Conversation
Jarkko Sakkinen
jarkko
Edited 3 days ago
I'm not sure if I have any practical use as a maintainer for any possible AI guidelines because such priority emphasizes the task of writing or generating code. It is for me as important for a code review as whether one used search-replace functionality of a text editor while writing the kernel patch.
#linux #kernel #ai
#linux #kernel #ai
1
0
1
The situation is a bit analogous to AI generated apps.
E.g., one can create a Tetris clone with AI but it is not likely to be the one consumers end up buying.
Every kernel patch is also similar puzzle i.e., we are seeking for the best possible option, not just any possible option. And if the best known option is not good enough, a feature is likely get simply postponed.
As of today, AI is not great for code with is aimed to be "commercially viable" i.e., usable e.g., in data center across the globe. It's more like something substituting the same space where e.g., Microsoft Visual Basic used to live, when it comes to writing software.
In the context of kernel development it is literally comparable on using GNU sed and similar text processing tools, and I'm not going to throw stones on anyone using those tools either :-)
E.g., one can create a Tetris clone with AI but it is not likely to be the one consumers end up buying.
Every kernel patch is also similar puzzle i.e., we are seeking for the best possible option, not just any possible option. And if the best known option is not good enough, a feature is likely get simply postponed.
As of today, AI is not great for code with is aimed to be "commercially viable" i.e., usable e.g., in data center across the globe. It's more like something substituting the same space where e.g., Microsoft Visual Basic used to live, when it comes to writing software.
In the context of kernel development it is literally comparable on using GNU sed and similar text processing tools, and I'm not going to throw stones on anyone using those tools either :-)
1
0
1