41–50 of 203 posts
In other words, is the issue in the defaults or is it impossible for AI to respond from other cultures?
Can’t recommend the book that coined this acronym enough: The WEIRDest People in the World Book by Joseph Henrich. It is such an eye-opening piece that explains so much of the world around us. He’s an anthropologist that goes into the psychology of it all. Touching on points like how religion plays a part in shaping the America of today and even how humans are worst at discerning faces today because we need to discer…
> humans are worst at discerning faces today because we need to discern letters and words and dedicate brain power for that. I have absolutely nothing to back this up, but my gut tells me this risks being one of those bold claims that grows legs and runs for a while until we debunk it.
I'm wondering the same thing, in addition to the related question of “Would an LLM perform better or worse if prompted with languages other than English?”.
Can’t recommend the book that coined this acronym enough: The WEIRDest People in the World Book by Joseph Henrich. It is such an eye-opening piece that explains so much of the world around us. He’s an anthropologist that goes into the psychology of it all. Touching on points like how religion plays a part in shaping the America of today and even how humans are worst at discerning faces today because we need to discer…
> women there tend to prefer to be 2nd wives of a better man rather being the only wife of a lower-ranking man This seems like a bit of a tautology; how are they 'ranking' these men?
And probably could change based on the roles of first and second wives and, yes, how male status plays out and how it influences the life of the wife.
We don’t have legal polygamy but in many places there’s not much stopping people from living in an unmarried multi-woman household with a man (or vice versa). But it’s not a very common arrangement, and it’s interesting to think about why.
Earlier quoted context omitted.
> humans are worst at discerning faces today because we need to discern letters and words and dedicate brain power for that. I have absolutely nothing to back this up, but my gut tells me this risks being one of those bold claims that grows legs and runs for a while until we debunk it.
After finding out even Think Fast, Slow (a book from a very creditable researcher and nobel laureate) is full of replication crisis, I approach pop-sci as entertainment instead of self-education.
Ummm... doesn't the AI have to scrape the data of those non- WEIRD cultures to work then? What am I missing here? There are parts of the world where constant person-electronic connection isn't a thing. Is that your point?
“Fancy autocomplete better at completing documents similar to ones it has seen before” isn’t as headline-worthy.
“Do non-American LLMs (e.g. DeepSeek, Mistral, Apertus) perform better or worse here? Do they have their own cultural biases in-built?” I'm wondering the same thing, in addition to the related question of “Would an LLM perform better or worse if prompted with languages other than English?”.
(I did then go and check the book myself; ChatGPT in English was right, the name is there)
Ummm... doesn't the AI have to scrape the data of those non- WEIRD cultures to work then? What am I missing here? There are parts of the world where constant person-electronic connection isn't a thing. Is that your point?
I don’t have the data but I assume the corpus available to train an LLM is majorly in English, written by Americans and western counterparts. If we’re training the LLMs to sound similar to the training data, I imagine the responses have to match that world view. My anecdote is that before LLMs I would default to search Google in English instead of my own native language simply because there was so much more content i…
But my guess is that the data sets used from the other languages are smaller (and actually, even if it had perfect access to every single piece of data on the internet, that would still be true, due to the astonishing quantity of English-language data out there compared to the rest. Your comment validates that). With less data, one would expect a poorer performance in all metrics for any non-Anglophone place, including the "cultural world view" metric.
“Do non-American LLMs (e.g. DeepSeek, Mistral, Apertus) perform better or worse here? Do they have their own cultural biases in-built?” I'm wondering the same thing, in addition to the related question of “Would an LLM perform better or worse if prompted with languages other than English?”.
ChatGPT is worse in Russian. Example: after accurately noting that a name appeared in a particular Russian book, it asked if I wanted the direct quote in Russian. I said yes. At this point it switched to Russian output but could no longer find the name in that book, and then apologized for having used what seemed to have been "approximations" about the book before. (I did then go and check the book myself; ChatGPT in…
For example (not actual output):
Input: "こにちは"(konichwa) Qwen Thinking: "Ah, the user has said "こにちは", I should respond in a kind and friendly manner.
Qwen Output: こにちは！
It quiiiickly gets confused in this, much quicker than in English.
Earlier quoted context omitted.
yeah I just ordered a copy of "We Survived the Night" based on a post here. Never would have heard of it otherwise.
Mind linking to the post for the curious?