AI-Assisted Steno

My thinking about AI-assisted hobbyist steno:

1. Realtime shorthand is so complex because of the inherent difficulty in translating a shorthand into English in real time (see Theory in Practice).
2. The goal in using AI is to allow steno to be more imprecise, which is probably not a good thing. We should be making simpler theories, not relying on AI to fix what's wrong with our theories.
3. AI lacks any context beyond what has been written by the stenographer. Context-dependent shorthand is often dependent on not just what has been written but what will be written, which cannot be encoded in a realtime system.
4. There is very little guarantee that a system like this could work with any amount of accuracy, especially as context dwindles. What happens when you write a single homophonic word in response to something? How is the AI supposed to know which spelling you intended? AI doesn't really solve much of the complexity inherent to realtime steno, it only adds more.

If the goal is to reduce or remove conflicts, there are ways to get rid of them all, except for homophones, which are inherently difficult to translate in real time.

I don't think the percieved benefit of allowing realtime steno to be more imprecise is worth the inherent complexity that comes along with doing so.