"vision and speech are solved"?
WTF is Gates smoking? It wasn't that long ago a Tesla plowed into a fire truck at full speed because it couldn't "see" it, and those of us who talk to phones, smart devices, IVR systems etc. know how fallible that is.
I agree with him that understanding text is a holy grail, because that will make speech recognition actually useful, instead of just a gimmick to tell your phone or whatever to do a simple task you could do yourself in a few seconds.