On: Finland's ChatGPT equivalent begins to think in Es...

[benatkin]

This will be easy, it will go from mostly English to hyperpolyglot quickly.

[DoingIsLearning → benatkin]

If it's training on available online corpus then it will go quickly mostly for English and Mandarin.

Most countries' classic texts and books are still undigitized sitting in Libraries and public archives.

Also book publishing market and online publishing are proportional to total population, smaller country means less content.

[ben_w → benatkin]

The issue raised in the article is that there may not be enough training material in many languages to do this.

I find this very plausible.