When I sit next to my friends kids or my nephews the conversations they have are far more nonsensical when looking at one page of chat. And then they are well educated; there are examples of chat and sexting in particular which look like gibberish altogether.
I would catch this bot out on the first line because of the grammar: people not only use shortcuts for everything (I was asked to check a resume and it contained u instead of you.....), they also do not consider grammar or syntax: their there ther tere ter tr are the same thing etc. Not sure if all of this goes for Chinese as well but in English just spouting nonsense using those rules will get you somewhere. It also helps most chats are with many at the same time (the company driver in China was chatting with 30+ people on wechat simultanously, rapidly switching between users), people generally do not remember or read context for that reason (my younger 20something colleagues some of which are brilliant, when something scrolled of the screen and you refer to it they ask what you are talking about).
The problem is that people on HN want Einstein level convo with perfect syntax and grammar while most people, if you do not tell them it is not a human, would call this chatbot intelligent. They just need to make it more human like short attention span (a 10 minute continues chat is unlikely as are immediate answers) and ofcourse not telling people upfront it is a bot...
That is chat bots: I would think a Facebook Britain First bot would be more feasible; I do not think anyone would ever guess it would be a bot. Markov chains with some heuristics won't do (much) worse than people like [1].
It is also about the vague defition of AI; even if it can learn then a lot of people would not call it AI anymore. Seems for many AI needs the be AGI to really be considered.
>When I sit next to my friends kids or my nephews the conversations they have are far more nonsensical when looking at one page of chat.
Sure, but you are referring to physical conversations which contain all sorts of context and levels of non-verbal interaction which are not captured in a transcript. They aren't really comparable, and I suspect that childrens' chat logs would make a lot more sense than transcripts of in-person conversations.
I would catch this bot out on the first line because of the grammar: people not only use shortcuts for everything (I was asked to check a resume and it contained u instead of you.....), they also do not consider grammar or syntax: their there ther tere ter tr are the same thing etc. Not sure if all of this goes for Chinese as well but in English just spouting nonsense using those rules will get you somewhere. It also helps most chats are with many at the same time (the company driver in China was chatting with 30+ people on wechat simultanously, rapidly switching between users), people generally do not remember or read context for that reason (my younger 20something colleagues some of which are brilliant, when something scrolled of the screen and you refer to it they ask what you are talking about).
The problem is that people on HN want Einstein level convo with perfect syntax and grammar while most people, if you do not tell them it is not a human, would call this chatbot intelligent. They just need to make it more human like short attention span (a 10 minute continues chat is unlikely as are immediate answers) and ofcourse not telling people upfront it is a bot...
That is chat bots: I would think a Facebook Britain First bot would be more feasible; I do not think anyone would ever guess it would be a bot. Markov chains with some heuristics won't do (much) worse than people like [1].
It is also about the vague defition of AI; even if it can learn then a lot of people would not call it AI anymore. Seems for many AI needs the be AGI to really be considered.
[1] http://imgur.com/RJThyAX