Hacker GitHub

← All repos

ogkalu2/Human-parity-on-machine-translations

View on GitHub ↗ ★ 31

Bilingual (or Multilingual) Large Language models and In-context Learning- The key to human parity on machine translations

17 mentions on Hacker News by 1 person: 2 stories, 15 comments · first 2023-02-15 · latest 2023-09-14

Mentions per year 17 mentions · busiest year 2023 (17)
20082026
AllStoriesComments NewestTopOldest
  1. It's weirdly flown under the radar but GPT style models as translators are a lot better than state of the art Machine Translators (Deepl, NLLB, Google etc). Like above, you can already see the difference with close language pairs where…
    comment by og_kalu · 2023-09-14 · in Playing Around with Machine Translation
  2. GPT-4 is a much much better translator than Google Translate and the like. You should absolutely be using GPT for translations especially for distant language pairs that quickly devolve into nonsense with Google Translate, Deepl etc…
    comment by og_kalu · 2023-06-29 · in Gitlab’s AI-assisted code suggestions
  3. If a typical LLM has decent representation of the languages in question (and you'd be surprised how little decent is with all the positive transfer that goes on during training) then outsourcing translation is just a downgrade. a pretty…
    comment by og_kalu · 2023-05-10 · in Hugging Face Releases Agents
  4. I know what the paper says but actually using both, GPT-4 is far ahead of google and Deepl. I think the isolated one sentence datasets used for evaluations are no longer up to snuff. Trying something longer and more comprehensive and the…
    comment by og_kalu · 2023-05-09 · in LLMs are not greater than the sum of their parts: researchers
  5. ▲ 1
    Large Language Models for Machine Translation
    story by og_kalu · 2023-04-18 · 0 comments
  6. https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-04-07 · in What it feels like to work in AI right now
  7. I didn't try with old poetic stuff. Passages sampled from 5 books released in the last 2 decades. You can see what I did thoroughly here. Before GPT-4. Basically a comparison between GLM-130b (English/Chinese model) vs Deepl, Google…
    comment by og_kalu · 2023-03-28 · in ChatGPT outperforms crowd-workers for text-annotation tasks
  8. i don't compare with GPT-4 and it's chinese but https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-03-27 · in Ask HN: What is a specific use of GPT-4 that you think is remarkable?
  9. ▲ 2
    Chinese-English Machine Translation Comps with LLMs and Traditional SOTA Models
    story by og_kalu · 2023-03-17 · 1 comment
  10. Tested these before GPT4 but 100%, Bi/Multi-lingual LLMs are the key to solving Machine Translation. https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-03-16 · in Anyone else witnessing a panic inside NLP orgs of big tech companies?
  11. I would disagree. We have image generation with a variety of architectures. Diffusion models aside, it still takes a lot less parameters to model State of the art image generators with transformers (eg Parti). Simplifying a bit, mapping…
    comment by og_kalu · 2023-03-10 · in Visual ChatGPT
  12. The complexity of what is learned is rooted in the complexity required to complete the task. Predicting the next token may seem deceptively simple but you have to ask yourself what it takes to generate/predict passages of coherent text…
    comment by og_kalu · 2023-03-10 · in Visual ChatGPT
  13. Bilingual Large Language Models are much better than the ilk of google translate. https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-03-10 · in Show HN: ChatGPT-i18n – Translate websites' locale json files with AI assistance
  14. I tested Chinese-English translations on a properly bilingual LLM and the results are amazing. You might be interested in seeing https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-02-15 · in What is ChatGPT doing and why does it work?
  15. I think it's a matter of training corpus. Here's how a bilingual equally corpus bilingual LLM does on Chinese-English translations. https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-02-15 · in What is ChatGPT doing and why does it work?
  16. Hey you might this. Bilingual LLMs really are human level translators. I don't know why this frankly mindblowing fact isn't discussed or researched more but they are. https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-02-15 · in What is ChatGPT doing and why does it work?
  17. Bilingual LLMs are human level translators. I don't know why this frankly mindblowing fact isn't discussed or researched more but they are. https://github.com/ogkalu2/Human-parity-on-machine-translati...
    comment by og_kalu · 2023-02-15 · in What is ChatGPT doing and why does it work?