Title: Why chatbots fail
Context: Chatbots are the new hotness. Or as a chatbot might say: “Can you rephrase that? I didn’t understand what you meant.”
Synopsis: Everybody and their natural language processing framework wants to build a chatbot these days. And why not? Isn’t it so much nicer to have a two-way conversation rather than click some buttons? Of course it is. Unless of course, clicking some buttons is infinitely easier and more efficient than hoping against hope that the 8000 nested if/then statements that comprise the brainpower of most chatbots are capable of parsing your typed commands. Most likely they can’t. Because language is hard. What’s more, most humans have a hard time discerning emotion, sarcasm, and the occasional autocorrected nonsense word from a brief line of text, so how can one expect some mere computer code to manage it effectively? Most likely you shouldn’t. But these are early days yet. Certainly, we are but a simple technological breakthrough away from pseudo-sentient chatbots capable of handling the most arcane and illogical human communications. Realistically though? Most likely we aren’t.
Best Bit: “Bots that do one thing well are more helpful that bots that do many things poorly.”

via chatbot.fail