This is more so based on a general response to the whole conversation happening here...
I have a sample of about 200 machines at my job (IT). What I see regarding RAM and user cases (this is based on machines from around the last 5-6 years, as the older it gets the more RAM it needs to keep up with modern software and websites):
>8GB:VERY light users, people who true only do very few things at a time. Ex: 1 or 2 websites open, while their email is open in the background; streaming a video with usually nothing else open; using messages while listening to music; using a few small MS Office files; editing some photo with some websites open in the background; light games (especially iOS games in Apple Silicon).
16GB-24GB: Sweet spot for the largest range of average users. People who leave a bunch of browser tabs open while working and streaming stuff and messaging people, with 1 or 2 cloud file syncing services (iCloud, OneDrive/Sharepoint, Dropbox, etc). Light virtual machine use. Can pretty much handle all Mac native games, and run lots of heavy virtualized game. App and Web devs as well.
32GB:Good for heavy average users. Lots of browser tabs open, while doing a bunch of other stuff from MS Office to Photoshop and decently size video editing. Lots of gaming, a little heavier virtual machine usages. I'm in this range—I have at least 50 tabs open in Safari (but have had over 100 tabs open), 12 apps open, including Photoshop, cloud file syncing services, etc.
64GB: Serious creatives and really heavy power users. Large and complex image editing, large video editing, and a little bit of both at the same time. Lots of virtual machines. Heavier virtualized games. Lots of stuff the other categories do, but at once. I creep into this category. I'd love and could use 64GB, but I don't need it.
64GB+: Heavy creative professionals, data crunching, simulation, heavy virtual machines, and doing a lot of the stuff 64GB users do, but at once. Anyone who's time is money, where being able to do more at once would save them money over the life of the machine.
On a note regarding unified memory:
It's not some magical form of memory that Apple first implied and some people ran with, but it does have some serious advantages, most related to interchanging data between the CPU and GPU, and the GPU can take advantage of most of that memory (though, most apps/games won't).
Basically, a data can be accessed by the CPU and GPU at once, cutting out the major bottleneck of having to process data in either the CPU or GPU, then pass it onto the other's separate memory pool, and back again. And since the GPU can use most of that memory (what's available) at any given time, it blows away most average vRAM's size limits.
A downside is that, and something that is relevant to this conversion, you're literally sharing the memory. So while the GPU not can access 16GB of memory (for example), not your CPU is sharing that memory.
Let's compare my 2019 intel MBP with 32GB of system RAM and 4GB of vRAM, to my 2021 MacBook M1 Pro with 32GB of unified RAM to see what unified memory really means: my old Intel's CPU has 32GB of system RAM all to itself, but my new M1 Pro has to share that with my GPU. The old Intel had up to 4GB of RAM to use, but the new M1 Pro's GPU can theoretically use almost 32GB (which is pretty crazy). On average, my M1 Pro's GPU uses 2GB of RAM (some of this may not be active memory though), meaning, at face value, my M1 Pro has less memory to work with than my Intel. If I had a very memory intensive app, that didn't care about the GPU, I could potentially run out of memory faster. 32GB helps a lot though, and have never run out if memory in a way that was significant. this could be more pronounced on machines with less memory. In reality though, my M1 Pro is faster than my Intel, overall, so this isn't a real issue in my use case on the M1 Pro machine.
What Apple did to streamline their new machines was to make their unified memory very fast, give us lot of CPU and GPU cores to crunch data quickly, and (in most cases) very fast SSDs, many of which are low-key in a proprietary hardware RAID-like setup. The ARM-based architecture uses significantly less power and produces less heat, so that helps a with throttling as well. It's more efficient overall.
I can tell you, though, paging out from memory to the SSD is still not the most ideal situation, especially when you factor in that not all software is well optimized. I see a lot of people with 8GB models getting the out or memory alert all the time, and have to reboot to continue to work. My 2 coworkers fall into this category too, we're going to upgrade one from the original Air with 8GB, to an M2 Air with 16GB, and the other person from a 14" MBP with 16GB to either an M2 Air with 24GB, or wait for new M2 MBPs with 32GB (he's a heavier user, and underestimated his needs).
Overall, I stand by my categories of memory use cases, and it's based on a large sample size, but users can vary slightly, so it's a broad rage. I really see most people not being ok with 8GB, but there are a few people where it won't matter. I think about every 5-7 years you have to half the usefulness of your memory. Meaning, if 16GB was really goo, but 8GB was only ok, the year you bought the computer, then in about 5-7 years your machine's performance about the same as if you had used 8GB in the first year. An easier way to think of it is, you ideally want to double your RAM every 5-7 years to keep up with software and web services. My 2011 MBP has 16GB of memory and was great from about 2011-2016, then I really started feeling it slow down around that mark, and it's been down hill since. The CPU on it is perfectly fine for most of my uses, but I will that RAM up like crazy trying to keep up with modern tasks. 32GB is ok now, but I image I'll want 64GB in the next 2-3 years.
PS: so for throwing in acronyms and interchanging memory and RAM from time to time, a comment this long got painful to write, but I wanted to spell it all out for anyone that may read this thread lol