Issue #52: Improve description of Generative AI elements

Dear all,

I made some updates to 
https://github.com/w3c/voiceinteraction/issues/52  Improve description 
of Generative AI element according to the review comments from the last 
call.

Open

  * Update all relevant figures with layers for Generative AI and
    Traditional NLU
  * Clarify if dialogs are to be used for Generative AI, e.g. for error
    handling

Please, check if you have any further comments on this. The current 
draft can be read at 
https://w3c.github.io/voiceinteraction/voice%20interaction%20drafts/paArchitecture/paArchitecture-1-3.htm

Additionally, we may want to discuss the open point that popped up for 
me when I was describing the LLM post processing component.

So far, I was under the impression that the concpet of specialized 
dialogs with their description how to cope with input and decide upon 
the output would only be relevant for traditional NLU based systems. 
However, in the LLM post processing step we may also be in the need to 
make use of, e.g., error dialogs if a call to the LLM could not be 
established.

It might be a limited usage of dialogs here, but this concept may still 
be useful for LLM based system.

Dirk