It’s a shame because “improve an off the shelf llm ti translate in line with this large dataset we prepared” is precisely the kind of project people love to work on. It could have been a chance to immortalize the hard work they did up until now.

▲

eCa 2 hours ago | parent | next [-]

> this large dataset we prepared

You mean the two+ decades of labour of love was always done to be a nameless contribution to the AI machine? Somehow I think he would have picked another hobby if he had known that back then.

▲

ehnto 5 hours ago | parent | prev [-]

Is it? I don't think you quite understood the issue.

This issue is specifically centred around the human element of the work and organisation. The translators were doing good work, they wanted to continue that work. Why it's important that the work done is by a human is probably only partially about quality of output and likely more about authenticity of output. The human element is not recorded in the final translation output, but it is important to people that they know something was processed by a human who had heart and the right intentions.

	▲	ninkendo 4 hours ago \| parent [-]
		> The human element is not recorded in the final translation output, but it is important to people that they know something was processed by a human who had heart and the right intentions Not that I entirely disagree with the conclusion here, but… It feels like that same sentiment can be used to justify all sorts of shitty translation output, like a dialog saying cutesy “let’s get you signed in”, or having dialogs with “got it” on the button label. Sure, it’s so “human” and has “heart”, but also enrages me to my very core and makes me want to find whoever wrote it and punch them in the face as hard as I can. I would like much less “human” in my software translations, to be honest. Give me dry, clear, unambiguous descriptions of what’s happening please. If an LLM can do that and strike a consistent tone, I don’t really care much at all about the human element going into it.