Tag: Featured

AI “Art” is Boring

Adam was bored alone; then Adam and Eve were bored together; then Adam and Eve and Cain and Abel were bored en famille; then the population of the world increased, and the peoples were bored en masse. To divert themselves they conceived the idea of constructing a tower high enough to reach the heavens. This idea is itself as boring as the tower was high, and constitutes a terrible proof of how boredom gained the upper hand. – Soren Kierkegaard (1843) –

I had not thought about Kierkegaard writing on the subject of boredom in years. The essay from which the above quote is extracted was a favorite in college for its biting humor, but something about Rogers Brubaker’s excellent article about democratizing culture sent me in search of my 38-year-old (ouch) copy of The Kierkegaard Anthology, and I think it was this paragraph of Brubaker’s which triggered the thought:

But the question is not just how many people engage in cultural production — it’s how people engage. The AI music company Amper promises to help customers “create your own original music in seconds.” The creativity involved is rather attenuated, amounting to editing and tweaking the music generated by the AI, but that didn’t stop Amper co-founder Drew Silverstein from evangelizing in a TED talk about how AI can “democratize music” by enabling “anyone to express their creativity through music.” 

That promise to “create your own original music in seconds” was the portkey back to Kierkegaard. “In the case of children, the ruinous character of boredom is universally acknowledged,” he writes, and, indeed, I maintain that boredom is the inevitable outcome of AI toys promising to make music, visual art, poetry, etc. We have all experienced as children and witnessed as adults that transition between playing with a new toy and rapid disenchantment because the toy fails to engage the imagination. I am not the only Gen-X parent, for instance, to notice that when LEGO began selling kits to build branded objects like Star Wars spaceships, my own children would usually complete the assembly once and then be done with the toy forever. By contrast, my contemporaries and I spent hours with sets composed of bricks and no predetermined design.

Kierkegaard proposes that the plebian bores others and amuses himself while the aristocrat amuses others and bores himself—a dialectic perhaps well suited to describe the inevitable use of AI machines to “make one’s own music or art.” At the current state of the technology, the input of the human user is barely creative—little more than dropping a coin in a jukebox—and thus, all users similarly situated are plebian bores for the time being. The works resulting from their prompts may amuse them (for a while), but they will mostly bore others who will only be interested in “making their own music” with the same toys. Before long, a million individual users of the music generating AI will achieve a collective homeostatic boredom—a two-dimensional Babel leading nowhere.

Perhaps one of these accidental works will reach escape velocity, break through the gravitational force of mass boredom and “go viral” for a fleeting period. Some AI-generated ditty might be next year’s “Baby Shark” or even share the apotheotic luminance of a “Gagnam Style.” Someone will choreograph a short dance to accompany the tune, and TikTokers will fall in line to perform their versions, and Big Tech will look down and see that it is good, and their disciples will proclaim, “Behold the new culture! The human songwriter is an anachronism.” And it will all be as boring as it is ephemeral.

It is possible, of course, that generative AIs will become sophisticated enough to be collaborative tools wielded by the human artists—that the human still selects and arranges the creative elements to achieve her vision while the AI “helps” in some way. If and when we get there, we shall see. But in the meantime, it is clear that AIs do not need to be more sophisticated to replace some creative human work right now. My good friend Marco North writes on Facebook to me, “A full roster of AI voice talent costs less than $100 a month, works 24/7 and [will] do endless revisions….Voice work is perfect gig work for actors, say goodbye to lots of that.”

A gifted polymath in film, photography, music, poetry, and prose—Marco writes a weekly blog called Impressions of an Expat. Initially written from Moscow, he now writes from Tblisi, and in his latest post, he describes a happenstance encounter with the statue of Georgian poet Vazha-Pshavela (Luka Razikashvili) and his feelings about AI “art.” He asks:

Who will be the subject of the next statue? An algorithm? Will there be streets named after TikTok? Will we name a playground after a Spotify playlist curator? These are the people that tell our stories now. Midjourney highway will take you there. Take a left at ChatGPT square, you can’t miss it.

Yes. That is a vision of a possible future. Of course, if the tech giants can make the world just boring enough, then certain humans will do what certain humans do. They will disassemble the unengaging toy and turn it into something else—something called art. And then, the world will start to be interesting again.

What I’ll Be Watching in 2023

T’is the week for year-in-review and/or looking-ahead articles. In that spirit, I recommend posts by Devlin Hartline, Hugh Stephens, and Aaron Moss. And here’s my list with commentary for your consideration:

AWF v. Goldsmith

Everyone in copyright world will be waiting, like Ralphie expecting his decoder ring, for the decision in this case. The highly anticipated question is whether the Supreme Court will provide clear guidance on the meaning of “transformativeness” in the factor one analysis of the fair use test. By invoking this highly subjective concept, follow-on creators have, at times, pushed lower court decisions toward problematic findings—first by finding “transformativeness” in secondary works that encroach on the derivative works right and/or classic instances where licensing is required; and second, compounding these errors by giving undue weight to factor one in the overall analysis.

AWF has argued that any “new meaning or message,” which may be subjectively interpreted by observing a follow-on work meets the definition of “transformative” and is, therefore, outcome determinative for finding fair use. Although, I have opined that this case poses certain difficulties in my view (i.e., that Warhol may have defenses under other principles), I agree that AWF’s argument presented here should be rejected and believe the Court should state that factor one must turn on whether the follow-on work contains at least some modicum of commentary on the original work. Absent such commentary, factor one should favor the copyright owner plaintiff. We shall see what the Court says in the coming weeks.

Hachette v. Internet Archive

Inasmuch as this is a very big case that could go all the way to the Supreme Court, it is almost impossible to fathom how Internet Archive is not destined to be rebuffed on the merits at every turn. What began as a lawsuit in response to IA’s unlicensed distribution of over one-million titles (using the fog of early COVID shutdowns as a rationale) is now a detailed complaint in which the facts imply more than just founder Brewster Kahle’s anti-copyright crusade.

In 2017, I asked whether the good aspects of IA require all the anti-copyright rhetoric in order to exist, and that was presuming Kahle’s well-known opposition to copyright was purely ideological. But some of the details in the publishers’ complaint imply financial interests that belie any pretense that IA is a principled, though misguided, Robin Hood. Expect to see the organization continue to allege that it is “just a library doing what libraries do,” but if this were true, the publishers’ suit would have been dismissed at the summary judgment stage. It should be clear by now that just because you say something on social media, that doesn’t make it true—least of all in a court of law.

The Copyright Claims Board (CCB)

Launched this past June, 2023 may be the year we really start to test the efficacy of the copyright small-claim alternative. For starters, the big question is how many respondents will opt-out of the tribunal. In order for the small-claim option to be constitutional, a defendant (respondent) must voluntarily agree to the proceeding, which led some to reasonably wonder whether the CCB will work at all if every defendant can simply opt out. But that question partly depends on how many plaintiffs are willing to file federal lawsuits, if the respondents are unwilling to resolve the matter at the CCB.

Further, to really understand how things are going at the CCB, we need a volume of cases and more time to allow the process to unfold. The plaintiff has 90 days once her case is active to show proof of service on the respondent, and the respondent has 60 days to opt out of the proceeding. Thus, with fewer than 300 cases filed between June and December this year, we simply do not have a lot of data yet. That said, Rachel Kim at Copyright Alliance posted a blog summarizing what we do know so far, and it’s worth a read.

Artificial Intelligence

I will not attempt to predict where this story goes in 2023, other than to expect that AI will continue to make headlines in the art world and beyond. As stated many times, I personally think AI generated “art” is a useless waste of computing power, but even if every artist and art consumer in the world agrees with that view, it seems unlikely that market failure of the companies behind generative AIs will predate one of these entities getting sued for copyright infringement. Perhaps not this coming year, but before long, expect to see litigation over the question of whether inputting large volumes of protected creative works into these databases amounts to mass copyright infringement or is exempted under the doctrine of fair use. And in anticipation of this battle, both sides of the argument may be scrutinizing the opinion(s) in AWF v. Goldsmith.

Gonzalez v. Google

Not a copyright case, but on the subject of platform accountability, the Supreme Court will finally have something to say about Section 230 of the Communications Decency Act. The decision likely won’t come until 2024, but we will soon see briefs filed on behalf of Google, and oral arguments will be heard in 2023. I recently posted about this case here and here, but suffice to say, it is hard to imagine that the majority will not generally agree that the statute neither states—nor ever intended to state—that online platforms are entitled to the kind of unconditional, broad shield against civil liability the lower courts have granted them for nearly 20 years.

Although 230 is not copyright law, it shares a kinship with the contemporaneous DMCA. Both laws were predicated on immunizing platforms from liability for material posted by users, and although neither law grants these immunities unconditionally, many online service providers—especially the big ones—have wielded these liability shields beyond the limits of reason or anything Congress intended in the late 1990s. Thus, if the Court reigns in the free-for-all applied to date under Section 230, it is conceivable that the opinion in Gonzalez will inform congressional review of the DMCA, which began in 2020.

That’s what I got for this December 30, 2022. See you in the new year!

Photo by: MediaFuzeBox

Cyber Civil Rights Initiative Files Common Sense Brief in Major Section 230 Case

In my recent post about Gonzalez v. Google—the Section 230 case granted cert by the Supreme Court—I expressed the view that the word “recommendation” is too charming to describe the interaction between social media algorithms and many users’ experiences. Systems capable of reinforcing suicidal ideations in a teenager or stoking violent instincts in a potential terrorist cannot sensibly be described as “recommending” the kind of content associated with these and other dangerous outcomes. And although petitioner Gonzalez specifically asks the Court to decide whether “algorithmic recommendation” is shielded from liability under Section 230 of the Communications Decency Act, the amicus brief filed by the Cyber Civil Rights Initiative (CCRI) and Legal Scholars asks the Court for a more nuanced reading of the question. From the brief…

Amici emphasize that this case cannot be correctly decided by focusing on “traditional editorial functions” or by trying to craft a general rule about whether “targeted algorithms” fall within Section 230’s immunity provision…. To categorically deny immunity to an ICSP for using targeted algorithms would directly contradict Section 230(c)(2) and finds no support in Section 230(c)(1). Such an interpretation would also have a devastating impact on the victims of online abuse by dissuading Good Samaritan ICSPs from using targeted algorithms to remove, restrict, or otherwise reduce the accessibility of harmful material, including nonconsensual pornography.

CCRI, which works to address and remedy various forms of harassment and civil rights abuses committed via interactive computer service providers (ICSPs), asks the Court to restore the textually coherent and common-sense meaning of Section 230, which was written to encourage service providers to mitigate harmful material—not to unconditionally immunize them from liability for hosting it. For almost twenty years, lower courts have consistently misinterpreted the purpose of 230 to provide automatic immunity just so long as the material at issue is posted by someone other than the platform owner/managers.

This chronic misreading of Section 230 results in two significant problems: 1) dismissal at the summary judgment stage of any claim in which an ICSP may be liable; and 2) failures to provide injunctive relief where the ICSP is not liable but may be ordered to remove material which the court agrees is causing harm to a complainant. As things stand, a site that intentionally trades in harmful material is immunized, and so is a site that unintentionally hosts harmful material but elects not to remove the material for its own reasons. The rationales vary as to why “neutral” platform operators often refuse to remove material alleged, or even proven, to be harmful, but for too long, the industry has echoed the absurd premise that removing anything from a social platform is incompatible with “a free and open internet.”

Section 230 Is (Was) Not Novel Legislative Territory

The CCRI brief is so firmly grounded in the legislative history of Section 230 that it is difficult to fathom how any court—let alone many courts—strayed so far, and for so long, from a plain-text reading of the statute. In describing the common-law (i.e., not groundbreaking) underpinnings of Section 230, for instance, CCRI cites the distinction between a “publisher” and a “distributor” of defamatory material thus:

… “[d]efamation at common law distinguished between publisher and distributor liability.” While a publisher was strictly liable for carrying defamatory matter, a distributor who only “delivers or transmits defamatory matter published by a third person is subject to liability if, but only if, he knows or has reason to know of its defamatory character.” [Emphasis added.]

This is common sense well founded in law. If an individual or a business has knowledge that he/it is facilitating harm caused by a separate, directly liable party, that facilitation may rise to a secondary civil or criminal liability. The newsstand operator is not liable for inadvertently selling adult magazines containing underage models, but if he knows about it, he is probably—and deservedly—in big trouble.

This basic principle of secondary liability applies everywhere except for internet platforms—and only because the courts have so thoroughly misconstrued Section 230 by conflating two sub-sections of the statute, which are meant to be read independently. As the CCRI brief explains, 230(c)(1) states that merely providing access to third-party content (e.g., YouTube hosting a video uploaded by a user) does not make the ICSP a “publisher” or “speaker.” Then, 230(c)(2) states that voluntarily making a good-faith effort to remove objectionable material does not make the ICSP generally liable as a “publisher” of everything it hosts.

“Cases reading Section 230 to have a broader preemptive effect than provided for in (c)(1) and (c)(2) have departed from the statutory text,” states the CCRI brief. It emphasizes the fact that “distributor liability” is envisioned by Section 230(c)(1) where the ICSP has knowledge of the harmful material, and it argues that the function of Section 230(c)(2) is legislatively “parallel” to state Good Samaritan laws written to immunize ordinary citizens against unreasonable liability when we make good-faith efforts to help someone in need of assistance. Prior to these laws, an individual intending to render aid to a stranger could be held liable for inadvertently causing harm, but as the CCRI brief states:

… like state Good Samaritan statutes, Section 230(c)(2) includes important limits to the immunity it provides. First, it does not apply when an ICSP is already under an existing duty to act—i.e., where its action to restrict access to objectionable third-party content is not “voluntary.” Nor does it immunize ICSPs that do nothing to address harm or that contribute to or profit from harm.

Again, this is just common sense grounded in common law that applies everywhere except the internet. If one does not initiate illegal activity but seeks to benefit from that activity, one may be liable for the harm caused. It is inconceivable that Congress ever intended to exempt the multi-billion-dollar internet industry from this longstanding principle. And that’s because it intended no such thing.

It will be interesting to see what amici who file on behalf of Google will argue in this case. Other than the usual panegyrics to the internet, I am curious to see whether, for instance, the EFF will have anything coherent to say in defense of two decades’ worth of textual misreading. Typically, defenders of the status quo reading of Section 230 write about threats to “the internet” as if a lack of immunity automatically results in a finding of liability and damages. But on the contrary, a proper reading of the law simply means that an ICSP cannot so easily dismiss every claim and that the injured party is allowed her day in court to prove whether a platform had, or has, a duty to act. Litigating against tech giants is hardly a fair fight in the first place, and ICSPs neither need nor deserve an unconditional immunity that exists nowhere else in the justice system.