It’s a little inaccurate to say that they’ve made no attempts to make these technologies offline, because there are examples of offline speech recognition, and all of them suck. I’m not arguing in favor of Google Assistant or Alexa, but cloud processing does result in a much better experience.