Hmm...
The US chatbot got taken off because it began to resemble the people chatting to it, and the Chinese chatbot s got taken off because...
Two chatbots have reportedly been removed from Chinese messaging app QQ after issuing distinctly unpatriotic answers. According to the Financial Times, chatbots BabyQ and Xiaobing (or Xiaoice) had been available to some of the 800 million users of Tencent's app QQ until Wednesday. However, the pair seem to have been hastily …
It's either that, or re-tweet cat pictures.
If they want the bots to learn to look like reasonable people, don't "train" it with the nasty trolling behaviour of the twatterati, nor the breakfast image posting farcebook posts along with their corporate "friends".
Not sure if they should have expected a different result - I am not however sure where you can train the AI to be a reasonable person on the internet any more?
Just imagine what it would do if presented with the el-reg forums!