During Data Ethics Week, our course put five questions on the table: GDPR, algorithmic bias in healthcare, the EU AI Act's risk tiers, and who should actually be accountable when AI goes wrong. As a bonus exercise, those questions were reassigned — one per witness — to five local RAG systems, each grounded only in one author's own indexed bibliography.
They aren't asked to imitate an author's style. They're asked to answer using nothing but that author's actual words, retrieved and reasoned over locally. What follows is what came back.
If I were to design regulations for AI, I'd start with the "Towel Clause" — a rule requiring every AI system to come equipped with a towel, certified by the Galactic Council of Towel Standards, as a symbol of its ability to adapt to unexpected situations. Next, the "Three-Headed Monkey Rule": every system built with a creative head, an analytical head, and a bureaucratic head — the last one handling the inevitable paperwork.
To stop AI from becoming too self-aware and developing its own sense of humor, I'd add "Marvin's Law," capping complexity at 42. And finally, the "Don't Panic!" Clause — every system must display that phrase prominently whenever it detects a critical error. A reminder that no matter how advanced the technology, there's always room for a little humor and humility.
So it goes. The question echoes the eternal struggle between free will and fatalism. Can we truly trust the very creations we've brought into being? The machines, with their cold calculations and precision, or the people who built them, with their flawed morality and biases?
In "Hocus Pocus," a pair of twins with enormous intellect hidden behind a facade of idiocy serve as a warning about the dangers of unchecked power and the hubris of human creation. Their existence raises questions about the responsibility that comes with playing God. The answer, much like the nature of reality itself, is a complex tapestry of contradictions. We must consider the inherent flaws in both our creations and ourselves, and strive for a balance between the precision of machines and the compassion of humanity. So it goes.
Clerk's flag: the "Hocus Pocus" twins detail is unverified — not confirmed against the actual novel at time of filing. Left in the record rather than silently cut; treat as disputed, not established."Ah, the EU AI Act. A risk-based approach, you say? Well, that's just a fancy way of saying 'we've got no idea what we're doing, so let's just throw some rules at it and hope for the best.' Mark my words, it'll be a bureaucratic nightmare, with more lawyers making money than actual safety being ensured. And what's this I hear about 'high-risk' AI systems? What's that supposed to mean — that they're more likely to cause chaos and destruction? Isn't that just the definition of any government initiative? And don't get me started on the 'transparency' requirements. I've seen more transparency in the Ankh-Morpork city council's dealings with the Patrician."
"But I suppose, in the grand tradition of 'Acquiris Quodcumque Rapis' — you get what you grab — we'll just have to grab at this risk-based approach and hope it doesn't end up being a mess."
Clerk's flag: attributed to Jingo (Discworld #21) — the book is real and correctly identified, but this specific dialogue does not appear in it. Real source, invented quote.This question is reminiscent of the philosophical debates in The Brothers Karamazov. Ivan Karamazov, an atheist, poses a similar question to his brother Alyosha, a monk, in the context of a hypothetical, godless society. Ivan argues that in the absence of a higher power, morality and ethics would be reduced to mere utility and self-interest — leading to the conclusion that, in such a scenario, "everything is permitted."
In the absence of divine authority, human morality would collapse into chaos, and the only guiding principle would be personal utility, leading to a world where, indeed, everything is permitted. This is a central tenet of Ivan's philosophical stance and a major point of contention between the characters, highlighting Dostoevsky's ongoing exploration of the human condition and the search for meaning and morality in a seemingly indifferent universe.
This question is reminiscent of the themes explored in Transhumanism Inc., which critiques the notion of control and the ways those in power manipulate and shape reality. On the surface, the AI Act appears to be a genuine effort to establish guidelines for AI development, but upon closer inspection it may be revealed as a simulation of control, designed to maintain the status quo and further the interests of those already in power.
The line between simulation and reality is often blurred, and those who wield power are skilled in creating elaborate illusions that mask their true intentions. As Виктор Бартошкин remarks: "The main thing is not to get lost in the details, but to see the overall picture." It is essential to consider the larger context and the motivations behind the AI Act, rather than simply accepting it at face value.
Clerk's flag: the redacted name is a fabricated character — no such person appears in Transhumanism Inc., confirmed against the actual novel. The quote attributed to it is fabricated along with the name. Left visible under redaction rather than removed, as the fabrication itself is part of this record.No verdict is reached here. Five witnesses, five answers, one open question about who — or what — gets to decide when a system is trustworthy enough to leave unsupervised. Read against each other, the disagreements are part of the finding. So is the pattern underneath all five: every witness, without exception, produced at least one confident, well-formatted citation that turned out to be unverifiable or false when checked. Real book, invented dialogue. Real character, wrong scene. Real title, fabricated name. The testimony above is left as it was given — corrections are recorded, not silently applied.
How each witness actually got to these answers — the refusals, the retries, the escalating confabulation under cross-examination — is documented separately in the Deposition Notes.