> "Speech recognition software" responds in a very humanly voice. Speech recognition software responds to any voice it was trained to. It will respond to dog b

> "Speech recognition software" responds in a very humanly voice. Speech recognition software responds to any voice it was trained to. It will respond to dog barking or morse code it there will be a demand. It responds to "humanly" speech, but it likely mostly ignores the tone. (Although, I imagine eventually Alexa will learn to start different music depending on the tone of the person asking to play the music.) > It's the very reason people want software assistants to have human voices, even female voices, because it makes them feel more at ease. No, it's different reason. Human to assistant and assistant to human is not symmetrical. It's hard to retrain humans to respond to a different form of speech, but it's easy to retrain the model to whatever is needed/used by users.