[aunty_helen]

I’ll put this out there, currently there’s no better platform than macOS on apple silicon for developing ai systems.

I’m not a stranger to Linux or the command line. I own, use, configure servers as part of my business, including the dreaded on metal cuda install. In fact, the terminal integration in macOS is one of the biggest things over windows for me.

But, every time I try linux desktop, for the past 20 years, it’s been a horrible time sink and has driven home the point that building a competent and most importantly consistent gui based os is harder than everyone gives it credit for.

I stopped using Linux mint after installing it on my desktop and having the screen saver require a hard reboot -sometimes- when trying to wake.

[ActorNightly]

>there’s no better platform than macOS on apple silicon for developing ai systems.

Lmao WHAT?

The ANE system isn't even remotely useful, since is primarily designed for running Apple AI stuff. This is why its integration is so spotty. IIRC, Tinygrad is faster on apple silicon than pytorch at this moment, solely because they did a whole bunch of reverse engineering.

Laptops for ML is just a lost cause as far as matrix multiply is concerned. Nobody is actually doing any serious work on ML stuff on laptops.

>In fact, the terminal integration in macOS is one of the biggest things over windows for me.

Which is funny, because Windows has WSL2 which works incredibly well, has native CUDA integration for ML tasks thats quite good, has an X server that lets you run GUI apps, and is actually linux (not BSD), without anything to get in your way, and its better because its an isolated system that you don't have to worry about bricking and not having a usable computer.

> stopped using Linux mint after installing it on my desktop and having the screen saver require a hard reboot -sometimes- when trying to wake.

The standard argument of "here is a particular bug that doesn't exist on Macs, therefore Macs are better" lol.

Like I said, its not really that difficult to say that you just prefer the Mac OS experience and end it there. You don't have to go on these weird tangents.

[aunty_helen]

>Nobody is actually doing any serious work on ML stuff on laptops.

Was just training a 52gb radiology model. Please point me in the direction of any other platform that can do that for 5k.

>Windows has WSL2

Macos can run VMs too.

>here is a particular bug that doesn't exist on Macs

Yes, my laptop doesn't require a hard reboot after closing the lid. Linux fans won't understand this. If you want more particular bugs, network drivers, amd video cards, trackpad not clicking on the login screen, screen resizing, external monitors via usbc, TweaksUI as a concept. There's no point pretending this isn't a thing, everyone has a story.

As I said, I'm not new here. I've been using linux for 20 years. Linux desktop is garbage no matter the tangent you want to explore. I have 3 developers using windows so know all about developing ai with wsl and the fun problems that come along with that.

[itsoktocry]

>The standard argument of "here is a particular bug that doesn't exist on Macs, therefore Macs are better" lol.

Yeah, this article is literally about an annoying MacOS bug. I use a Mac 8 hours a day for work; it's a great machine, but I bet I'd have 20 glitches and annoyances that I work around, by reflex, all the time.

MacOS is good, but it's certainly not flawless. And if it wasn't for Apple's magnificent hardware, I'm not even sure it's the best.

[ActorNightly]

> And if it wasn't for Apple's magnificent hardware,

I would agree with this sentiment if people said something like "Apple has gotten the hardware right with the current gen, but the intel macs were colossal pieces of shit in both hardware and software".