These Macbooks will be superseded anyway, so I'd rather wait until the software I'm using is fully supported and optimised than to jump into the first generation of M1 Macs with unoptimised / unsupported software running in rosetta.
But right now, in general? No. they are not good for deep learning.
> In general, they’re fantastic.
The unoptimised software and the missing developer tools says otherwise. Especially for users of deep learning, if it lacks the tools they will not use it at all for this use case.
> The question was "Are they good for deep learning" and the answer was no. No one asked about them in general.
Don't you think the answer is to skip the M1 altogether and in general for developers and deep-learning users? At this point, there is no reason on getting an M1 Mac at all since the software required is not even ready for M1 and the hardware will almost certainly be obsolete this year for M2.
I wouldn't want to be an early adopter on a system that has unoptimised software on it and would be running on Rosetta. The answer to the question above lies in whether if the hardware in the newer generation Mac products is powerful enough for deep-learning. In this case, it is not. So just get a desktop with a RTX 3080 instead or wait for an M4 / M5 Mac.
At a certain point, you need just raw throughput, which requires power, which means a bigger chassis with louder fans, which ruins all the aesthetics of macs.