For years, the common wisdom about robots was that they’re taking over repetitive, routine jobs. But at least they’re not creative. We’ll always have art and music to ourselves, right? Wrong. Douglas Eck works at Google. He created the Magenta project, dedicated to using artificial intelligence, or AI, to make art and music. Eck is using a form of AI called Machine Learning. His software analyzed thousands of sketches done by people; and now, it knows how to complete a drawing itself.
Digital technology has given rise to a host of novel questions concerning the authorship, ownership, and exploitation of creative works, from the right to re-sell digital copies to the copyright status of works produced by artificial-intelligence agents. But a new legislative skirmish in New York State could take the debate beyond the realm of copyright into the realm of privacy and the right of publicity.
Assembly bill A.8155B would create a new right of publicity for individuals concerning the use of their likeness in a “digital replica.” In what is believed to be the first such legislative effort by a state, the bill is meant to prohibit the use of “face-swapping” artificial intelligence technology to overlay an individual’s face onto another actor’s body without the individual’s consent, particularly for pornographic purposes.
According to the bill, “Use of a digital replica of an individual shall constitute a violation if done without the consent of the individual if the use is in an audiovisual pornographic work in a manner that is intended to create and that does create the impression that the individual represented by the digital replica is performing.”
The bill is strongly backed by the Screen Actors Guild-American Federation of TV and Radio Actors (SAG-AFTRA), which claims it is necessary to combat the growing scourge of “deep fake” videos, in which well-known individual are made to appear to be performing pornographic scenes.
“Individuals turn to image rights to sue corporations that use social media accounts or publicly available images to promote products or services without consent or compensation. These rights will also provide individuals, often women, relief if they are inserted into commercialized Deepfake sex scenes.,” SAG-AFTRA said in a statement supporting the measure.
With time in the legislative session running out, however, the major studios and the Motion Picture Association of America (MPAA) have mounted an all-hands effort to block the bill, according to the Hollywood Reporter, claiming the bill’s imprecise language could limit the production of biographical films of real-life individuals and chill technology innovation.
“If adopted, this legislation would interfere with the right and ability of companies like ours to tell stories about real people and events,” the Walt Disney Co. wrote in a letter the bill’s author. “Unfortunately, the proposed bill would transform New York from a jurisdiction that is friendly to and protective of such expressive endeavors to one in which they become encumbered by uncertainty and risk.”
In a separate memorandum, NBCUniversal warned, “The bill creates an unprecedented new category of protection for “digital replicas” of
living or deceased individuals. These provisions have potentially far-reaching implications, yet there is scant time left in the session for New York’s legislators to explore and consider them.”
The bill is still pending and it’s fate is uncertain at this point. Either way, though, it’s unlikely to be the last word in the debate over the uses (and misuses) of face-swapping technology and other forms of artificial intelligence in the creation of media content.
We’ll tackle some of those questions at the upcoming RightsTech Summit , at a panel titled What to Make of Machine-Made Art? Click here for more information on the summit, and for information on how to register.
As a formal matter, the litigation in People for the Ethical Treatment of Animals v. David Slater, the “monkey-selfie case” is now over. Although the parties nominally settled the lawsuit last year, the U.S. Ninth Circuit Court of Appeals put an exclamation point on it last month by issuing a formal ruling and opinion anyway upholding the district court’s ruling in favor of the human photographer Slater that was under appeal at the time of the settlement.
While the court conceded that, under the Ninth Circuit’s peculiar precedent in Cetacean Community v. Bush, animals like Naruto, the crested macaque who took the famous photos and on whose behalf PETA nominally had brought the case, may have standing to bring a lawsuit in human courts, they do not have statutory standing under the Copyright Act to bring an action for infringement:
We must determine whether a monkey may sue humans, corporations, and companies for damages and injunctive relief arising from claims of copyright infringement…Our court’s precedent requires us to conclude that the monkey’s claim has standing under Article III of the United States Constitution. Nonetheless, we conclude that this monkey — and all animals, since they are not human — lacks statutory standing under the Copyright Act. We but therefore affirm the judgment of the district court.
So, case closed.
Yet as some legal experts have pointed out, the Ninth Circuit didn’t really settle the question of who does own the copyright in the photos at issue if not the monkey. Under the settlement between the parties, Slater, whose camera Naruto borrowed for his self-portrait, is free to license their use on the stipulation that he donate a portion of the earnings to a fund to protect crested macaque habitat — a sort of professional tip of the hat to Naruto — neither the district court nor the Ninth Circuit explicitly determined the copyright to be his. Each ruled on the issue of standing, but neither reached ruling reached the merits of the infringement claim.
Slater set up his equipment in the Indonesian forest where Naruto and his fellow macaques live and left it unattended to see how the monkeys might respond to it. But beyond that, he didn’t have any particular creative input to the photos they took. So on what basis would his copyright in them rest? Is mere ownership of the equipment sufficient? Was providing the opportunity for the monkeys to take the photos by leaving the camera unattended a necessary step? We don’t really know.
Another question not directly addressed by the litigation is what made the monkey selfie so compelling in the first place? Had Naruto managed merely to take a random photo of his foot, or the sky, it’s doubtful anyone would have bothered litigating its authorship.
What made the monkey selfie so compelling, I think, was its seemingly human-like intention. Serendipitously well-framed in the shot, Naruto seems to exhibit human-like self-awareness as he appears to smile into the camera, an effect Slater himself recognized in titling the book in which it first appear Wildlife Personalities.
The whole case, in effect, functioned as a sort of photographic Turing test: Was the photo a creative act of original expression if perceived that way by humans, even if the law does not recognize its proximate creator? If so, how should the law regard it? Who, if anyone, should have an exclusive right to exploit it commercially, and on what would that claim rest?
While the issue may appear abstract and theoretical, those questions are getting less theoretical by the day, as researchers develop ever more machine-based systems capable of producing Turing-sufficient works of expression. Like monkeys, machines would likely lack standing to bring a claim under copyright law. But that doesn’t mean they can’t produce works that humans perceive as expressive.
In a paper published last month, researchers from Microsoft and Kyoto University describe how they trained a neural net to produce original poems, one of which was accepted for publication by the human editors of a literary journal based on a blind, online submission.
Last year, researchers from Google published a paper describing how they taught a neural net to recognize the aesthetic elements of photography. The artificial intelligence system then produced original works of landscape photography that professional photographers had trouble distinguishing from similar works produced by humans (take that, Naruto!).
Computer scientists as Italy’s Politecnico di Milano, last month submitted a paper describing how they used a two-stage artificial intelligence system to create new levels of video game play on the fly that even other A.I. systems could not distinguish from human-created levels.
Copyright law is designed to incentive creativity by giving legal force to the author’s claim on its economic value. Our systems of commercial licensing of creative works rest on exploiting the exclusive rights of authors spelled out in the Copyright Act.
But as the monkey selfie case has shown, works can have considerable economic value even where there is no obvious or legally recognized author.
How that value is to be realized through our existing author-based licensing systems, and how disputes over that value are to be adjudicated within our author-centric copyright law are questions we’re only beginning to grapple with. We’ll be grappling with some of them at the 2018 RightsTech Summit in New York.
Anyone interested in addressing them can contact me at [email protected].
The U.S. Patent and Trademark Office last week held its second public meeting on Developing the Digital Marketplace for Copyrighted Works, where one of the topics of discussion was whether and how works authored by a computer running artificial intelligence software should be regarded for purposes of copyright law (disclosure: I was an informal adviser to the PTO on the program for the event).
The question is not merely academic. Companies like Amper Music in the U.S. and Aiva Technologies in Luxembourg are using A.I systems today to create original production music, while Google engineers have taught A.I. systems to create music and art, and to take photographs that professional photographers have trouble distinguishing from professional landscape shots. Works created at least in part by machines are already entering the stream of commerce.
There is a broad, although not universally shared, consensus among legal experts that under U.S. copyright law, non-human actors, whether machine or monkey, cannot be considered to own a copyright. But that still leaves a host of vexing questions likely to occupy courts for many years. If the machine cannot own the copyright on a machine-made work, can anyone, and if so, whom? The owner of the machine? The author of the software program? The human who pushes “start”? Or, should machine-made works be considered in the public domain?
If a copyright is to be assigned, how much creative input must a human provide, and at what stage of the process, to claim it? If there is no copyright, what if any other legal basis is there for licensing their use?
At the Digital Entertainment World conference in Los Angeles February 5-6, KlarisLaw and KlarisIP managing partner Edward Klaris will explore some of those questions in a special presentation called From Art to Artificial Intelligence.
Klaris previewed some of his thoughts on the topic in a blog post this week for Intellectual Property Watch:
The concept of encouraging the production of creative work by protecting it — incentivizing authors financially — is embedded in our Constitution. The Intellectual Property Clause expressly aims “to promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries.”
In drafting the black-and-white clarity of this clause, our framers could hardly have anticipated the highly gray area of bots making copyrighted works. You don’t have to incentivize a bot; a machine simply does what it was programmed to do without any need for financial motivation. That is why the court declined to award a copyright in a work created by a monkey. Monkeys are not financially incentivized to create works, and even if they were, the monopoly afforded copyright holders was not intended for animals.
In a world where bots may eventually dominate the creative space — manipulating, arranging, color-correcting, filming, and ordering literary, audio and visual content – courts may decide that works created without human input belong in the public domain with no protection. Or, if copyright is granted, bots’ output would be protected for potentially more than 100 years under current copyright law. Which is better? What path best promotes our country’s fundamental interest in “the progress of science and useful arts”? And, should copyright subsist for fewer years under certain circumstances?
With the long-running litigation between People for the Ethical Treatment of Animals (PETA) and photographer David Slater over the now world famous “monkey selfie” still pending before the U.S. Ninth Circuit Court of Appeals, the two sides have agreed to settle the case and have asked the court to dismiss the appeal, according to a joint statement put out by the parties.
But the settlement of the lawsuit isn’t likely to end the debate over how courts and the law should regard works created by non-human authors. In fact, the debate may only be getting started.
For one thing, as The Hollywood Reporter’s legal columnist Eriq Gardner notes, the “settlement” between PETA and Slater doesn’t fully settle the case between them. In their joint motion to the court to dismiss the appeal, the parties also ask that the original lower-court ruling that monkeys do not have standing to enforce claims under the Copyright Act. If granted, that would presumably leave the door open for PETA or someone else to bring a future claim with a different set of facts.
“PETA contends it would be just and proper to not bind Plaintiff Naruto [the monkey] by the judgment of the district court in light of the dispute concerning PETA’s status to file the complaint which resulted in that judgment,” the complaint says.
In a footnote, Slater’s side demurs, “Pursuant to the terms of the parties’ settlement agreement, Defendants also join PETA’s request for vacatur, without joining or taking any position as to the bases for that request.”
Moreover, the statement released by the parties does not say whether they have in fact agreed on who actually owns the copyright in the photographs taken by the Naruto. According to the statement, Slater has agreed to “donate 25% of future gross revenue from the Monkey Selfie photographs to charitable organizations dedicated to protecting and improving the welfare and habitat of Naruto and crested black macaques in Indonesia.”
Does that mean he has agreed to some sort of shared, or split ownership of the copyright between himself and some other entity on behalf of Naruto? Does PETA, or someone else, have the right to an accounting of Slater’s earnings from the photographs? Who will have standing to bring claims against future unlicensed uses? The statement doesn’t say.
Since the case was settled, whatever provisions the parties may have made with respect to those questions will remain private, and do not establish any sort of legal precedent. Yet, while monkeys may not be rushing to set up Instagram accounts to show off their selfies, works created through machine-learning processes and artificial intelligence are becoming more common. And the question of how (or whether) to fit works created by non-human agents into existing licensing structures is only likely to grow more urgent.
A.I. researchers researchers at Google, for instance, have taught an artificial neural network called Creatism to take aesthetically sophisticated photographs that in A/B tests many professional photographers were unable to distinguish from human-made shots.
Music composed by artificial-intelligence agents is increasingly finding its way into the soundtracks of advertisements, videogames, apps, and other contexts where suitable human-authored works are either not available or too costly.
It may not be long before A.I.-composed pop tunes start turning up in Spotify playlists alongside the works of anonymous or pseudonymous artists already popping up there.
Those questions will be the focus of two special presentations at this month’s RightsTech Summit in New York.
In one, Edward Klaris, managing partner in the media and intellectual property law firm KlarisIP will offer a global perspective on artificial intelligence and copyright protections.
A second, roundtable presentation, will focus on the specific case of machine-made music. It will feature Drew Silverstein, founder and CEO of Amper Music, an A.I. composer, performer and producer for creating music for soundtracks, and Arnaud Decker of Luxembourg-based Aiva Technologies, which has developed the first A.I. agent to be registered with a collective rights management organization.
The RightsTech Summit will be held on September 27th at the Museum of Jewish Heritage in lower-Manhattan as part of the 2-day New York Media Festival. Click here for a full list of speakers at this year’s summit. Click here for information on how to register for the RightsTech Summit and the full NY Media Festival.