[unit_circle]

- Amazing translations (this alone is a game changer)

- Language learning... The ability to improvise realistic conversations is huge. I can ask to talk about cooking a specific dish or a sport!

- As others have noted, refining documents similar to grammarly

- Looking for a product with extremely specific features (though it isn't very good at comparing yet)

- Searches that are too vague and complicated to articulate to a search engine or use exact matching ...

[spaceman_2020]

I don’t think people dismissing AI have access to GPT-4 yet. GPT-3.5 did feel like a gimmick.

GPT-4 feels like a team of really capable human interns.

The potential really goes wild once you connect it to the internet and use stuff like autoGPT

[rrrrrrrrrrrryan]

This makes a ton of sense.

I've been struggling to reconcile my personal experience with what I'm reading - it was so strange reading such dismissive comments by such a knowledgeable community about a new technology that's such an obvious game changer.

GPT-3.5 was easy to dismiss, but GPT-4 is incredible.

[riffraff]

Didn't we have amazing translations for a long time?

I do not notice a significant improvement between new models and last year's deepl/Google translate.

[mxmbrb]

Sounds like you're a native english speaker? Translations to english often worked "well enough" with google. But as soon as you tried the other way arround, or from one non english language to another, the results where often fully incomprehensive. Btw. the same goes for t2s and s2t. It only got slightly usable in the last few years, therefore wasn't really adopted in many non english countries. As you can see, there is a huge market opening up. Then take voice synthesis into the picture. I think the simple amount of recent changes, including LLMs, will steamroll a totally new media envoirenment and with it substantial social and economical changes.

For me its less the capabilities of LLMs, but the speed and the inevitability of change. You can choose to ignore it, but you soon will be outdated then. Just like if someone would try to work an office job without using digital machines. Maybe you can still do it, but who would hire you?

[throwaway2037]

Language pairing is important. Machine translation between English and German/Nederlands has been excellent for years. English and Korean/Japanese: Awful for years. On the last couple, did it get much better. I am sure ChatGPT12 will have virtually native level translation. Maybe it will be integrated into Kindle so you that can buy books in 25 diff languages, then get a ChatGPT-translated version with one button click.

[FranzFerdiNaN]

DeepL is really good, better than Google Trandlate. ChatGPT is about the same, but it can also explain the translation and the grammar. It might sometimes be wrong but it’s usually good enough at explaining things to point in the right direction.

[skybrian]

For translations, I use Google translate. Should I switch?

[fomine3]

Yo, check this out! DeepL is like way better than Google Translate, so you should totally use it. But, GPT-4 is like a game-changer, dude! It's obvi way better than DeepL, and it can do all sorts of stuff like rewording, explaining in detail, and changing up the style. This text was translated from Japanese to gamer-style English by ChatGPT-4, no joke!

[frabcus]

LLMs are better at context - as far as I can tell Google Translate only does one sentence at a time.

With GPT-3 it infers things like gender and formality from earlier context.

Not clear yet how good GPT-4 is as OpenAI won’t say what training data it has, even rough volumes, in each human language.

Needs some thorough research testing it.

[dodslaser]

DeepL [1] usually does better in my experience, but I suspect Google Translate has gotten better since last time I used it.

[1] www.deepl.com