want to understand how llms work? 

☺︎
↓

llms, explained with a minimum of math and jargon

want to understand how llms work? 

☺︎
↓

llms, explained with a minimum of math and jargon
https://www.understandingai.org/p/large-language-models-explained-with

Two Q's if you don't mind...

1: Are "word vectors" & "embeddings" the same? Any reason to use the former vs the latter?

2: I read somewhere that word vectors remain largely the same across languages. Does this make word vectors a "universal language?" This would have wild religious / philosophical implications. 🙏

ᴀᴜᴅɪᴏᴠɪꜱᴜᴀʟ - 𝚠𝚛𝚒𝚝𝚎𝚛 - 𝔅𝔢𝔯𝔩𝔦𝔫

https://lilinter.net
https://newmodels.io
https://channel.xyz

1. yes, “word vectors” and “embeddings” generally refer to the same concept, w/ “embeddings” being a broader term that can also apply to other types of data beyond words.

2. cross-lingual embeddings link languages but can’t fully bridge linguistic & cultural gaps since they represent concepts / relationships mathematically rather than fully capturing the depth & nuance of human languages.

Thank you! Re: language, in my head it was something like the “shape” of the relations between all word vectors (imagined as some kind of multidimensional mesh) ends up being very similar.  But maybe it’s just because at that scale the “shape” is just a model of the human experience of reality.