Hi all,
I understand this topic might be controversial.
First and foremost, let’s be clear: being part of GNOME means being part of a community. What many of us look for in this community is a safe place to build relationships, to learn, to be ourselves, and to express ourselves for the common good.
All that being said, productivity is clearly not what contributors primarily look for.
I understand that in some teams with limited resources, it might be tempting to use generative AI to assist with the translation process. Some contributors might already be doing so without our knowledge. Please keep in mind that many of us have already checked a translation or a grammatical form using tools like Google Translate or DeepL, just to ensure that these tools “understand” the same thing we humans do.
Stating that current AI tools used for translation - especially large language models or specialized platforms like DeepL - do not provide quality translations is inaccurate. Microsoft, for instance, uses automatic translation, and while it can sometimes produce poor results, we should not focus solely on failures to draw general conclusions.
In languages with extensive linguistic resources (such as Spanish, English, and French, languages I have worked on during my PhD), the question of automatic translation is largely settled for everyday usage. For languages with fewer resources (such as Finnish or Estonian, which I have also worked on), achieving high-quality standards has historically been more challenging. However, this may have changed with recent progress in large language models.
The main criterion, in my view, is consistency. It is unacceptable to have different translations for the same term within a single module or across multiple modules when used in the same context. A glossary provided to an LLM could, in principle, help address this issue.
So, this is not a question of productivity, nor is it really a question of performance for widely spoken languages (even though variations exist - for example, French as spoken in France differs from French in Québec or Senegal).
The GNOME community is deeply concerned about the future of our planet and the quality of life of the people living on it. It is also concerned about copyright issues and the appropriation of energy, work, and effort by large tech companies to train their models.
So, my conclusion would be the following (and it is neither definitive nor authoritative): we do not prioritize performance or productivity above all else. We want to foster a welcoming and safe environment where contributors can express themselves. Participating in translation projects is a meaningful way for many contributors to be involved: it is not a full-time commitment, often happens in bursts a couple of times a year, and feels “safe” in the sense that contributors are unlikely to break anything.
That said, I can understand the use of LLMs for low-maintenance applications as a form of assistance. Maintainers remain responsible for their work, and for some, these tools may help them stay involved during transitional periods in their lives (for example, when raising a child).
| 
. You are the one with the reminder.