The next step in AI: Chatbots?

Will Nesbit | @notwillnesbit

Earlier this month, the White House introduced a chatbot that allows users to contact the President via Facebook’s Messenger app. Though more of a novelty (the bot carries out the same function as the “contact” page on, the move is a hint at the emergence of chatbots and similar artificial intelligence (AI) in social media and public relations.

From the onset of increased automation in the early 20th century, people have been fascinated with the idea of conversing with machines. Such dreams were first realized on a large scale in interactive voice response (IVR) technology in telemarketing and over-the-phone customer service in the late 1990s. The tradition was continued in the new millennium with Apple’s Siri and similar voice-recognition software in hands-free car navigation systems. Chatbots being a relatively new phenomenon, first appeared in notable numbers on Facebook Messenger within the last year or so.

Facebook chatbots are pre-programed plugins for Messenger that can serve a wide variety of purposes, but tend to continue the legacy of their IVR predecessors as customer service implements. They are generally able to address and resolve simple, regularly-occurring requests and complaints, but falter in the face of more complex inquires. Many bots tend to quickly delegate responsibility relatively early in a conversation by directing users to call or email their parent page operators.

This is because machines are still inept at dealing with language despite tremendous improvements in AI in recent years.. So much of communication relies on subtlety and context; things that we rarely consciously notice in our daily linguistic interactions, but are paramount in our interpretation of a conversation. In order to reach logistical  conclusions, computers must be programmed to take into account the variables in a  situation, assign the variables statistical precedence, and perform  a mathematical calculation. Furthermore, they must be able to learn from past interactions and alter behavior accordingly. There is much still to be discovered in the field of linguistics, but as of now there is no single agreed-upon way to go about improving computer cognition to perfectly mimic that of humans.

Chatbots are most likely here to stay and will get better at communicating. They offer yet another way to automate a task that had only been carried out by humans in the past. Though technology is currently quite far from C-3PO, we may one day be able to have a conversation online and not be able to tell whether we are talking to a human or a chatbot.