Glad to see this post in The Economist.

"The Internet" is probably close to 99% English. I believe there are efforts to train LLM's in various European languages - Dutch are doing something, French of course. However, I doubt any of these will have large enough dataset, especially in scientific domains, and that which does exist might well be already AI translations of the original English.

China might (again) be the only non-English alternative, but even here there are will be significant constraints not least the contextual nature of the language, the differences between written and verbal, as well the prominence of superapps whose data will not be easily accessible as stuff on websites.

Future is going to be dominant English AI vs WeChat AI