Best post in this thread, good job.
3 Likes
I’m worried that sort of tone might be counterproductive… 
and I don’t think productivity concerns are entirely unjustified for any project to think about.
2 Likes
its a welcome reality check in my books
4 Likes
It’s not entirely unjustified to think about, but the gains are questionable and the costs are massive.
2 Likes
High viscosity truths need low gauge syringes.
3 Likes
I would very much like to see Delta Chat adopt a policy like the one that the Zig language adopted recently:
Strict No LLM / No AI Policy
No LLMs for issues.
No LLMs for patches / pull requests.
No LLMs for comments on the bug tracker, including translation.
English is encouraged, but not required. You are welcome to post in your native language and rely on others to have their own translation tools of choice to interpret your words.
Delta Chat is a project meant to help connect people and build community. As many other folks have argued in this thread, LLMs and other contemporary GenAI tools do not do that. They centralize power in the hands of some of the worst people I can think of. These are people who say things like “We see a future where intelligence is metered like electricity.” Also, recent research indicates that usage of an LLM stops new learners from developing skills, and causes atrophy of skills in people who already knew how to do something.
Let’s build tools for people. Let’s provide learning opportunities for new developers. Delta Chat can be a tremendous force for good if we set solid community standards in addition to the excellent technical standards already in place.
5 Likes
Are other kinds of machine translation still allowed, though? We had usable rule-based engines since the '90s and then decent “small” statistical models since the '00s. The model used in the present offline translation feature in Firefox weighs a few megabytes per language.
Although, the problem with relying on each recipient to translate the content on their own is that each of them could arrive at a slightly different meaning. This is aggravated by the presence of poor grammar on the input side that a recipient couldn’t judge either.
I feel like translations might be part of the least worrying problems with AI. In terms of both morals as well as potential real confict and claims.
(I could be completely wrong here, this isn’t legal advice.)
I personally don’t like AI translations once the output gets baked into a FOSS project. I’d rather have humans do that then, just to be safe.
2 Likes