Hacker News Reader

Comment by yubblegum | Hacker News Reader

Interestingly enough for this morning's walk I was musing over the tension between the hypotheses that: 'LLMs can map between languages in the vector space' (thus languages are ~equivalent); and 'Language affects thoughts' (as in German is good for Philosophy and English for getting things done).

If both these thoughts are true, then it would appear that languages have topological characteristics. We can (topologically) map from one to another, 'thoughts' (that is a complex of words) form 'paths on the language manifold' and certain paths may be more 'natural' in one topological form than the other.

View on HN

My take is the human brain learns concepts primarily through differentiation. To a newly born child who has no concept of door or a wall, has no reason to see the two as being different parts. Different languages form different differentiations, but one can always compound concepts, and differentiate them differently.

To extend: there will also be general alignment tendencies towards those readily mapped and expressed concepts within available language. Hard but useful concepts can get mapped to idioms. Modes of categorization will be influenced by these factors, which in turn influences many processes.

What is a word in one language is a collocated words in another, possibly context-dependent.

We can look no further than English: "man can do something," "man can do not do something" (i.e., can do but does not have to), then pretty straightforward "man can not do something" and, all of sudden, to express that man cannot decline some obligation, we say "man can not help but do."

It is not translation per se, but shows that some parts of language were evolved to tiptoe around non-customary things, in this case, double negation. And double negation is very easy in some other languages.

"'Language affects thoughts' (as in German is good for Philosophy and English for getting things done). If both these thoughts are true..." Well, the second one isn't true (I omitted the first one in this reply). It is simply not the case that German is good for philosophy and English for getting things done, and similarly for most other such claims (French is better for talking about love, Italian is better for operas, etc. etc.).

There is the platonic representation hypothesis, which speculates that as LLMs get larger and more multimodal, they all end up learning isomorphic representations of reality. Maybe for humans something like this is true as well, since ultimately language must be rich enough to capture and communicate reality.

Reply on news.ycombinator.com