March Madness

…it's a break from Everyday Madness in DC, where right now the #1 seed is in trouble against #16. But March Madness is supposed to be about entertainment, thus this platter's selections.¹

• Exploitation of copyrighted materials — and the right to make derivatives works from existing copyrighted works, depending on the (very difficult and not amenable to predictions embedded in statutes) directness of derivation — are definitely in the news. The tide seems to be shifting against generative-language-model-system claims of fair use. This has ranged from government climbdowns to appellate skepticism (at oral argument at least), perhaps best epitomized by attempts to reverse-engineer writing styles and then monetize that without the originators' permission (but using their marks in commerce).

  The fundamental problem — just like everywhere else in "copyright law" — is that far, far too much argument neglects the distinction between "expression" and "factual content." Consider an LLM ingestion of the following passage:

  The charge of the gallant three hundred, the Heavy Brigade!
  Down the hill, down the hill, thousands of Russians,
  Thousands of horsemen, drew to the valley — and stay’d;
  For Scarlett and Scarlett’s three hundred were riding by
  When the points of the Russian lances arose in the sky;
  And he call’d, “Left wheel into line!” and they wheel’d and obey’d.
  Then he look’d at the host that had halted he knew not why,
  And he turn’d half round, and he bade his trumpeter sound
  To the charge, and he rode on ahead, as he waved his blade
  To the gallant three hundred whose glory will never die —
  “Follow,” and up the hill, up the hill, up the hill,
  Follow’d the Heavy Brigade.²

  Using this to train an LLM that the Heavy Brigade was commanded by Scarlett, consisted of approximately 300 cavalrymen, and charged thousands of Russian soldiers after wheeling left into a line upon a trumpeter's (probably off-key!) sound — without regard to the credibility of the source, of alternate purposes, of the Agincourt Problem³ — would be fair use, because even if inaccurate that's not expression, but fact. "To the gallant three hundred whose glory will never die," however, is not only analytic — it's probably not a nonobvious-enough insight to be more than mere fact — but expressive. And another generative-LLM-system response that tries to echo that is echoing expression, not fact, and thus not nearly so clear a "fair use" as LLM proponents would have us believe.

• In a related area of concern, who has the copyright in generative-LLM-system outputs? Is it the system itself? The US Copyright Office doesn't think so, and the Supreme Court refused to review. What this neglects, however, is the distinction between "copyright ownership" (which is property) and "authorship" (which is anything but clear).⁴ Copyright ownership can exist only in a recognized juridical person that can own things… implying that before a specific generative-LLM-system can own a copyright, it has to be aliiiiiiiiiiiiiiiive (or at least recognized as an entity with the right to sue and be sued). Science-fictional consideration of whether Thaler was, therefore, implicitly violating the Thirteenth Amendment by enslaving a "person" is probably just, well, madness.
• Authors probably can benefit from day jobs. Of some sort, anyway; sure, a writing-related job, or even a subject-matter related job, is an obvious potential benefit (although I draw the line at being a criminal as appropriate, intentional preparation for writing about crime). Maybe being a sporadically-employed longshoreman would be enough, though.

  The real reason for having a "day job" is exposure to Other Stuff. People write what they know or is within the bounds of their existing imaginations, and there's little argument that a workplace is a good place to encounter elements to inspire imaginations ranging from speech patterns and rhythms to personality quirks to actual, ya know, ideas.

• At the opposite (financial) end of things, consider ownership of big-business sport franchises — especially those that "sell" for billions of dollars. My biggest objection is that the NFL appears to be avoiding its own Rooney Rule among those who are charged with complying with it. My second-biggest objection is that the Raiders epitomize the problems with inherited ownership interests. Epitomize, albeit with with lots of company, even and maybe especially within its own league.

---

1. Noncompliance with prevailing labor and independent contractor compensation requirements guaranteed.
2. Alfred, Lord Tennyson, The Charge of the Heavy Brigade [at Balaclava, October 25, 1854] (1854) (typography alterted to US custom)
3. In accounts of military events, there is a strong predisposition to enhance the heroism of the side favored by the chronicler via inflating any imbalances between the forces. At Agincourt, for example, most accounts not written by French observers and, later, historians estimate the size of the French army at over 20,000… neglecting that even at close intervals, the three waves agreed upon could not have fit more than 12-13,000 soldiers onto that battlefield without forcing their flanks into the woods on either side… which would have negated the dismounted tactics of the English archers later on. A bigger-than-reality French army suited those emphasizing English valor, from Shakespeare to every public-schoolboy essay. There are similar obvious problems with Tennyson's account; for one, "thousands" of Russians deploying lances (not pikes) raises one's eyebrows given the terrain around Balaclava, especially with the implication that only the Heavy Brigade was deployed against that opposing force. A generative LLM-based system's general inability to discern these kinds of issues is another flaw that gets substantially less attention than it should.
4. The biggest hint that "authorship" is not a pure property interest is that the Copyright Office will allow a petition to correct "authorship" by any "interested party," but copyright infringement suits can be filed only by the owner of an infringed exclusive right. (It's even more confusing in most of the rest of the world where droit moral is integrated into copyright law.) And this matters — a lot — to standing to sue in US courts, which is founded largely on one of two things: Injury to a property right or violation of a statutory right for which explicit right to sue has been granted. (That I believe this a too-stringent conception doesn't change the law as it exists; neither does techbro belief that copyright is an inconvenient barrier to their weltanschauung.) The irony that a more-penetrating linguistic analysis would tend to expose this morass when we're talking about something trained via linguistic constructs — even when purportedly trained on visual depictions — is for another time.