In this episode we talk about artificial intelligence in the world of invention. My guest recently co-wrote an article for the Journal of Robotics, Artificial Intelligence & Law about a recent decision from the U.S. Court of Appeals for the Federal Circuit that expounded on the principle that only human beings -- not machines -- can be named as inventors under U.S. patent law. The decision applies a straightforward interpretation of patent statutes, our guest says.
Beyond invention, what about that initial spark of innovation? What about the decision might make it difficult to obtain intellectual property protection for inventions generated by advanced AI systems? Isn’t AI kind of like using computer modeling? Don’t inventors already get considerable assistance from technology? What did the court say about all that?
Joining me to answer these questions is Robert A. McFarlane, an intellectual property litigator and registered patent attorney and partner with Hanson Bridgett LLP in San Francisco. Rob chairs the firm’s technology practice, co-chairs its IP practice, litigates and advises on a variety of IP matters in the U.S. and abroad, and teaches patent law at the University of California College of the Law San Francisco (formerly Hastings College of the Law). Rob earned his J.D. from the University of California College of Law San Francisco and his B.A.S. with departmental honors, in Industrial Engineering & Political Science from Stanford University.
I hope you enjoy the episode. I mean, we get to talk about everything from Tom Jefferson to monkeys with cameras. That's five-star material right there!
***********
This podcast is the audio companion to the Journal of Emerging Issues in Litigation. The Journal is a collaborative project between HB Litigation Conferences and the vLex Fastcase legal research family, which includes Full Court Press, Law Street Media, and Docket Alarm.
If you have comments, ideas, or wish to participate, please drop me a note at Editor@LitigationConferences.com.
Tom Hagy
Litigation Enthusiast and
Host of the Emerging Litigation Podcast
Home Page
LinkedIn
In this episode we talk about artificial intelligence in the world of invention. My guest recently co-wrote an article for the Journal of Robotics, Artificial Intelligence & Law about a recent decision from the U.S. Court of Appeals for the Federal Circuit that expounded on the principle that only human beings -- not machines -- can be named as inventors under U.S. patent law. The decision applies a straightforward interpretation of patent statutes, our guest says.
Beyond invention, what about that initial spark of innovation? What about the decision might make it difficult to obtain intellectual property protection for inventions generated by advanced AI systems? Isn’t AI kind of like using computer modeling? Don’t inventors already get considerable assistance from technology? What did the court say about all that?
Joining me to answer these questions is Robert A. McFarlane, an intellectual property litigator and registered patent attorney and partner with Hanson Bridgett LLP in San Francisco. Rob chairs the firm’s technology practice, co-chairs its IP practice, litigates and advises on a variety of IP matters in the U.S. and abroad, and teaches patent law at the University of California College of the Law San Francisco (formerly Hastings College of the Law). Rob earned his J.D. from the University of California College of Law San Francisco and his B.A.S. with departmental honors, in Industrial Engineering & Political Science from Stanford University.
I hope you enjoy the episode. I mean, we get to talk about everything from Tom Jefferson to monkeys with cameras. That's five-star material right there!
***********
This podcast is the audio companion to the Journal of Emerging Issues in Litigation. The Journal is a collaborative project between HB Litigation Conferences and the vLex Fastcase legal research family, which includes Full Court Press, Law Street Media, and Docket Alarm.
If you have comments, ideas, or wish to participate, please drop me a note at Editor@LitigationConferences.com.
Tom Hagy
Litigation Enthusiast and
Host of the Emerging Litigation Podcast
Home Page
LinkedIn
Welcome to the Emerging Litigation Podcast, a co-production of HB Litigation and critical legal content, custom content for law firms and litigation service providers, and the newly formed VLAC's Fast Case, your World of Legal Intelligence and our friends at LawStreetMedia. I'm your host, tom Hagee, litigation Content Producer and enthusiast and an average bongo player. Contact me if you have an idea for an episode. In addition to often being polite, I'm always looking for new twists on the law, whether it's a new regulation, legislation or an important new opinion, or it could be a development in the world that will test existing law or anything you're dying to share with other litigators, organizations or individuals. And, if you like what you hear, give us a rating. That always helps. And now here's today's episode.
Speaker 1:So this episode we're going to talk about what old Tom Jefferson had to say about patent law, among other things here. You just thought he was writing declarations of independence. That's all I thought he did. No, that's not true. I believe he was also president. And why can't a monkey copyright his own selfie? It seems patently unfair, or is it patently? We won't discuss that.
Speaker 1:And there are issues raised, of course, when artificial intelligence, when a machine driven by artificial intelligence, cooks up an idea. Specifically, there's one inventor who tried to throw a monkey wrench into how Patent and Trademark Office determines inventorship. We're going to talk about such AI engines, one named Dabbas, which stands for a device for autonomous bootstrapping of unified science. There you go, everybody should have one if you've known already. How does an invention come about? Can it be conceived solely in a human mind, and what did the federal court recently have to say about some of these issues?
Speaker 1:So joining me to discuss these questions and more, because there's always more, is Robert A McFarlane, an intellectual property litigator and registered patent attorney and a partner with Hanson Bridget LLP in San Francisco. Rob chairs the firm's technology practice and co-chairs its intellectual property practice. He focuses on patent infringement and licensing disputes, trade secret misappropriation all the good things technology related commercial matters and software performance disputes, counsel's clients on a range of IP matters. He's argued, of course, cases around the country, but he's also been retained as an expert on American patent law before the High Court of England and Wales and has handled matters before the International Trade Commission. Rob teaches patent law and is an edging professor at the University of California College of Law, san Francisco. It's formerly the Hastings College of Law. He previously taught patent litigation and IP courses at Golden Gate University School of Law and Copyright Law at Santa Clara University School of Law. He earned his JD from the University of California College of Law, san Francisco again formerly Hastings and his BAS with departmental honors in industrial engineering and political science from Stanford University.
Speaker 1:With that, here is my interview with Rob McFarling. Hope you enjoy it. Rob McFarling, thank you very much for speaking with me today. Great to be here. Before we get to the 21st century, tell us about Thomas Jefferson and Enrico Fermi.
Speaker 2:Well, sure, I mean it's interesting. Patent law always focuses on cutting edge issues. You know, today it's everything from artificial intelligence to the new CRISPR technology that is used to actually edit DNA sequences. So everyone thinks of these cutting edge deals. They think about practicing patent law, dealing with everything that's new. But really the foundations of our patent law are as old as our country. It's actually in the Constitution. All patent lawyers like to open their arguments to a jury with a reference to the Constitution. The Constitution actually says Congress shall have the power to promote the progress of science and useful arts by securing, for limited times, to authors and inventors the exclusive rights to their respective writings and discoveries. So it's in the Constitution and so it's.
Speaker 2:At that time, thomas Jefferson, all the Founding Fathers are thinking about patent law and a few other things, I guess, but of course for patent lawyers the most important thing is patent law. And so Thomas Jefferson actually wrote the first paragraph of the patent statute and it defines what you can patent, and in Thomas Jefferson's words, you could patent a new, and he said art, machine, manufacture or composition of matter. And that phrase is still in the patent statute, with one phrase, with one word changed. Now the patent statute says you can process any. You can patent any process, machine, manufacture or composition of matter. So those are the four things you can patent. So every utility patent falls within one of those four categories. And then the courts have carved out exceptions to those.
Speaker 2:You can't patent a law of nature, and you can't patent a natural phenomenon, and you can't patent an abstract idea. And so that brings you, I guess, brings us, to the question of Enrico Fermi. And so, under those, when you think about the law of nature, einstein could not patent. E equals mc squared. That is a law of nature. It's been there for eternity, it's there for anyone to discover and use, and so that law of nature can't be patented, but an application of that law of nature can be patented. And so that's where Enrico Fermi came in. Whereas Einstein could not patent the law of nature, enrico Fermi could and he did patent nuclear reactors that use that law of nature. And so that's what you can patent. You can patent an application of a law of nature and application of an abstract natural phenomenon. You just cannot patent the law of nature itself.
Speaker 1:So, coming forward to the case we're going to talk about, is Taylor versus Vidal was whether an AI system could be listed as an inventor. So if you could give us a snapshot of what the issue was there, including some of the facts presented in the parties, and who's Taylor what is? Is it Dubbus or Debus? How do you say that?
Speaker 2:I've heard it. Dabbus is how it's been.
Speaker 1:Dabbus okay. Option three Okay, go ahead. What's up with that?
Speaker 2:And so so Stephen Thaler is an AI research and artificial intelligence researcher and he's been filing this and other cases trying to establish what legal rights are in inventions and creations of AI systems. And so in this particular case he had two inventions by Dabbus, and Dabbus is an artificial intelligence engine and Dabbus stands for device for autonomous bootstrapping of unified science. That's why you call it acronym Dabbus. Under the US patent system, the inventor gets the patent, and up until AI system is always, the human being that conceives of an invention is awarded the patent. And so conception is the mental part of invention. It's creating in your mind and the mind of the inventor the idea of the patented invention as it is its full and complete form, as it's thereafter to be used, and so that mental act of conception generally earns a patent. And so the interesting thing about the Thaler case is Thaler put in affidavits into the US Patent and Trademark Office and the subsequent court that the inventions that issue had been conceived of entirely by Dabbas, this artificial intelligence engine without assistance or direction from a human being, and that was the record. That wasn't challenged. There was no further further facts of that, but for the court decision and the US PTO decision. That was the fact, and so they took it that Dabbas created these two inventions.
Speaker 2:And the two inventions that issue were two very different things. One was called a device and method for attracting attention. Think about, like if a fire alarm goes off in a building, you see a flashing light on the little fire alarm. But the idea of this invention is that if it's a very chaotic environment smoke, noise, fear if it's an actual fire, just a single flashing light doesn't attract human attention very well. And so the idea of this invention was that if you have the light source pulse at a specific frequency that's tuned to human brainwaves, it will attract human beings attention better. And so, according to the application, dabbas had used fractal mathematics, fractal dimensions, to synchronize the frequency of that flashing light to human brainwaves, thus making it better at attracting attention. But that idea of the frequency, the method of flashing, how it goes about interacting with human brainwaves, was calculated entirely by Dabbas, no human being intervention.
Speaker 2:So, according to Thaler, dabbas had invented that. And then other thing Dabbas invented this isn't quite as abstract, it's a food container, and again it uses fractal geometry. And so the idea of this if you have a six pack of Coca-Cola or six pack of Budweiser, you have to have the six ring binder that you put the cans in, so that creates waste. And so the idea here is well, if you have a fractal design for a container, they have edges that can then be used to interlock with each other, and so you can have interlocking food containers, and so that's that invention. And again, stephen Thaler said my AI engine, dabbas, had created this design all on its own, no intervention for me, so it should be the inventor. And so the issue in the Thaler versus Vidal case was can Dabbas and AI engine be an inventor?
Speaker 1:What arguments were presented in court for and against naming Dabbas as an inventor?
Speaker 2:For we talked about the Constitution a moment ago, and so that's what Stephen Thaler relied on. He said you go back to the idea of what is the purpose of patent law. What is the purpose of awarding patents? And it's to encourage inventions. It's to encourage people to invest in the R&D to create those inventions. And then it's the purpose is to permit their disclosure. If you look at the difference between patents and trade secrets trade secrets you can't disclose or they lose their value. Anybody else can copy a trade secret if you disclose it, whereas patents are give you rights against anyone in the marketplace. If you have a US patent, nobody can patent or can practice your patented invention within the United States. And so if you award a patent on a valuable invention, then it encourages the research and development to make that and it encourages its commercialization and it encourages its disclosure. And so Stephen Thaler said look, this is a valuable invention. I'm using an AI engine. It has created this valuable invention. I want to commercialize it, I want to disclose it so other people can build on it, and that's what the patent system is for. And so it makes sense to award a patent, even though the inventor is an AI engine. So very policy, very policy based arguments.
Speaker 2:The patent, the trademark office, the USPTO argued that's all nice, that's all fine. We're not arguing with the statutory purpose of the constitutional purpose of patent law. But the patent statute says whoever invents. All the other patent statutes refer to the inventor as an individual, they refer to it with personal pronouns and the inventor is he or she or they. If there's more than one inventor, you cannot, a corporation cannot be an inventor. All the patent rights has to go back to the human inventor, the individual inventor, or if it's a joint invention, the handful of inventors. There's nothing in the patent statute that says a non-human can be an inventor. So the patent on trademark office, the government just argued kind of dryly, based on the language of the statutes. It says a person for all intents and purposes, a dabbist is not a person, so dabbist cannot be an inventor. Those were the two arguments. On the other, side.
Speaker 1:Okay, so where did the federal circuit come down on this?
Speaker 2:They came down with the government and they said we're not going to get into the metaphysics of the nature of an inventor, the nature of an artificial intelligence, we're not going to answer any question of what rights, if any, an AI system might have. We're just going to look at the statute. And they said, going through all the statutes, they kind of followed the government's argument it refers to a person, it refers to a natural person, and so that's what they ruled, just as a matter of very. They didn't answer the big questions, it's just strictly what does the statute say? What is Congress written in the patent statutes? And it requires a natural person to be an inventor.
Speaker 1:Okay, so that's inventorship, but you also wrote that the court itself invented something, a new category of inventions. What do you mean by that?
Speaker 2:Well in this case. So if you have a patent that has an artificial engine or artificial intelligence engine as a inventor or the inventor, under this statute it just can't be patent. So it's kind of an orphan category of inventors. It may be a valuable invention, but it cannot be protected through the patent system, and so really, this decision gives rise to the question of okay, if you cannot patent that invention, how do you go about protecting it? How do you go about commercializing it without the ability to patent it? So really, it gives rise to that whole new category of things that companies will have to think about.
Speaker 1:Okay. So when I talked to you about this before, I was wondering. Not that anybody would endorse this, certainly not URI but you're supposed to divulge everybody's role in the process that you use to come up with an invention or development invention. But what's the stop somebody from saying AI didn't have a role or just not mentioning AI? It was used?
Speaker 2:I guess two things really. One is, whenever you file patent application, you do sign an oath, under penalty of perjury, that you are in fact the first true and sole inventor. And so if, in this case, you know Steven Thaler saying that artificial intelligence agent created this, conceived of this invention, you know, if you signed that oath it wouldn't be true. So you know. But there is one issue there. And the second issue is that if the patent does not name the correct inventor, then it can be invalid. And so if you had a patent that actually was conceived of by an artificial intelligence engine without input from a human, you know, and you put a human being's name on there, that human is not technically the inventor, If you sued anyone on that patent, those facts would come out during the litigation and then the accused infringer would be able to argue that the patent is invalid for failure to name the correct inventor.
Speaker 1:If you didn't show your work already. You really have to show your work if somebody challenges Right. What do you see the pros and cons of this ruling? Do you think this is good for innovation?
Speaker 2:Well, I think it's a limited ruling in its scope because you know the facts before the court were very, you know, kind of artificially presented by Steven Thaler. You know he said there was no human input into this invention. And so you know on that, but that doesn't answer, is okay. If a researcher is using artificial intelligence as a tool during their research, you know how far can you go in patenting that. And I think as long as the human you know, like today you've got, you know any number of inventions where you're using computer modeling you know just anything from a wind tunnel to, you know, computer modeling and pharmaceuticals, anything and the inventor is actually using those pieces of software as a tool for their inventive process. And so there's nothing in Thaler that says you can't do that using AI, as long as the inventor is still conceiving the idea, you know, directing the AI and its research, the human is still, you know, the inventor, under the statute, I would argue, and so Thaler applies to the extreme positions and so it doesn't really take away the ability to patent inventions, that where human beings are using AI as a tool. You know, in that narrow context this is a good decision or a bad decision and we really have to go back to what we talked about.
Speaker 2:What is the purpose of patent law? It's to advance technology, it's to encourage research and development, it is to encourage the disclosure of inventions, the commercializations of inventions. And so there's always a balance there. You want to give enough patent protection that encourages that those activities, encourages research and development, but you don't want to give so much patent protection that it discourages other companies from conducting research, from commercializing property. And so if you go back, like around the turn of our century, around the 2000s, there was the internet boom and out of the internet boom became large numbers of very basic patents having to do with internet, having to do with computer networks, and then all of those companies or many of those companies you know, dissolved and failed in the dot-com bust, but they're platinist states persisted and so a lot of those patents were acquired by non-practicing entities, targets of non-practicing entities called patent trolls. That's a little bit of a derogatory term in some circles, but you can call them non-practice entities, npes or patent trolls.
Speaker 2:And so there was a lot of writing, a lot of argument, even in Justice Breyer wrote some concurring opinions talking about this issue where you awarded so many what were deemed low quality patents that came out that actually discouraged inventions because, you know, non-practicing entities particularly could then sue practicing companies and extract royalties from them based on patents of limited value. And so the balance is okay. You don't want to give so many patents that they become a drag on companies that are creating products or creating services. That's too much patent protection, too little patent protection and undermines research and development, research. And so in this, you know, you have to look at how much. How do you get the balance right here? You know, is denying from a policy perspective, is denying all patent ability of things, of inventions created solely by AI. You know, is that too far? Or if, as long as the person who's using the AI is directing it, then you can give and using it a tool, then you can give the patent to that person.
Speaker 2:And so I think you know at the end of the day, you know, I think it's probably going to end up as a balance between those extremes of denying it. Yeah, let me let me say that part again. Sure, you know, at the end of the day it's balancing the protect, protection of the patentable inventions. I think in almost all cases, you're gonna have a person directing the AI and directing what the AI is doing and using it as a Tool, just like you're using software modeling technology today, and then that person will still be an inventor. If there are situations, as AI advances, that it actually is Sol inventor on patentable projects, then I think you have to look at, you know, does the patent stats you have to be tweaked a little bit so that the AI can technically be the inventor and then the ownership of the patentable invention rests with the individual or the company that is controlling funding, operating the AI.
Speaker 1:Okay, well, you led into the next question around Working under the direction of an inventor, and then you also had brought up the, the concept of work for hire In the copyright realm. What more can you say about those?
Speaker 2:Right, and this is what's kind of interesting and one of the differences between patent law and copyright law. So in copyright law if you are an employee working for a company, you know it's your the copyrightable work that you create in the course of your appointment. It's considered a work for hire and belongs to your employer. Patent law is very different. There is no concept of work for hire and patent law the Property ownership of the invention rests with the human that creates the invention or conceives of the invention, absent a written assignment. And so if you look at tech companies, there's always an inventor Assignment form. They science usually an employee handbook and the employment contract and says any inventions you conceive of during the course of your employment here belong to the company and you do. You hereby assign those Inventions to the company and you will assign. You know you'll basically execute an assignment form if that's when the time comes for it.
Speaker 1:Right, so that the human being is still the inventor, but the ownership goes can go to the company or to whomever they assign.
Speaker 2:Exactly, I got you and so the question of inventorship is who actually conceived of the invention, who actually conceived of the idea? And then the ownership is who has legal title to that invention? Has the Inventor effectively assigned it to a company? If they have, has that company assigned it to somebody else, or who actually has that legal title?
Speaker 1:I got you All right. So, as you said, this was the. This case, taylor, was in the patent context. Did you want to say more about AI in the copyright area, because we're seeing seeing why that come up with with authors and Folks like that. So how might they were be applied in that context? Or, and, and the most importantly, I want to ask about the monkey selfie. So if you could get into that.
Speaker 2:Well it's. It's exactly the same issues that arose in this and in the Taylor case. You have an artificial intelligence engine creating a patent bill invention. But there are instances where an artificial intelligence engine generates Work, a work that if it was created by a person would be Copyrightable. And Taylor, you know, has really been trying to push the the law in this area, not only in patents but in copyrights. And he had another artificial intelligence engine that created an, an image called the entrance of paradise. It's a very beautiful floral scene. It kind of looks like a railway spur Kind of disappearing into this very peaceful setting. And he tried to get a copyright on that, try to register it with the US copyright office. And the copyright office Came to a very similar conclusion, said under the copyright statute, you have to have a human author. You cannot Register a copyright for something that is not created by a human being and so that the, the history on that is originally photographs.
Speaker 2:There was a big dispute back in the 1800s whether you could Copyright a photograph, because the theory was that you know, a photographer is just putting the magic box on its pedestal and, you know, opening the shutter right and it's just a mechanical act that there's no act of authorship in that. I think anybody who's ever taken a picture knows that more goes into taking a picture in terms of framing. You know Opposition, and and so the courts ultimately decided that, absolutely, you know, it's that's why people you know We'll pay thousands of dollars for beautiful or artistic photograph. So the courts eventually came to that view that said well, human, there is a human author to a photograph if a human takes an off a picture of that. And so the case came down.
Speaker 2:The monkey selfie case just a few years ago came down and Answered the question of, you know, whether you have to have a human author. And so this, you know, made all the rounds on the internet and many, many means were done about it. But you've gotten the Ruto and the cac monkey Took a wildlife photographer's camera and snapped a really nice selfie of himself. It's very dramatic, very, very handsome photograph of this macaque monkey, and so a wildlife, you know welfare organization took up his case and said you know, naruto should be the author. He took this picture, naruto should have the copyright on this. And the the Ninth Circuit Court of Appeals here in California just came down said well, naruto is not a human. You have to have a human as an author or you can't get a copyright. And so you know again, just very cut and dry, it's just a line in the sand that you can't, you cannot cross. And so you know that that decision was relied on when Thaler tried to Copyright his image of the entrance to paradise, the copyright office, and later the courts, just said look, there's no human author, you cannot get a copyright in the absence of a human author. And then so the subsequent decisions in the copyright office kind of take up the issue. Okay, so you have a situation where a person is using AI as a tool, person's relying on AI for some things but contributing other things.
Speaker 2:And so there was a case that had to do with an author named Christina Kastanova, who had created a graphic novel called Zaria of the dawn, and to do that, you know, all of the images in the graphic novel had been created by an artificial intelligence.
Speaker 2:You know one of the one of the I think I can't remember which one it was, but was, you know, one of the standard image generating a eyes, and she worked for months, you know, really, to get the, the text with the, the prompts right and everything just to get exactly the images she wanted. And then she took those images and arranged them into a graphic novel and then put her own Human-generated speech bubbles into them. And so what the copyright office said is, look, the images themselves lack human authorship, so you cannot copyright the images themselves. But you, christina Kastanova, took an authorship step in arranging them, curating them, putting them into the form, the order, the arrangements, the sizing, and you put the speech captions into them, and so there is a limited copyright then on the compilation as she put it together, and then on the human-generated speech, and so again there's kind of the balance of anything that's purely generated by AI can't be copyrighted, but the humans basically steps of authorship. The humans' compilation and use of that is subject to a more limited copyright protection Right.
Speaker 1:Okay, I kind of remember that coming up in the context of like directories and things you know and information's out there. But you've also added value to it, you've organized it and all that. So okay, well, that's, it's interesting. I'll find that case and put a link to it. But I mean, we are asking some big questions here, like what is a mind? What constitutes a mind? Discussing photographs is really a good one, I think as well. I mean, you're, yeah, you are taking nature and yet you're transferring and doing things and enhancing it, so yeah, so that that I think that must be a big part of it, as you said, with the, with the book, the graphic novel, all the enhancement and organizing that goes on. But I guess that is kind of the question of what is a mind Is that is that kind of the big question.
Speaker 2:Absolutely, absolutely. And you know, you've got, you know the courts right now are taking very, you know, dry statutory interpretation. What does the statute say? The statutes require a human author, the statutes require a human inventor. But the bigger question, you know, is kind of the thing of science fiction, you know that's what a lot of people are really interested in is. Okay, you have these autonomous systems and they're becoming more autonomous. When do they, you know, cross the line into being more than just a computer tool, to be more than just software? I think a lot of those questions will have to be answered at some point as they become more and more autonomous, more and more sophisticated.
Speaker 2:You know it always brings to mind, you know, like it was probably in the early 90s. There was an episode of Star Trek, the next generation, where Lieutenant Commander Data, the Android, came across, came to the attention of a Starfleet researcher and he wanted to basically download data and dissect him and see how he worked, and so there was actually a court proceeding, a judicial proceeding, where they I think the episode was called the measure of a man where they decided where their data was on autonomous entity, who had legal rights, or was it just a piece of hardware that could be disassembled and downloaded. So you know, luckily for all the Star Trek fans, he was not disassembled and downloaded and they decided that he had legal rights. But you know, at some point you know right now, I think you know the software.
Speaker 2:You know artificial intelligence engines don't know what they're doing. You know they're very, very, very sophisticated mathematical programs that can imitate human speech and imitate, you know, text. You know you think of chat, gbt and the very, very sophisticated things it's doing. But it doesn't know what it's doing. It is working on mathematical rules, very, very complicated mathematical relationships, you know. But is an AI system sentient yet? And I think we're a long way from that? But you're right, I think at some point that will have to be answered.
Speaker 1:Yeah, pay attention to sci-fi. And two, I had a law professor on one podcast and a judge on another and they're talking about you know, can you plug case facts and law into a machine and have it spit out a ruling? Of course, you know. In both cases I said of course a human being will want to review that. And I'm blown away by you know, google Bard and his grammar is excellent. Yeah, ask your questions. It's like wow, it does sound like somebody actually wrote that. It's pretty good.
Speaker 2:I mean, if you train those AI engines on good data, you know they can replicate the quality data that you've provided, but they, again, they don't know. I mean, like ChatGbt predicts what the next token or the next word should be, but it doesn't know the meaning of it. It just knows, based on the data that I've reviewed and statistically and with the temperature adjusted, and what the next word should be, and it does that one word at a time, but it doesn't know what that word means. So that's how you know.
Speaker 2:There was another decision that came out last week of an attorney relying on ChatGbt and filing a brief that had essentially made up case sites in it. You know, at this point, you know the ChatGbt knows that the next word should be part of the site, but it doesn't know what a case says, it doesn't know if it's correct, it doesn't know if it's true, and so I think we're you know we're a long way from having all the lawyers put out of business by ChatGbt, because we actually know, we vet something, we know if it's true, we know if it's accurate, we know what we're saying and, at least at this point, chatgbt and anything like it doesn't actually know what it's saying.
Speaker 1:Yeah, a conclusion that I've come to, speaking of lawyers using it or anybody using it. When you point it at your own data and it can organize your own data and it's predicting based on what you've done before, I mean that's that seems more reliable and less risky. So it's like, rather than you know, I've had it pointed articles, I've written, it does a nice job putting them together, but I know where everything came from. So and I would always disclose that I use that assistance and I did look for cases. Somebody asked me about a couple of cases once when I was looking for them and it was ChatGbt. It spit out like three or four cases and they sounded like cases and they said where they were from. They did not exist. They simply did not exist, you know, and a paralegal was who was really good at this I'm good at legal research and this paralegal is way better and and they were being kind they said, well, maybe it's a, it's a small court, maybe in a county. That's what I'm like. No, it didn't happen.
Speaker 2:That is a risk and that is a big risk right now. Yeah, yeah, yeah.
Speaker 1:But I think it's always that I know there are people coming up with rules, but always check it on the way out, always, absolutely.
Speaker 2:Absolutely.
Speaker 1:Well before we are there any other recent rulings that you wanted to mention that anybody should know about. I think we just kind of put that in there.
Speaker 2:Oh sure, I think. I think the other big issue, like all of the cases we've been talking about now during our conversation, are how does patent law and copyright law apply to the outputs of an AI system? But there's a lot of cases right now about how particularly copyright law applies to the inputs to an AI system. So there's class actions. Everyone from John Grisham to Sarah Silverman have filed lawsuits about their copyrighted material being used to train artificial intelligence, and so the theory is look, I'm the author of 20 books 25 books I think John Grisham is one of the, so it's probably a lot more than 20 or 25.
Speaker 2:But Sarah Silverman was the author of, I think, one prominent book that she said was basically copied input into the AI engine and then the AI engine can basically appropriate her creative creativity, her style, her words and then create material that will compete with her, and so a lot of those cases are very early on saying look, is that a copyright infringement? Are you violating those authors' copyrights when you're digesting their information, their works and then outputting a novel in the style of John Grisham? And so those are going to be working their way through the courts and through the appellate courts, I would say over the next two or three or four years, and those questions will have to be answered. Right now, just in the very early motions to dismiss stage, the courts are narrowing those cases and trying to determine okay, have you alleged just the elements of a copyright infringement here, and then that's going to be decided in the near term in the trial courts and then the appellate courts will have to kind of decide what the law is and should be on those.
Speaker 1:Yeah, yeah, a mimicking part. And when I was in college, we had a writing class that one of the assignments was to write like a famous author. So write something like Kurt Vonnegut would write, and so we would do that, and then some people were really, really good at it. But I'm sure you could train. You could train AI to look at the speech patterns, like you say, just predicting words. Well, this is typically the kind of words Kurt Vonnegut would use and his head would spin if he were still alive. But yeah, and also, too, when they pull things from the Internet, there's enough. There's enough of Sarah Silverman interviews and her stand up and stuff and excerpts of the book that were mentioned in reviews. Yeah, they could piece them together. So it's crazy. And I think too, wasn't that a part of the actors saying you know they could use your image forever? Oh, I think it was extras background people, correct?
Speaker 2:Right, Like clip art. We'll just stick you back there now, Exactly, Exactly, and so the description kind of raises the types of words, the questions that are being asked. So if you have an AI engine and it you copy a copyrighted work and input it into the engine, you know that's a, you know you've copied it, so that's potential copyright infringement and the issue there. You know there's some cases going back to the 90s that have to do with video game programs and whether you can make an intermediate copy of a piece of software in order to create compatible games and things like that, and so the courts are looking at whether that intermediate copying is a copyright infringement in the case of training and AI. And so you've got that input question and then the output question.
Speaker 2:The big threshold they're going to have to pass is typically for a copyright infringement there has to be substantial similarity. So if you take a copyrighted painting and copy it, you know, just copy all of its elements but create a reproduction of it, that's a copyright infringement. But if you don't take those elements and, like you were just saying, create a painting in the style of another artist but you're not incorporating specific elements, that's much harder to show. That's copyright infringement because it doesn't have substantial similarity. And so those are the questions here Is the output of the AI substantially similar to the copyrighted work that you're alleging is infringed? And so every, you know, all of those are going to have to be answered based on the facts of the individual cases that are working their way through the courts.
Speaker 2:And then you know, at the beginning of the internet, you know, you, the courts created the whole new set of laws on, you know, copyright over the internet, privacy over the internet, all these new and novel issues when the internet came up, and there's really that broad of an arrangement of laws now that have to be determined with AI, you know. But everything from training it, how it's written and then what the outputs, you know, how do the existing laws apply to those and to the existing laws need to be tweaked to address artificial intelligence, Because it really is. You know, such a wholesale change in the way content can be generated, the way that copyright material can be used to generate new content. Now does the statute need to be tweaked a little bit? So I think we'll be looking at that for, you know, many years.
Speaker 1:Yeah, I'm sure it's going to be changing. There's going to be a lot coming down in the coming months and years, so I'll have to get you back on here because I know we'll have more to talk about Fantastic.
Speaker 2:Yeah, oh, I'd love it.
Speaker 1:Yeah, all right. Well, rob McFarlane, thank you very much for talking with me about this today. It's a really interesting subject.
Speaker 2:It's been my pleasure. Thank you for having me.
Speaker 1:That concludes this episode of the Emerging Litigation Podcast, the co-production of HB Litigation, Critical Legal Content, VLex Fast Case and our friends at LostG Media. I'm Tom Higgy, your host, which would explain why I'm talking. Please feel free to reach out to me if you have ideas for a future episode and don't hesitate to share this with clients, colleagues, friends, animals you may have left at home, teenagers you've irresponsibly left unsupervised, and certain classifications of fruits and vegetables, and if you feel so moved, please give us a rating. Those always help. Thank you for listening.