It is valid when comparing similar workloads on a common architecture and has far more relevance than picking the RAM typically required for an Intel system and arbitrarily applying that to the M1 SoC - that is the comparison that doesn't make any sense.
Sorry, but 1GB of bitmap images or sound samples loaded into RAM on an Intel machine will occupy exactly the same 1GB of RAM in an Apple Silicon machine. Running 8 Virtual Machines that each need 2GB of RAM is still going to take up 16GB. Unified Memory
may well speed up the rendering of images from RAM to screen but it's not - in any systematic way - going to make them take up less RAM - and compared to the higher-end dGPUs in some cases it's going to use
more RAM because the ASi GPUs don't have separate VRAM for screenbuffers etc.
The A-series & M-series chips are certainly faster at graphics rendering, compression/decompression and image processing than comparable (i.e. low power, integrated graphics) Intel chips. We
know that the A14/M1 have new Metal-optimised GPUs and hardware-accelerated codecs and neural engines etc. For some workloads
that aren't limited by the *size* of RAM this is going to make them faster, or able to do things that Intel machines can't.
Maybe, in some scenarios, you'll be able to reduce the RAM buffer size and save a bit - but that will depend on exactly what you are doing, what file formats you are using, what storage you are using and - not least - what software you are using: FCPx and Logic will doubtless have been heavily optimised by Apple as "flagship" programs for their new silicon, but other people need to use third-party, cross-platform applications.
Generally, 8 or 16GB RAM on a Mac is more than enough for the majority of uses and will let most applications do their basic job smoothly. The main reason for expanding your RAM beyond that is because you
need to work with large chunks of data loaded into RAM and those sorts of needs aren't going to go away with Apple Silicon.
Unless, of course, you don't
actually need that much RAM on Intel, either (fast SSDs replacing mechanical hard drives have greatly reduced the performance hit when your machine starts swapping RAM - but Intel Macs have had fast SSDs with seek/multi-file access times an order of magnitude better than spinning rust for years now, the slightly faster SSD on M1 Macs isn't going to be a night-and-day difference there).
(I mean, I probably don't really need the 24GB of RAM in my iMac either - but because it was an iMac and I didn't need to pay Apple's atrocious RAM prices, there was no point in skimping...)
This really is a daft argument - 16GB max is perfectly adequate for these low-end Macs and no different from the last generation. The ridiculous 8GB upgrade price is no more ridiculous than it was for the matching Intel models (and slightly less ridiculous than $200 for a pair of plain old SODIMMs for the iMac). No magical Unified Memory fairy is going to appear and turn that 8/16GB of "Intel RAM" into 32GB of "ASi RAM". (8GB -> 9GB? Maybe.)
When the higher-end Macs appear, Apple will need to offer more RAM, and probably will.