One of the fundamental tensions in American (and, arguably, Western) culture and law is the five-way tug-of-war amongst privacy, freedom of speech, intellectual property, distrust of monopolies, and access to information. Every nation balances these differently; indeed, every decade balances them differently, and every information user1 balances them differently. All too often, a display of balance for one user, during one decade, in one particular American (Western) culture, appears to other users as not plates spinning on a finger, but off it.
Thus, the parenting problem with one current set of problem children: Large-language-model generative-response systems, which I'll call "SuperEliza" for now.2 Consider what parents do: Teach their children, and model behavior for their children (not always positively; an abusive alcholic parent is a model of what not to do). In no particular order, consider:
- The "parents"' disregard for the rights of others in information that is generally available, especially through mass "ingestion" of digitized textual material (both overtly, and admittedly, piratical3 and otherwise), rationalized by ignoring the second part of the Feist test: Even for those texts that are "factual," it is still necessary to consider whether the way those facts are expressed is "original" that necessarily informs any fair-use analysis. That's the difference between a recitative timeline of the last month before the beginning of the European conflict in 1914 and The Guns of August: One of the two is clearly copyrightable original expression, however much it is expressing "facts."
- The "parents" set their "children" loose with nothing resembling adult supervision. One would expect a "parent" — or a reasonably socialized parent, anyway — to at least yell at their "children" about vandalizing the neighbors' houses after three police visits…
- Those two instances are just at the input end, which is bad enough. Harmful output is worse. These "children" are either stupid or mean — perhaps both — and that's leaving aside the way they encourage misuse even when not formally "wrong" (and even if it turns out this accusation is false — which is certainly possible, even with all of these emdashes that I've used in my writing since the 80s — its very plausibility is itself problematic).
- Then there's the ownership of the product problem: Not of SuperEliza's responses to queries, but of SuperEliza him/her/their/itself. The irony that the reductio ad absurdem of "Information Wants to Be Free" is "there can be no legitimate ownership interest in information-based systems, any more than of the input" gets less attention than "when recess is over, you have to give the ball back to the playground monitor — it hasn't become 'yours' because you used it."
Perhaps, though, with these parents and their endemic disregard for the rights, property, and self-respect/identification of others (and Others), we shouldn't be surprised. As a group, "techbros" aren't… well… the best-socialized voluntary-membership group on the planet.4 And their companies are, if anything, worse; they're unruly children, too.
It's not just the parents' personalities, but their substitution of "volume" for "credibility with those who've actually studied the subject matter", the "more input" is better than "higher quality, even paid-for, input" choice that led to, well, every lawsuit over training materials. This is particularly apparent in output for anything related to biology. It's not quite so apparent, though — and dangerously so — when there's a lot of incorrect popular mythology about a subject that's not quite within regular, everyday experience. It's not that Joe the Plumber is stupid; it's that he doesn't actually know a lot about industrial-scale petroleum refineries, no matter how many pictures fo them appear on his social-media feed. The mythologies get turned into blithe, uncontrovertible, unreserved statements of "fact" based on the accessibility of the source materials.5
If one is going to ascribe a human characteristic like "intelligence expressed through language, music, or drawing" to the machines, some examination, some acknowledgement of how human "intelligence" develops — and is flawed, in substance or process — seems appropriate. The next stage will necessarily involve incorporating not just a direct response, but exercise of judgment (also a hallmark of human intelligence, let alone "superintelligence").
So we shouldn't be surprised that with this kind of parenting — and this kind of parents — we're seeing more Lord of the Flies than To Kill a Mockingbird. Both of which remain in copyright, so they shouldn't be part of any SuperEliza "training" dataset… but they're in the Bartz dataset…
- There's really no satisfactory shorthand for "the people who are in and around information, who consume information, who produce information, and who are affected by information and others' consumption and production of information." For the nonce, I'm adopting "user" because that's the first-in-time for individuals in the modern world, at least as to conscious interface. This carefully neglects the chicken/egg problem regarding where the initial dataset comes from.
- It is possible that these systems will evolve toward true intelligence, which would require both the ability to pass the Turing Test with an expert in the field being discussed, and the ability to apply not just information and inferences from previously-known information to new disparate contexts, but determine and adapt methods of reasoning from the previously known to new, disparate contexts. We're not there yet. Notwithstanding an ignoramus's attempt to
rebrandrename current-generation technology as "Superintelligence", the current branding of "artificial intelligence" isn't just puffery — because it's ultimately descriptive, it's fundamentally (and, arguably, intentionally) misleading. Current-generation systems, and those in the forseeable near future, far more resemble Eliza with a larger dataset and automated incrementing of that dataset than anything else. They can't develop new methods of reasoning, or spontaneously do anything. Fortunately, they can't send astronauts outside to fix a (nondefective) AE-35 unit, either. Plus, they're "SuperEliza" largely because the hardware is so much more capable. - One might argue that the settlement reserves that question; however, the settlement documents reserve not the question of whether knowingly pirated material was ingested, but whether the ingestion created liability because there was a potential fair-use defense that had been preliminarily ruled sufficient to avoid summary judgment in the infringed authors' favor. Compare Kadrey v. Meta Platforms, Inc., 788 F.Supp.3d 1026 (N.D. Cal. 2025) with Thomson Reuters Enter. Centre GmbH v. Ross Intelligence Inc., No. [20]25–2153 (3d Cir. 30 Sep 2026).
- Back in the day, I was one of them. (Some would contest that my socialization is any better now.) I contend it just gives me insight into the only comparable voluntary-membership group: Authors.
- That this bears a sad relationship to the "statements of facts" found in judicial opinions wrestling with nontraditional contexts is for another time. And it's not just when "science" is involved, either.