With a lot of FEC, you can sort of approach it, but still not quite. Human brain is a lot better at picking out voice out of noise than any algorithm currently known. That is , for example, why your Alexa cannot take your commands in the middle of a 100-person party, but your interlocutor can hear you and respond.
That is not very relevant since humans don't use an error correcting code that is easy to understand for machines when talking to each other. So Alexa's ability to filter your voice out of the noise has little to do with the ability of two machines understanding each other over a noisy channel when they use appropriate coding.
