I don't like the smart replies, but for a different reason. It will only work under very generic situation with borderline functional replies. For me, that kin

I don't like the smart replies, but for a different reason. It will only work under very generic situation with borderline functional replies. For me, that kind of conversation is just a small part of the every communication. Human conversation is more organic than the demo tries to lead you to think. Details in the words and sentences convey very subtle but identifiable personality traits that the current algorithm cannot capture. So, it is a really cool feature and a great showcase of recent advancements in applying deep learning to natural language. But it is pretty much a gimmick for now, and I will stay sceptical about whether this is going to turn user around and become a killer feature of Allo, after all if my friends didn't use this, neither will I.