Thursday, August 11, 2022

IPSC Breakout Session 3, Language and Authorship/ownership

What Is the Relationship Between Language and Thought?: Linguistic Relativity and Its Implications for Section 230 and Copyright

Christopher S. Yoo

Cognitive linguistics says language may not be independent of thought: language shapes how we think/what we can think about. Very much a debate in the literature: Sapir-Whorf v. debunkers. “Thinking like a lawyer”: you become adept at thinking in certain ways and language is a part of that. Structure of a language: subject verb object; only eight options for order—only two use one of the options, and Klingon deliberately chose that one to be more exotic.

Contrast: Chomsky’s universal grammar. Children don’t get enough examples to do what they do with language. Stimulus alone isn’t enough.

New empirical studies: effects of language on directions, horizontal/vertical/size metaphors for time, lack of numbers in Piraha and inability to do math, feral children.

What does this have to do with law? 230: Roommates.com was the first exception for 230 because Roommates established categories for people to use. By creating a mandatory category structure, that was consistent w/ the idea that structuring a language has effects on the people who respond to it—a determinative component to doing that. Has implications for autocomplete and prompts, and other forms of AI.

Constructed languages: attempts to use copyright to protect them. Tolkien: C&D, but no lawsuits. Klingon: Axanar. Tasmanian Aboriginal Centre and palawa kani: trying to put fragments together to preserve it. Copyright claims attempted to preserve/protect/avoid splits.

Can’t © structure as idea, nor words as too small, but what about the area in between? A new alphabet, a set of words, a grammatical structure? Especially when done for dramatic purposes: should it be copyrightable? His conclusion: no protection for syntax but yes for the language as a whole.

Esperanto fragmented—people keep fiddling with it so it can’t become standardized. See the same complaint in open source software—everyone has to add a bell or whistle. Loglan, created to test Sapir-Whorf, also forked and had a copyright dispute. Palawa kani: asked Wikipedia to take the website about it down, don’t want anyone to use the language except natives until it’s fully reconstructed.

Software: machines aren’t limited by biology, but programmers use the structure of source code. Thinks there’s a difference between protecting SSO syntax and terms; a particular naming architecture might be protectable.

Can we rely on language creator to keep it optimally permeable? Unix died from forking into 20 different versions; no one had the authority to assert control of it. Benefits of widespread use need to be balanced with potential to squelch competition.

Josh Sarnoff: say more about merger. Once it becomes a language, you’ve lost ability to control it.

A: if you create a language that’s entirely yours, you never lose control. It didn’t develop organically. A business code communicating within a community designed to be closed. If public function, no, but if private, yes. Merger is off the table. [So if you bring kids up speaking this language, you will definitely have lost control.]

RT: “Unix died from forking into 20 different versions; no one had the authority to assert control of it.” Replace Unix with English. Why didn’t English die from being control-free?

A: Software is less forgiving than spoken language.

RT: ok, then this is two papers, one about software and one about spoken language.

A: He finds the comparison compelling.

Samuelson: CJEU decided that SAS language was not something that they had a © interest in; they developed the language so their userbase could make programs to run on the platform. WPL took programs written by SAS users and executed them on an alternative platform. That’s a good example b/c it’s a completely made-up language; she thinks that GvO indicates that users’ investment in constructing programs within the language matters. She thinks there’s more flex in merger than Yoo does.

A: author of Esperanto consigned it to the public domain. Authors have options: they can surrender to the public; they can assert complete control; or something in between. SAS: does inviting people in some of the way invite them in all the way? His conclusion: people are better off with more options/ability to say this far and no further.

Felix Wu: Not sure sees 230. If it were true that Roommates.com were based on structuring the language, that doesn’t really distinguish the dating site cases. Don’t think you need it. Feels very different from constructed languages.

A: if you’re just a passthrough and not speaking yourself then you get immunity but if you structure the discourse then you don’t.

Chander: more on Yoo’s side.

RT: very much not: you can absolutely structure discourse and stay immune.

Yoo: You’re protected by the good-faith provision, but it’s complicated.

[Sites are immune from liability for content supplied by users. Ravelry will remove your non-knitting/crochet-related posts. It is covered by 230, including the immunity for user-supplied content, not just the good faith (c)(2) provision, and it should be. The paper may well be more detailed, but there is a difference between an affordance and a requirement. The other dating sites offered affordances; on the facts alleged in Roommates, the site required selection within regulated categories, thus forcing people to select descriptors the law didn’t allow them to select, which was attributable to the site because the site decided what was mandatory. Affordances can also make things more likely/thinkable but as long as fields are optional and are only illegal if the user misuses them, the current jurisprudence means 230 protects them.]

The Right to Speak a Brand: Rethinking the Interaction Between Trademarks and Speech in the Age of Expressive Branding

Alvaro Fernandez-Mora

Tweak of original mark to convey parody/comment: adictas three cocaine stripes v. adidas three stripes; enjoy capitalism in Coca-Cola font; Lost-it in Post-it font. TM owners make dilution and confusion claims.

There’s a lot of literature and case law; focus of courts and scholars is on the extent to which freedom of expression ought to operate. Compare: TM registration, FACK JU GÖHTE—courts finally said that freedom of expression operates in registration as well. Seeing a convergence in application and infringement: freedom of expression as sword, not just shield. Compare: public law measures encroaching on TM use, such as plain packaging for cigarettes (probably impossible in US) which is now the norm in many jurisdictions—prevents use of registered TM. Rightholder tries to challenge validity of rules on freedom of expression basis. South Africa’s prohibition on depicting healthy, happy babies on infant formula containers.

Lessons from this taxonomy: it’s a two-way street. It’s not about balancing incommensurable property v. speech, but speech v. speech which is more feasible. TM can both limit and validate TM rights.

In response to question from Ramsey: Would lead to some situations in which unauthorized use would damage expression function of TM, but very very rarely.

Rothman: we have speakers using someone else’s TM, and speakers using their own, which seem like they’re more similar than you’re presenting. What does validation of TM rights mean in this situation? You need a functioning TM to have a valid TM; can you reject it even if it functions as a mark? The First Amendment comes in only if it functions as a mark (so far).

Labeling and regulation—limiting uses of the mark—are two different things, but they don’t limit TM rights [against other people]—they’re just part of the regulatory machine. In some instances they limit what people can do; in other cases it’s compelled speech. [Here I guess I’m actually with the brand theorists: restrictions on using trade dress/requirements to only use block text seem very likely to impact the ability to build goodwill/secondary meaning. But it’s a question deserving an empirical investigation]

A: maybe grounding is a better term in US discourse than validating. The argument is the huge warning label affects the rightsholder’s ability to convey its own branding message in the tiny remaining amount of the package, but it’s true they aren’t identical.

Jake Linford: how different is the concept “use” from the concept “speech” in your conception? We speak of use as a mark/use of a mark. But is that the same as a speech act in terms of what the owner’s rights are, a third party’s, the government’s?

A: fuzzy distinction [I think that use and speech are identical here—if not used to communicate, then not speech and not a use as a mark at the same time].

Creation, Room 90

The Neglected Role of the Family in Copyright

Tal Itkin

Literature emphasizes social environment’s effects on authors, but rarely discusses immediate family’s effects on authors. Psychological research supports impact of family. Intimacy defines the influence of the relationship.

Many jurisdictions accept special relationship between author and work and give the author control greater than a spouse would have over other property acquired during the marriage. In most cases, the spouse doesn’t contribute copyrightable creativity. Argument: spouse’s contribution should not be seen through copyright law but through family law. Ownership should require a lower burden of proof: comes from the relationship between the spouses and not from the connection to the creation.

Rosenblatt: compare treatment of inheritance where US © law treats inheritance very differently than many other rights/properties.

A: her concern is divorce law, though comparison is worthy.

Gerhardt: compare right of publicity, moral rights?

A: family supplies raw material for work of art. The moral rights can be seen in some cases as joint, but that will be controversial

The Revolution Has Arrived: AI Authorship and Copyright Law

Ryan Abbott and Elizabeth Shubov

UK has clearly allowed AI authorship for a while, but AI authorship has not been any good for so long there was no real risk of infringement/point in litigating. But the US has not. Recent patent case: patents must come from humans. Seeking cert on that. In ©, the cases are about spiritual mediums and the monkey selfie.

Neither the Act nor the Constitution defines authors; corporations can be authors; so who’s to say a program can’t be an author, especially given that the standard for originality is so low. WFH: an AI can’t sign a written contract or be an employee, so it doesn’t really fit. The machine has no rights, so there’s no worry about exploiting it.

Arguments against: machines can’t create, but that seems technically unsound. Only the thing the person doing is the sort of thing that ought to be protected. But we don’t really know what people are doing; people in philosophy of mind and computing think that what the computer and the mind are doing is very similar—generative algorithms. Consequentialist benefits of just protecting human creativity: relational—it’s good for people to create. But the history of © law is about encouraging creation and dissemination of creative works, and if we happen to help some authors along the way that’s good. AI doesn’t need incentives to create, but that’s a straw man b/c you will encourage people to make and use generative AIs that make creative works, and these can be expensive and also you need incentives to disseminate. Denying © protection sets up a 2-tiered system where Marvel will be able to generate a movie that people will love, but won’t be able to do that and keep © even if the human-involved version is less efficient and less good.

Fagundes: Worth pointing out that all creativity is mediated by some technology. The camera does work for the “author.” But who is the person?

A: the AI should be the author.

Fagundes: WFH seems analogous. But it’s not a given that a visual artifact has to have a ©--Meshwerks, Corel.

A: people like works that they value as creative works. Now AI is finally likable and tech people are making lots of AI art.

Fagundes: if likable and cool, where are the incentive needs?

A: To get more?

Paul Goldstein: there are differences, like the inevitability of death for a human. These programs may be costly, but that points to patents or even sui generis scheme to promote investment in AI, but not giving protection to products created at the cost of electricity.

A: you have to pay someone to make this if a person does it, but if a machine does it, you don’t, but you still need incentives to distribute. Could do it sui generis/

RT: notable that you just said that people are “making” art with AI. But your avowed position is that the AI is the author. Freudian slip?

I’m kind of with the artists: Marvel should have to pay the occasional human to get a ©.

Independent creation: when you can’t trace, which is true for humans, then it may be justifiied to give ©, on a sort of tie-goes-to-the-runner principle, but we can trace exactly what the inputs were for AI even if we don’t know why it ended up generating this particular image.

A: But people train on datasets as artists; all creation is remixed from what’s come before. A machine that has trained on 650 million images creates something that looks original.

No comments:

Post a Comment