Hi!
I agree that ai generated code should not silently be added to packages.
It could be acceptable to use ai to do translations, if the message
files are text. No ai-based change of code which prints stuff, please.
It can be okay to have app code written with help of ai, but I want
to have the choice whether I want to use such packages. Sometimes,
people who cannot write software do ask ai to write software, which
means there will be no proof-reading and quality will be limited.
This also means that if somebody adds ai stuff to an existing
package, I want to be able to select the non-ai version while
the ai-enhanced version should start a fork.
It is okay to use ai to find bugs, in particular security flaws,
but fixes for those bugs have to be written and checked by hand.
I doubt that ai is useful for documentation. If a package fails
to have a decent /? screen and readme, ai is unlikely to improve
the situation. I expect ai to write wordy, eloquent descriptions
of what the software seems to do, while missing key points :-p
Of course this could get better as ai will continue to improve.
Regards, Eric
PS: What is the conclusion regarding firefox here?
https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791
The ai based firefox hardening probably did make sense:
https://blog.mozilla.org/en/firefox/hardening-firefox-anthropic-red-team/
_______________________________________________
Freedos-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freedos-devel
