Live data from Hacker News

Xiaomi MiMo Reasoning Model

github.com

71–80 of 203 posts

Re: Xiaomi MiMo Reasoning Model

#71
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

All LLMs are trained on the same basic blob of data - mostly in English, mostly pirated books and stuff.

Re: Xiaomi MiMo Reasoning Model

#72
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

The mandarin language models obviously exist, but what would you do with them if they provided access to them? And what knowledge would be in them? What is the body of knowledge encoded in Mandarin? What does that look like?

Sad reality is that not many outside of China have the facility with Mandarin to use those models. Even non-native Mandarin speakers who claim to be "fluent", are often messing up intended meaning in text. Or making literal translations that wind up making no sense.

Inside of China, llm use will be Mandarin based. Outside, it seems to me English is the natural choice.

Irony of Irony, probably the best way for a non Mandarin speaking layman to test a Mandarin based model would be to use another LLM to translate prompts to Mandarin.

It's a sad future we're looking at.

Or a brilliant one.

Time will tell.

Re: Xiaomi MiMo Reasoning Model

#74
post #70
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

I assume a large portion of high quality training material is in English

You'd be correct. The largest portion of all languages in Common Crawl (aka the "whole open internet" training corpus) is English with 43%. No other language even reaches double digit percentages. The next biggest one is Russian at 6%, followed by German at 5%.

Re: Xiaomi MiMo Reasoning Model

#75
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

English won. The Chinese youth struggle to write their own calligraphy characters they can read now. Typing favors English.

Re: Xiaomi MiMo Reasoning Model

#76

Open Source or Open Weights?

And this point everybody will open source their models or weights. The only one which will not is open AI.

> The only one which will not is open AI.

I think you meant Anthropic. OpenAI is "planning" to release an open weight model this year likely competing against the Llama models. [0]

I have not seen an open weight AI model ever being released by Anthropic at all.

[0] https://openai.com/open-model-feedback/

Re: Xiaomi MiMo Reasoning Model

#77
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

Nearly everyone in the urban areas of China spoke some English when I visited way back in 1995. It's a bilingual society.

Re: Xiaomi MiMo Reasoning Model

#78
post #72
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

The mandarin language models obviously exist, but what would you do with them if they provided access to them? And what knowledge would be in them? What is the body of knowledge encoded in Mandarin? What does that look like? Sad reality is that not many outside of China have the facility with Mandarin to use those models. Even non-native Mandarin speakers who claim to be "fluent", are often messing up intended meanin…

For it to be brilliant, AI needs to be a benevolent tool all the time. It would take just a few malignant actors to turn our world upside. I suspect it'll follow the same Internet and social media path. Great at first, grow markets, bring us together and then take a turn.

Re: Xiaomi MiMo Reasoning Model

#80
post #65

Why are there so many English-first AI models from China? Are they not interested in serving their own population? Or is it that if they publish Chinese-first models it won't get publicity in the West?

One reason is that there is no "good" search engine in China. The most popular one, Baidu, is like garbage compared to Google search. The most useful training data in Chinese would likely be from the social media and video sharing platforms, which I guess is much more difficult to crawl and clean up.
Post reply on HN