RE: New NSA patent explicity mentions machine transcription

[William P. McGonigle <William.P.McGonigle@artoo.hitchcock.org>]

--- Fisher Mark wrote:
Once
you have speech-to-text for multiple speakers working
--- end of quote ---

That's the hard part, I think.  MIT's Jupiter is super-cool, but I think it's
only working in English and it can take 5 seconds or so for a response from
their cluster (but what a good job it does). 

I'd think the NSA would need to do better in real-time, for hundreds of
thousands of channels simultaneously.  To me, this seems infeasible with the
MIT-level tech, even for the NSA.  They can't really thrown petaflops at
problems, can they?

So, they (in my opinion) must've figured out a better way to attack the problem.