Jeremy Howard: CUDA for Python Programmers [video]

It's not like you have a choice, unless you deliberately want to work with tools that aren't best in class.

At this point the better answer is invest in CUDA parity on other platforms, not throwing out the ecosystem to spite nVidia.

Not as bad as promoting X by putting the video on Twitter :-) Here the YT link: https://youtu.be/nOxKexn3iBo

Your comment about promoting X, however in jest, reminded me that I find JH to be...an interesting person. On one hand, who am I to question him taking money from the likes of a16z-"you do you" after all. On the other hand, Fast AI seems to have put considerable effort behind "data/AI ethics." Why bother after all?