I’ve often thought balloon internet aka googles abandoned project loon would be ideal for this use case. Specifically point to pint microwave to receivers near the front line.

This the way I interpret it and didn’t realize until reading this oddly.

This concept is assuming a formalized language would make things easier somehow for an llm. That’s making some big assumptions about the neuro anatomy if llms. This [1] from the other day suggests surprising things about how llms are internally structured; specifically that encoding and decoding are distinct phases with other stuff in between. Suggesting language once trained isn’t that important.

[1] https://news.ycombinator.com/item?id=47322887

I haven’t seen much discussion about this point other than “llm handle languages x y and z because there’s a lot of training data”. Watching Terence Tau using llm for writing proofs in Lean was a real eye opener in this regard.