

I’m not sure where you’re getting the idea that language models are effective lie detectors, it’s very widely known that LLMs have no concept of truth and hallucinate constantly.
And that’s before we even get into inherent biases and moral judgements required for any form of truth detection.
Very unconvincing. The only point they bring up which actually precludes RAM-only servers is hard drive encryption, which they only need to do because they store data on a hard drive. The whole article reads like them trying to justify a choice they’ve already made rather than a legitimate comparison RAM-only versus hard drives.
Their first point is literally that RAM-only doesn’t help when the power’s on. That’s like saying you shouldn’t wear a seatbelt because it doesn’t protect against someone smashing your window. That’s just not what it’s for.