1000x is ridiculous. What are you doing where that level of improvement is measurable. That means you are doing things that would have taken you a year of full-time work in less than half a day now.

I understand, but the improvement is actually more than that. It is not directly programming, but look at this page [1] for example. I spent years handcrafting parallel texts of English and Greek and had managed to put just under 400 books online. With AI, I managed to translate and put in parallel 1500 more books very quickly. At least 2/3 of those have never been translated into English, ever. That means I have done what the entire history of English-speaking scholars has never managed to do. And the quality is good enough that I have already had publishers contacting me to use the translations. There are a couple other areas where I am getting similar speed ups, but of course this is not the norm.

[1] https://catholiclibrary.org/library/browse/

... you know what. Whilst I suspect the quality of these translations is probably not great. Fair play this is a valid example.

Of course they are not perfect, but no translation is even close to perfect. The floor is actually incredibly low. All I can say is that many doctoral-level scholars, including myself and some academic publishers, find them to be somewhere between serviceable and better than average.

Knowing the quality of LLM translations between the two languages I speak, hearing it used like this by supposed academics invokes a deep despair in me. "Serviceable" is a flimsy excuse for mass-producing and publishing slop. Particularly given that slop will displace efforts to produce human translations, putting a ceiling on humanity's future output - no one will ever aspire to do better than slop, so instead of a few great translations, we'll get more slop than we would ever even want to read.

I guess it does depend on the languages involved; one study suggests that it's even worse than Google Translate for some languages, but maybe actually okay at English<-->Spanish?

> There were 132 sentences between the two documents. In Spanish, ChatGPT incorrectly translated 3.8% of all sentences, while GT incorrectly translated 18.1% of sentences. In Russian, ChatGPT and GT incorrectly translated 35.6% and 41.6% of all sentences, respectively. In Vietnamese, ChatGPT and GT incorrectly translated 24.2% and 10.6% of sentences, respectively.

https://jmai.amegroups.org/article/view/9019/html

I wouldn't have put it online if I didn't think it was a major improvement over nothing. Realistically, if we haven't translated it in the last 500 years, there is no point for the next several hundred years of history to stick with nothing as well. It takes a bit more than pasting sentences in chatGPT to get a serviceable translation of course, but significantly better results than that are possible. I have not tried translating into other languages, but I am sure having English as the target language is a help.

It's all right there on my website in parallel text, everybody can check and come to their own conclusion rather than driving by with unhelpful generalizations. And really, that is the primary scope of these translations: as aids in reading an original text.

This is my favorite type of interaction to see when browsing HN. I feel like this respectful mosh pit of ideas and practitioners is uniquely rad. Thanks for the serendipity dopamine this Sunday