On Mon, 2025-05-19 at 14:53 +0930, Justin Zobel wrote: > > On 19/05/2025 14:35, Konstantin Kharlamov wrote: > > > > > On Mon, 2025-05-19 at 10:03 +0930, Justin Zobel wrote: > > > > > > > > > > > > > > On 18/05/2025 16:41, Albert Vaca Cintora wrote: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > On Sun, 18 May 2025, 08:59 Justin Zobel, <jus...@1707.io> > > > > wrote: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > If the contributor cannot tell you the license(s) of the code > > > > > that was used to generate the code, then it's literally > > > > > gambling that this code wasn't taken from another project by > > > > > Gemini and used without their permission or used in a way > > > > > that violates the license and opens up the KDE e.V. to > > > > > litigation. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I'm no lawyer but I would expect that training AI will fall > > > > under fair use of copyrighted code. If that's not the case > > > > already, it will probably be soon. The benefits of AI to > > > > society are too large to autoimpose such a roadblock. > > > > > > > > > > > > > > > > > > > > Albert > > > > > > > > > > > > > > > > > > > > > > > > > > > > From my understanding (what others have told me), AI generally > > > does not produce good quality code though. So how is that a > > > benefit to society? > > > > > > > > > > > > > I wrote some lengthy answer here, but then I scratched that because > > I realized your question can really generate tons of lengthy > > replies that no one will read 😅 I will say you that: AI is useful > > for simple and tedious tasks. In general, you don't expect that AI > > will complete correctly whatever you asked it to do. Instead you > > expect it to give you some useful base, which you can > > change/correct/modify to fit whatever you actually need. > > > > > > > > > > Like, I dunno, do you have a friend in a foreign country who you > > want to write a recent story, but the story is in english? You ask > > AI to translate it, which will be don "almost good", so what you do > > then is you go over the text and correct everything to match your > > style. This is faster than translating everything manually. In > > fact, it well matches what people-translators were doing for > > decades: they typically translate texts in two phases, one is sort > > of writing a scratch, and the other one is polishing, like adding > > suitable idioms, etc. > > > > The problem is we're not talking about text here, we're talking about > code and code has licenses, on which language models don't care > about. I'm all for AI that helps humanity, but stealing code or using > code that is incompatible with KDE's license set is not it. > > I want AI to solve world hunger, prevent disease and help me do the > housework :) >
Well, you asked how "low quality AI code" benefits society. I have many different examples, I came up with one related to translation because it was the simplest one to describe (all others resulted in too much text) and its morally equivalent, because you similarly get poor low- quality base and you improve upon it. If you want code examples specifically: another simple one that comes to my mind is I have in my Emacs config a code that allows to quit emacsclient with vim-command `xb` and upon doing so it converts Markdown code to bbcode. And there's also similar one for textile. The "conversion code" is a terrible O(n²) algorithm written by AI, which didn't even work when AI wrote it. But I fixed it (as in, made it work), and use it just fine at work to write Markdown text and then insert into to either Redmine or Bitrix at their native markup. Don't care much for O(n²), because for amounts of texts I write it is instantaneous. Another example is I had to rewrite some project from one language to another at work. This is tedious task, so I similarly used AI to create a scratch, and then I went over that code refactoring it and fixing.
