> The approach used by desktop operating systems with TPMs is awful and makes security worse in a lot of ways rather than better. It's not at all the same thing, similarly to how what the desktop world calls secure boot is not a serious or complete implementation of it and doesn't provide nearly any useful security properties to end users unlike iOS or AOSP.
Interesting. Could you elaborate? I always thought that Secure Boot is reasonably secure (provided you can set it up in the first place).
The standard UEFI Secure Boot used with traditional Windows and traditional desktop Linux is an incomplete and insecure implementation of secure boot. It doesn't verify the vast majority of the OS, nearly always lacks an unbroken chain of trust from the hardware, does not verify all of the firmware and lacks downgrade protection for both most of the firmware and the OS. In the typical setup, it trusts an enormous number of keys and software. It barely provides any useful security properties.
Secure boot was widely used as a term prior to the UEFI usage of the term including many much more meaningful implementations. The term verified boot refers to the same concept but avoids it being confused with solely verifying a late stage bootloader and OS kernel from UEFI firmware.
Ah yes, so if I understand correctly, you're referring to the fact that, e.g., on Linux a typical Secure Boot setup will verify the kernel (if at all) but not the rootfs. Yes, that's of course a huge issue (as is mutability of the rootfs in the first place). I think what I had in mind is indeed more akin to the "verified boot" setup you described.
I think you're correct, but only if you add "for a given environment". It's good to develop the ability to store super efficiently fat when you're a Pacific Islander and food is scarce, but this advantageous trait becomes a liability with modern and plentiful (junk) food.
see thrify gene hypothesis; I know it's not a good explanatory theory but it's a good illustration of my point
That was kind of my initial point... there are far too many permutations that alter what you want to maximise to consider this as a good option right now. Environment, gene-environment interactions, epigenetics, polygenecity, pleiotropy, etc, etc.
Exactly! As I've argued here on HN before, such an "LLM in a box" might end up being serviced/upgraded once or twice a year by a company very similar to the one servicing the coffee machine at the office. In contrast to databases, storage, etc. it doesn't matter much if the box breaks at some point – they'll just come by and replace it with a new one – and there's barely any software on the box to speak of, at least none that requires continuous development and feature upgrades, beyond rolling out security patches. This makes the business case drastically different from cloud and SaaS offerings, where most of the moat is in the software and the state maintenance (and the vendor lock-in of course). The LLM in a box is destined to become a commodity.
Interesting. Could you elaborate? I always thought that Secure Boot is reasonably secure (provided you can set it up in the first place).