BoM/TWS Message Forum

I know it feels like beating a dead horse, but the spell protecting the Libra still suggests an interesting combination.

The way I see it, the spell preventing instant translation of the Libra is meant for the reader to understand the content the hard way - translating each word very carefully. The expectation is that the owner of the Libra has solid grasp of Latin, enough to understand the instructions without relying on magical help. This correlates to how you can't translate the book immediately and thus must go bit by bit.

While AI would definitely provide a reasoning for the translation (akin to how someone like Caleb would explain why X word means this and not that), it'd find multiple hurdles. For example, machine learning would still be unable to translate the contents of the book at once - you'd need to drip-feed the content. Attempting to do so would cause an untold amount of glitches, for one.

Even if drip-fed, there are points where AI will be stymied. Take for example the 8th spell, which has two possible translations: "use me to solve me" and "my use is my solution". Judging by how it behaves, something like ChatGPT would suggest both if prompted at different times, and if prompted what would that mean, the results would be esoteric - at that point, the idea should be that AI can only take you that far.

However, one other thing about machine learning is that, the more it's used, the more it adapts. Not just that - imagine if, say, the machine-learning algorithm suddenly gets fed the contents of Francken's book on sigils. The machine wouldn't be able to interpret the information at first, but at some point (probably with some added contents from other books, especially if those have real magical information), a prompt for "write me a spell" would eventually result in a real, functional spell. If that machine-learning algorithm is then trained on the Libra, at some point it may replicate one of the spells, or perhaps a derivate of the spell (like the extraction of the sigil that knocks out people) with full functionality.

There's also a possibility that the obtuseness of magical language renders any attempt for machine learning algorithms to decode magic downright impossible, but the suggestion is there. The Libra, or at least some of its contents, would be somehow fed to a different form of intelligence, one that may struggle to understand it at first but then it'll be capable of memorizing it. It's an entirely different beast, one very different from cryptids and ancient conspirations. (In fact, what if a modified mind band were to be set to a server that hosts an "AI girlfriend" algorithm and a ChatGPT-esque algorithm? If placed in a free golem, would it become the bridge that ultimately leads to the formation of true AI?

That said, to address the point you've been making - would it be trustworthy? Would the characters trust the translation, or will the deficiencies inherent to machine learning will be enough to distrust the whole thing? The answer to that has to consider that, the more the characters use that AI tool (for translation or for context), the tool will learn and adapt to it. As for its existence, I still suggest that it should be a TWS thing - that would make BoM less of a period piece and more of "alternate history", as the latter is much more robust than the former and you could suggest that AI development got stymied for reasons, therefore it releases later (arguably under Fane, which would be the worst of the possibilities since you could then create a machine-learning algorithm to sniff out the Stellae and their associates).