R2R vs Delta Sigma DACs - same results?

[antcollinet]

I'm not sure why people use AI to help write their posts other than maybe doing translations?

The thing is that many browsers and such are integrating Ai into every single box.... so maybe its just people pressing a single button to "rewrite with AI".

To be fair - if I were not writing in my native language - I *would* be using AI to rewrite my posts for grammar.

 

[Jimster480]

To be fair - if I were not writing in my native language - I *would* be using AI to rewrite my posts for grammar.

Yes, I use it nearly every day to write posts and translate posts in other languages as I am a part of many international chat groups in Telegram and elsewhere (such as doing international support for my customers).
LLM's truly allow people to communicate in ways never before possible due to proper grammar and formatting as well as the proper use of "slang" and "local" terminology when speaking to different people.
When using a good model (such as Qwen3.5 or newer, GLM 4.7 or newer) you can usually get an accurate picture of what is being said and also reply in an effective manner. It just takes some actual thinking / deliberation on the part of the models, so you can definitely eat up tokens. For me I have local LLM hardware here so most of those tokens are processed locally since I use it so often, but the result is great.

 

[mhardy6647]

Yes, I use it nearly every day to write posts and translate posts in other languages as I am a part of many international chat groups in Telegram and elsewhere (such as doing international support for my customers).
LLM's truly allow people to communicate in ways never before possible due to proper grammar and formatting as well as the proper use of "slang" and "local" terminology when speaking to different people.
When using a good model (such as Qwen3.5 or newer, GLM 4.7 or newer) you can usually get an accurate picture of what is being said and also reply in an effective manner. It just takes some actual thinking / deliberation on the part of the models, so you can definitely eat up tokens. For me I have local LLM hardware here so most of those tokens are processed locally since I use it so often, but the result is great.

Does it say precisely what one wishes it to say?

 

[somebodyelse]

Does it say precisely what one wishes it to say?

If you were in a position to know that then you wouldn't need to use a translator, whether human or machine.

 

[wwenze]

Dunno - astrophotographers always want a dark background.

But you get dark background by buying better power cables.

 

[RexrothPigeon]

Does it say precisely what one wishes it to say?

I use LLMs for translation of extremely nuanced texts very frequently in academic/art contexts, and I've had the good fortune of having a few truly bilingual human translators help assess the results. And yeah, pretty damn close if you give it accurate guidelines. My favorite move is to feed a translation and the original to a different LLM and ask it how well they match and where the weak points are. LLMs are even good at explaining the various choices available for slippery connotation sections.

 

[Jimster480]

Does it say precisely what one wishes it to say?

Basically, yes, if you use good quality models, I typically take a different model or a couple different places and translate what I have translated back into English and see how it comes out.

I use LLMs for translation of extremely nuanced texts very frequently in academic/art contexts, and I've had the good fortune of having a few truly bilingual human translators help assess the results. And yeah, pretty damn close if you give it accurate guidelines. My favorite move is to feed a translation and the original to a different LLM and ask it how well they match and where the weak points are. LLMs are even good at explaining the various choices available for slippery connotation sections.

Yeah, basically this. That's the way to do it if you really want to know that it's accurate.
With some languages, you may want to tell the LLM that you want a conversational or a relaxed form of the translation versus a formal form.