "this new artist uses techniques made famous by Van Gogh or Monet" or be even more critical and say "this is a cheap imitation of Van Gogh". Much like we say Cursed Child is a cheap imitation of the original 7 HP books.
Whenever something gets popular, someone else tries to ride the wave of popularity. Arguably popular books like HP and twilight and the hunger games prompted countless other books to be written that didn't live up to the popularity of those ones. Films like The Notebook prompted dozens of similar films that fell flat. So granted, AI can create work that is commercially viable but is not going to beat out the original inspiration. Now if the original inspiration starts to fall flat and the imitations become better (the "drake" AI songs by ghostwriter are better than the stuff drake has been putting out lately), then that's kind of a grey area.
Once your nation's written statute books/ law books and a large chunk of its recorded case law (a paltry total of just 57 volumes for the UK) has been 'fed' into its database AI tools will apparently render many/most of the services offered by today's lawyers unnecessary / unprofitable.
Given how much intricate knowledge one needs to have to practice law in each of its specialist/particular areas/fields, it seems likely to me that an AI tool (which 'learns' ) will undermine 'the law' as a profession at a shockingly rapid rate.
drawing as it will upon a very precise 'memory' of every single law in your nation/state and all of the case law that has effectively interpreted those laws over centuries, it will be able to output to the user a solid legal case to take before the court, no matter which specific area of law is involved . (or simply tell you that on the info the AI received from you, you are not likely to win your case, even giving you a precise % win/lose rate for cases of that nature in your nation/area ) potentially saving you time, money and much anxiety)
One can imagine that in relatively straightforward/simple cases (at least initially!) , rather than a lawyer the users, armed with such output. may instead choose to hire the services of an actor/other confident orator to represent them in court...simply to verbally present their case to the court ! (as specified by the AI tools output).
AI, armed with all that legal info/precedent could just as readily be imagined as able to assess the merits of cases too, assisting (eventually replacing) the work of magistrates/judges ! ... Eg; identify/remove/avoid any biases* from their decision making processes that may have unwittingly or otherwise , crept in over time.
*racial/social/class/religious biases for example.
I think question should not be: is it legal, but: should it be legal?
I don't know why people think old law will always be good for today world. Like what is stealing when digital come to world, because you not taking anything, so there is not direct lost. And law still don't take to account that coping something don't mean someone lose something. Because many research shows that people which illegal coping stuff, don't making lose in profit for owners of copyrights, so why court still think they do? When research shows they don't? Because law should change in direction for people, not corporations.
Personally I think copyright should be lost when author die, our civilization always base on pat experience, so we only hurting ourself by blocking progress for profit of few. Most discoveries would be done anyway, as it was common in past for few people in same time to discover same thing, because it was base on our understanding, not on they genius only.
To AI content creators : What the heck-a-roony is this Mrs. Black ; It is a candy dish Ned, $90. Well I guess you could put a lot of nice things in there... No its a candy dish Ned $90 !
This is the biggest misconception about Mechine Learning Algorithms, they don't copy when learning they observe and adjust. If you see a picture is online then it's open for observation. That's what you as a user on the internet intended when you published it. The web browser and the internet work by downloading content and showing it to you. No download, no observation, no content. What they do when training this models is they visit whatever page and observing it, the same way human do. I feel like everyone is missing this point because they are not educated of how software or Mechine Learning works. That's a big issue and really scarry that we have people trying to decide on disciplines that they don't understand. I think the problem will continue to build up because the progression of digital age.
I believe a good prompt is like a good google search. It's a skill. You must know what and how to ask the AI in order to create something worth looking at. So I believe the prompter has the copyright. But then, what about an AI that prompts another AI?
33:15 this is funny, because Mickey as way of drawing is as art style, so why Disney can so easily block people for using this style? If they don't name it Mickey and it would not copy other aspects?
Another idea here is: Could AI systems like DALL·E or ChatGPT, stick a copyright license agreement on any content their computers generated. And furthermore, could they oblige all their users to accept a terms of use agreement to that effect, as part of an account application process.
Open AI could concievably archive everything they generated and when a person published something, they could say our archives indicate that this was originally produced by us.
he said AI can't do that
If I had to guess from a technical solution:
AI-companies will be required by law to submit a digital signature to a block chain-like system to register the work, so people know it's AI-generated.
And original artists/authors will end up doing the same to register their works. And thus every new work will end up having a signature attached.
Maybe we'll do the same with e-mails, tweets, etc. so we don't end up with scams.
@@autohmae That's interesting. Then what about combining NFTs with digitally created art and text. It sounds like a similar thing. If you were trying to enforce copyright, then that could work.
it's anyway will not work if you run such model on your own computer.
It makes me quite upset to hear that an AI generated image prompted by me isn't own partially by me. At least you're trying to tell me that somebody can prompt the exact same image in the exact same way as me. I don't think so but this is very new stuff so I can understand the legalities are lacking
12:35 so he missed part that AI generate at last 4 images and human is selecting one which is use? Yeah, this would be great point to defend right to AI generate images, not all, but one selected.
Something is off with his video feed. Looks like he is AI-generated.
good stuff, underrated and still relevant
Intent is important. Thank you 34:55
Really appreciated hearing the lawyer perspective on all this. Will say that the generative art AIs are really only mimicking artists. Artists are already a struggling profession that’s hard to break into. If there is no longer a demand for entry to junior level artists because they’re replaced by AI, we will see fewer and fewer “highly skilled” artists who are actually creating original content AIs are trained off of. Worst case, art becomes strictly a hobby or profession for the wealthy, and we enter an era of stagnation when it comes to creativity.
unless you're exceptionally talented/gifted isn't 'art' already a profession that is effectively only an option for the wealthy ?
who in the 'west' , typically with a colossal mortgage debt and/or other incessant and inescapable costs to existing, ( student debt and/or ever increasing rent for very impermanent shelter for example) opts to study/work towards becoming a professional artist?
only those with the means to cover those inescapable 'western' expenses , and potentially for many years...people in that category are the wealthy, at least i think they are..
how many working/middle class kids currently make enough income to support just themselves from their work as an artist? Rhetorical Q obviously. but my guess is a maximum of only as many as those who displayed an exceptional talent for art from early age and therefore had a more than reasonable expectation (and likelihood) of making a living as an adult artist (should they go on to decide to study/work towards that goal.)
For those who fear AI: GIT GUD
lol
Interesting about the copyright. I know a guy who wrote a digital novel that changes different parts of its content on each download. I think he's still trying yo figure out how to work around copyright protection on it.
that sounds like books i read years ago where you get an option at the end of a page and pick one and go to that page
Finally, -some good fucking food- *A GUEST!*
Who else should we bring on as guests in the future?
@@ThroughTheWeb There's so many I'd love to see... Anton Petrov, Fraser Cain, Marcus House, or Sabine Hossenfelder for any kind of science or space discussions.
Your voice is music to my ears. I don’t know anything about AI or technology but the videos are always so interesting so I watch.
Love the podcast but can we chill on AI/Chat GPT for a bit? It's taking over every tech podcast to the point of becoming boring
I'm glad people have such a good understanding of neural network AI already, makes me positive we won't use it for nefarious means as much as if most people didn't knew how it worked. Just because knowledge makes it harder to exploit people.
The AI training issue and fair use make sense to me, but when you ask the AI for very specific images it will basically copy some pretty famous works almost 1:1. You got some great examples in your talk, as well. This might come down to the same laws like in music, where a few notes can't be copyrighted but a theme or whole melody potentially can. Might take seperate AIs to check AIs for infringement before the piece is published.
I'm sorry. They don't understand mechine learning. If you ask a machine learning model to write as Shakespeare you are not copyrighting shakespeare. The words that a models outputs is not copyrighted. You can copyright a book but not the words. The same thing goes for art. You can't copyright a scene, color, style. But you can copyright your art. The ML models don't copy they observe, the same way that those artists did when they learned. Thus making it not a care for copyright infringement. The other issue is I can ask an artist to draw me like Picasso. That's not copyright infringement. That's basically what everyone is doing when asking these models to output a new image. There is no case to be made here. This is just some companies and artists that are really angry. Sure they can be angry but there is no case to be made. Just use the new tools and become better artist, have fun
👍💪✌
Im convinced dagogo is ai generated, but ai was not strong enough to show his face until now.
That guy is a hack. Get Rekieta Law on instead.
Good interview. Thanks for sharing it. ✌️
When all of humanity has a Universal Basic Income - the point of copyright will evaporate (as copyright is strictly a financial concept).
Where AI ends up taking us is just utterly beyond our current ability to conceive. Afraid it's rather a "suck it and see" roller coaster ride into the future 😎😎😎
the last time music and film fought back and won
@@trevorjoneill707 The Lawyers won.
Excellent content!
Although style cannot be copyright, artists can actually sue when a song infringes their "personality rights". "Heart On My Sleeve" ft. fake Drake had to be taken down. Tennessee just passed the "Elvis act", and there's actually also the "No Fakes Act". Big steps have happened since this video.
Performers will have an easier time getting restrictions on the use of voice and likeness since that’s more relatable to the general public (and politicians) than for having their data used to train a model whose primary purpose is generating new content.
@@pokepress yup, producers and songwriters' styles will be mimicked freely. But then again, we've been doing that for centuries. We copy bach's style of fugues, to bossas, to blues, to edm...
There is no such thing as “personality rights”. The Elvis Act has been in existence for decades. The fake Drake song was not taken down because of any legal decision. Possibly they were afraid of the cost of legal action or were compensated. We don’t know.
Love this!! Thanks @LegalEagle for joining!
Legal eagle has TDS just another leftist hack that defends his party
Legal Eagle's Disney LIES & Why DeSantis Already Won
ua-cam.com/video/rRpNMNJR5TM/v-deo.html
Bring Dr Mike
Sound 🔥
"a work solely created by AI ... cannot be copyrighted" Seems like a misunderstanding of how these tools are used. They cannot create anything "solely" on their own, they must be given a prompt by a human, in order to do anything. Even human commission artists are capable of making art on their own.. lol
If I can copyright a descriptively written story, we should be able to copyright works created by using AI generation tools. A detailed description of a scene, aka the prompt, is just as "human produced" as anything else that can be copyrighted. The output is just as unique and no one else is likely to come up with anything that looks the same. Many AI artists write lengthy paragraphs, that describe every little detail of a scene, in such a way that the AI has no choice but to directly create what the user wanted, giving them full the control over the creative decisions. The creation of which can take hundreds of modifications, trial and error, and hours upon hours of work, to create the image the artist originally imagined in their mind. Just as much creative effort as any other form of art. We're seeing a HUGE bias against this, and a refusal to imagine the possibilities of how different people might use this with more effort than others.
As an artist, I appreciate hearing the nuanced and balanced take on the matter coming from Legal Eagle. It's reassuring to know that purely AI-generated content can't be copyrighted and the precedent for that is already pretty strong. But there's still a lot of gray areas that could go either way in court (like where the line should be drawn for partially-AI generated content for example), so it's great to see a detailed breakdown of that from a lawyer.
OMG FIRST TIME IVE SEEN YOUR FACE!!! THE VOICE DONT MATCH THE DREADS!!!!
We will have AI rights activists in 20-30 years, if not sooner. :P
The replacement of jobs is happening since the industrialization. People have been replaced by machines for aeons and then manufacturers always found out that there are certain aspects of a production you just can't automate. Tesla still has human workers in assembly because they have an eye and intuition for potential damage and failure, while AI can only deal with it after it already happened. So it's actually cheaper for Tesla to still have some humans working in assembly.
At 22:05 As far as I know there's no "copy" happening in that process. The companies that have dataset collections typically claim their data was collected legally, and the responsibility is on those companies, not the AI researchers. And as far as I was aware, some of these datasets are just links to existing publicly available content, so the data isn't even being stored anywhere other than original locations.. And to top it off, once the training process is complete, the data is essentially discarded, and all that remains is a model file full of mathematical "weight" values between "neuron" connections, no data whatsoever, not even in any compressed form. Otherwise they'd have just discovered the world's greatest form of image compression. The AI can never accurately reproduce images from it's dataset, without significant mutations being present.
I usually love this podcast, but bringing on LegalEagle... I'll sit this one out.
Appreciate the bluntness! Who'd you like to see us bring on?
@@ThroughTheWeb I don't really have anyone in mind, I think you two do an excellent job together; no 3rd party needed!
I appreciate you keeping (partisan & divisive) politics out of the stuff you do, so bringing guests like this who has a pretty heavy political leaning baggage will alienate parts of the audience. I personally can't stand LegalEagle and his politics (I'm probably in the minority, looking at other comments), so others would likely feel the same about someone I would suggest.
Sweet, never seen the ColdFusion teams faces before!
Epic conversation between two amazing UA-cam channels.
This is fascinating new direction for ColdFusiom , great work/ideas @Dagogo / @Tawsif