So why isn't translating a frontier model into a local language sufficient? And also what kind of local language does an AI system need before it can operate effectively in an organization?

Yeah.

I think the translation changes the words, but it does not create understanding.

If I translate an operating manual into, for example, Kata language, the model may read the words, but does it know which use is mentality, which asset the rule applies to, which person has authority, or which version is current, and what happens if it's to use config, right? So this is not a language problem.

It's a real-world model problem.

So I think every organization has its own small universe.

It has objects, relationships, use, and history.

What is your view on that?