4 ms·
Couldn't the LLM translate it to words that have similar embeddings in our own language? Translation is one of the tasks that LLMs excel at.
by pton_xd 2y ago
Couldn't the LLM translate it to words that have similar embeddings in our own language? Translation is one of the tasks that LLMs excel at.
- dotnet00 2y agoDon't you actually need manual translations to feed into the model first for that? AFAIK LLMs are not sci-fi-esque magic universal translators.
- truculent 2y agoNot necessarily. For example: https://engineering.fb.com/2018/08/31/ai-research/unsupervised-machine-translation-a-novel-approach-to-provide-fast-accurate-translations-for-more-languages/ https://engineering.fb.com/2018/08/31/ai-research/unsupervis... > Training an MT model without access to any translation resources at training time (known as unsupervised translation) was the necessary next step. Research we are presenting at EMNLP 2018 outlines our recent accomplishments with that task. Our new approach provides a dramatic improvement over previous state-of-the-art unsupervised approaches and is equivalent to supervised approaches trained with nearly 100,000 reference translations. To give some idea of the level of advancement, an improvement of 1 BLEU point (a common metric for judging the accuracy of MT) is considered a remarkable achievement in this field; our methods showed an improvement of more than 10 BLEU points. Although, this specific method does require the relative conceptual spacing of words to be similar between language; I don't see how that would be the case for Human <-> Whale languages.
- lossolo 2y agoNo, translation from one language to the other doesn't occur in vacuum, there are millions of examples of translated text done by humans, without it LLM wouldn't learn anything.
- afelixdorn 2y agoYou would need to align the vectors of words in our language to the ones in their language... which requires knowing their language (or enough of it)