You’re invited to join a confidential discussion on Working Effectively with Legal Recruiters. More events to accelerate your career are coming soon.
Welcome to AccelPro IP Law, where we provide expert interviews and coaching to accelerate your professional development. Today we’re featuring a conversation with Matthew Carey, Partner at Marshall Gerstein.
In this interview, Carey discusses artificial intelligence image generation and copyright, artificial intelligence and the patent system and the impact of AI on legal decisions.
We then cover his career path and transition from computer engineering to IP law and the responsibilities of being a hiring partner at a law firm. One of Carey’s main areas of focus in this conversation is evaluating the different ways that AI-generated content interacts with existing legal precedents and how existing interpretations could change in the future.
Listen on Apple Podcasts, Spotify and YouTube.
Interview References:
Matthew Carey’s Marshall Gerstein profile.
2:53 | Stephen Thaler v. Shira Perlmutter, No. 22-1564, (2023). United States District Court for the District of Columbia.
4:06 | Copyright Act of 1909, (1909).
4:23 | Brittain, Blake. US Copyright Office denies protection for another AI-created image. (6 September, 2023). Reuters.
13:40 | USPTO Initiatives: Artificial Intelligence. United States Patent and Trademark Office.
14:47 | Alice Corporation v. CLS Bank International, No. 13-298, (2014). Supreme Court of the United States.
TRANSCRIPT
I. AI IMAGE GENERATION & COPYRIGHT
Neal Ungerleider, Host: Can you give our listeners some background on Marshall Gerstein, and your role there?
Matthew Carey: We're an IP boutique based in Chicago. We handle all aspects of IP law–so patents, trademarks, copyrights. We have an IP litigation team and also do transactions. We have over 100 professionals, and that ranges from attorneys to patent agents to technical specialists.
Within our patent department, we have different practice groups that are segmented based on subject matter. So, we have an electrical software group, a biotech group, a mechanical group, a chemical group.
Our clients, are mainly larger Fortune 500 companies, but we also represent smaller startups and universities. My background is in Computer Engineering. so I'm part of the electrical and software practice group, which has about 35 to 40 professionals at the moment. And I also became our firm's hiring partner recently, which has been a rewarding experience thus far.
NU: For our listeners who may not be familiar: Can you tell us a little bit about the Thaler versus Perlmutter decision, and why it's so important?
MC: So an artist named Stephen Thaler has developed this AI platform that he called the “Creativity Machine.” And this platform output an image that Stephen titled “A Recent Entrance to Paradise.”
And it's a rather lovely image depicting railroad tracks that go under a tunnel. It has flowers; it has leaves. It's nice to look at. He attempted to register this artwork with the Copyright Office. What's important here is that this image was generated autonomously by this Creativity Machine. And because he owned the Creativity Machine, he was seeking to register the copyright on the work. And by virtue of the ownership, the copyright would transfer to him.
So, the Copyright Office initially denied registration of this work. That decision went to the U.S. District Court, which sided with the Copyright Office. It denied the registration. It found that human authorship has been a fundamental requirement for copyright under the law, as well as past precedent. And noted that the text of the Copyright Act protects “original works of authorship”, which implies a human creator.
The main takeaway from this case, really, is that works that are generated autonomously by AI are currently not eligible for copyright protection.
NU: And there was a decision in Fall 2023 by the U.S. Copyright Office around another AI generated image that's somewhat related to Thaler. Can you tell us more about that?
MC: There was an artist named Jason Allen who used Midjourney, which is an AI generator. He input a prompt into the program. MidJourney outputs this image. And it’s an intricate image. It depicts this ballroom with figures wearing fancy clothes.
There's this portal in the middle of the image that looks onto a mountain scene. Jason used existing editing tools, like Photoshop. He added additional details to the image. He made the figures a little bit more clear. And he actually entered this final work in the Colorado State Fair's annual fine art competition.
It won the competition as the first AI assisted work to win a reputable artwork competition. He then sought to copyright this image. The board of the Copyright Office reviewed this petition, or reviewed the application, and concluded that the original image remains in substantial form in the final work and therefore must be disclaimed.
Jason, for some reason, did not want to disclaim this AI contributed portion. So the board ended up denying the copyright registration. I think the main takeaway here is that the copyright office noted that modifications made to an AI generated image may rise to the level of copyrightable authorship. So whatever that delta is from the original output, to what the artists changed in the image, can achieve copyrightability.
NU: How are images generated by AI? For instance, by Midjourney or DALL-E? How does that process work?
MC: A user can go in and input a prompt. And these prompts can be as vague or as detailed as the person wants. The model then takes this prompt, interprets it; figures out what it's saying. And then, it's this black box depending on how the model is trained, that then outputs images. And it's usually a set of images, and the idea is that the images align with what the user is looking to have the model create.
NU: And from your point of view: What are the upsides and downsides of granting copyrights to AI generated works?
MC: I think some clear upsides would be the idea of increasing innovation and development in these generative AI systems. It could spur development of AI capabilities to augment existing human creativity.
So, my nine year old comes into my office, and asks me to help her generate pictures of a golden retriever kicking a ball around in the backyard. And we input that prompt, and DALL-E outputs images. One or two of them would be good. The other may be bad, or not relevant. But I think with the existing image generators, the human has little control over the final output. I think there is an opportunity for these technologies to develop so that humans have a little bit more control from inputting the prompt to the final image.
I saw a Twitter thread where this person had a prompt along the lines of, “Generate a set of eight-bit images of famous structures around the world.” And the one that really struck me was this output of The Bean in Chicago. It's a really famous structure in Millennium Park, and the image was an eight-bit image of The Bean that showed it from a common perspective, showing reflections of people and the skyline of Chicago on Michigan Avenue
So, this one struck me. I think AI can be used more as an image editing tool, like Photoshop. Or maybe even a camera where the human, through the prompt, sets the stage and the environment. And the AI “captures”, so to speak, that image.
I think by doing that, these technologies can allow, again, the user a little bit more control. Which I think helps the argument for granting copyrightability to these AI-generated works, because there is this greater involvement in human authorship.
Some of the downsides: I think it can be confusing if the Copyright Office is requiring authors to disclaim AI generated portions. I think it creates challenges in evaluating respective contributions of humans and machines, to determine fair attribution and ownership. Some of the main aims of copyright are to incentivize creativity and provide access to knowledge. Perhaps protecting AI works would create protections over works that are not fully conceived by human minds. Which may be contrary to some of these aims.
NU: So the issue of copyrighting AI generated works in the United States is still something that the courts are working through. Is that a fair way of putting it?
MC: I think so. What's clear now, at least as the law currently stands, is that fully generated works by AI are not eligible for copyright protection. The Copyright Office has notified claimants that they have to disclaim portions of an image that are generated by AI. And the remaining portion is eligible for protection.
But these AI generators are a new technology. And as technology has evolved, so does the law. So, I expect moving forward, the Copyright Office–and maybe the courts–to zero in on, “Okay, what is protectable here? How do we balance these AI generated outputs with the more human, controlled pieces of the artwork?”
—
II. ARTIFICIAL INTELLIGENCE & THE PATENT SYSTEM
NU: I want to shift gears a little bit and discuss patents. What are some of the considerations around listing a generative artificial intelligence as an inventor, or joint inventor?
MC: So U.S. patent law requires that to be listed as a joint inventor of a patent application, you need to have contributed to the conception of the claimed idea. This creates two schools of thought. So can an AI system conceive of an idea?
The first school of thought is if an AI, if an AI system, can be considered to have conceived of an idea that becomes part of a patent claim, then that AI system needs to be listed as an inventor. Of course, this contribution would need to be significant in quality. And again, I think there's a possibility that the USPTO may require applicants to notify the patent office of whether an AI system was involved in the conception of an idea.
And the second school of thought is that an AI system is merely a tool that is not capable of “conception outside of a human input.” So in this regard, the AI system is treated more as a tool in invention discovery, to aid the human inventor. And according to this school of thought, the AI system would not need to be listed as an inventor.
NU: And what are some ways that US patent law can help ensure that AI generated inventions benefit society similarly to human created ones?
MC: So, generally the patent system incentivizes innovation by granting these limited monopolies - i.e. patents of new and useful inventions. When a patent is granted, it becomes public. So anyone can go out there and review the patent, and innovate on that. Innovate on existing solutions - and, and eventually drive innovation forward.
With the introduction of AI, I think you want to maintain these same human driven incentives. Perhaps with some modifications - and I'm getting into pure speculation here – but perhaps one way that you could balance AI generated contributions to benefit society would be to expedite examinations of applications for AI inventions that address humanitarian needs.
So, for example, in areas like medicine and the environment. This could incentivize research for these types of initiatives. There's also been talk about heightening the non-obviousness standard for AI inventions. So in patent law now, a new invention needs to be non-obvious in view of any existing prior art. Well, if you heighten this non-obvious standard, you could perhaps prevent monopolizing more mundane outputs that AI's come up with.
NU: Can you talk about what the U. S. Executive Order on AI means from a legal perspective?
MC: The USPTO is keenly aware of developments in AI. And they recently stated that, “We want to ensure that we incentivize innovation in critical and emerging technologies such as AI, but also advancing national security and AI safety.”
So, this executive order directs the USPTO in two relevant respects. The first is inventorship, and the second is patent eligibility. Regarding the first, as we discussed, the USPTO has made it clear that listing only an AI as an inventor is not allowed. So, the question really becomes: Can and how can an AI be involved in the inventive process? The Executive Order, I think, will be helpful because it directs the USPTO to produce examples of AI's different roles, and explain how inventorship issues should be analyzed in each one. I think it will give practitioners a little bit more guidance on: Can an AI be a joint inventor of a patent application?
The second respect is patent eligibility. So, nine years ago, the Supreme Court came out with its Alice decision–which basically said that known activities that are done faster or more efficiently by a computer are not patent eligible. So with the introduction of AI, how does Alice align with these new technologies?
If you're just applying a computer to do X, that's not patentable. And now if you're applying AI to do X, is that still not patentable? The USPTO will hopefully release guidance on ways that AI represents an improvement on existing technology. Or perhaps, the USPTO will address questions like when the use of AI improves on an existing process in such a way that it becomes a patent eligible invention.
—
III. FROM COMPUTER ENGINEERING TO IP LAW
NU: I also want to ask you questions about your career path. You mentioned coming from a computer engineering background. What led you to intellectual property law?
MC: It's a winding path. So growing up, I didn't have any lawyers in my family. I didn't have a lawyer mentor in my early years. I remember senior year of high school, I joined my high school's law team. I enjoyed that experience.
As a senior in high school, I had already specified my major. I went to the University of Illinois. I majored in computer engineering, which is very different from the law. I never really knew that patent law was an option for engineers until after I graduated. I started working as a consultant. I was traveling around the country. It was a grind. I knew that I didn't want to do that.
And at that point, I started talking to some of my classmates from college, and they mentioned that they were going to law school. I did some research on that. I figured out that the study of IP law or the practice of IP law was a good way to combine my engineering experience with law. And I thought I found that very intriguing.
I went to law school, graduated in 2008 and then moved to D.C., started out at a firm out in D.C. and then worked my way back to Chicago a few years later. And here I am. And I think I always like to tell people that, what I do now is probably 75 percent engineering and 25 percent law. And it's been rewarding because I'm able to flex this legal muscle - that growing up, I never knew I had. And it gets me out of my strictly engineering brain.
NU: And how does coming from an engineering background help you as a lawyer?
MC: Typically, all patent lawyers have what I like to call the engineering brain. And I think it's crucial for practicing patent attorneys to be able to understand a client's technology. Or at least the ability to learn about a new technology. By virtue of being a patent attorney, we're always exposed to new innovations. So we take our building blocks and figure out how a new technology works. How a new technology perhaps is different from existing technologies. I think the big bar to this profession is being able to understand these technologies and to counsel clients on how to best protect them.
NU: What advice do you have for new hiring partners to succeed in their roles?
MC: So there are two main determinations whether to hire someone. First are they going to fit into our culture? We are an IP boutique. We're midsize. I think we are different from these larger general practice firms, but we're also different from these smaller firms. And, we have our culture and we like to see that a person could fit into that culture.
The second is whether someone has a relevant technical background to fit into these individual practice groups. We have the software practice group, so we're seeing a lot of AI based applications that our clients are interested in. I think figuring out which candidates have those relevant backgrounds to be able to talk the talk, and to ultimately help our clients protect these new technologies is important.
NU: Can you give an example of how you've relied on peers and not mentors or bosses inside or outside your organization to deal with tough situations?
MC: One of the big struggles that we’ve touched on is this uncertainty with patent eligibility, specifically with AI based concepts. I think this executive order will hopefully allow the USPTO to provide us with some guidance as to what is patentable; how do we structure a patent application to clear this subject matter eligibility hurdle.
I think it's helpful, even though I'm in the weeds on a lot of this, to talk to some of our junior folks or my peers to get their insight as to different strategies that they take. Different ways that they advise their clients on what they think is patentable; what they think is maybe not patentable.
Also in an examiner interview, I ask the examiner for their input on things that we could do to move the patent application in a positive direction. I don't have all the answers. But it's helpful for me to seek out conversations with other folks to help me advise clients more accurately and succinctly.
Listen on Apple Podcasts, Spotify and YouTube.
This AccelPro audio transcript has been edited and organized for clarity. This interview was recorded on November 8, 2023.
AccelPro’s expert interviews and coaching accelerate your professional development. Our mission is to improve your day-to-day job performance and make your career goals achievable.
Please send your comments and career questions to questions@joinaccelpro.com. You can also call us at 614-642-2235.
If your colleagues in any sector of the IP law field might be interested, please let them know about AccelPro. As our community grows, it grows more useful for its members.
