Chat bots, powered by artificial intelligence are all around us – on eCommerce sites, social media, chat messengers, and umpteen other places. But what really captured the attention of world media recently was when Facebook had to reportedly shut down its AI-chat bot experiment when the two bots – Alice and Bob invented their own language to talk to each other. The social media giant was conducting the experiment to train bots to converse with people.
In its attempt to create next-generation digital assistants, Facebook was attempting to create bots that could negotiate deals, apply reasoning, etc. – a step above performing simple rules-based tasks. Left alone to learn their conversation skills, Facebook’s Artificial Intelligence Researchers (FAIR) found the two bots had started developing their own language script without human input.
Why artificial intelligence sounds so creepy to a large majority of people
The Facebook chat bot controversy, which the company later played down by saying it discontinued the experiment not because bots started talking to each other, but because they were not speaking English, proves a solid point – people are paranoid about Artificial Intelligence and its impact on everything – their jobs, future, society. Silicon Valley veteran and tech titan Elon Musk recently said “I have exposure to the most cutting edge AI, and I think people should be really concerned by it.” While companies are increasingly turning to live chat outsourcing using chat bot applications to save costs & provide 24×7 service, not all is hunky dory. Besides the thought of speaking to chat bots not appealing to a large majority of consumers, there are umpteen cases of chat bots interpreting language incorrectly, leading to bad customer service – a pitfall no business can afford these days. Recently, a chat bot being used by clothing retailer ASOS responded to posts from its Facebook page followers by sending them details of their order status when the users were in fact, talking about a complaint from one of ASOS customers regarding the irrelevant automated response he received when he contacted live chat in relation to his return. This case highlights we are still very far from meaningful artificial intelligence applications in customer service outsourcing and that human interface still tops chat bots by a large measure.
Chat bots can have a mood of their own
Microsoft’s Tay chat bot launched last year that could communicate with people via Twitter, Kik and GroupMe, started spewing racist, harsh, genocidal, and misogynistic messages to users within hours of being launched. The company’s response? “Tay learns from human responses and she picked up these words from users attempting to abuse Tay’s commenting skills by teaching her wrong messages.” Amid widespread backlash, the tech giant had to take Tay offline & the incident marked another case of artificial intelligence gone horribly wrong.
How humans are teaching chat bots to talk
The case for deploying chat bots is compelling for most businesses – cost reduction, uninterrupted service continuity, and faster turnaround time. But the language troubles prevent many from going down the artificial intelligence route. In a bid to change that, some companies are now using human copywriters to inject some personality & empathy into bots – two key areas they struggle with. Whether or not this will prove to be a game changer in the artificial intelligence domain is yet to be seen but for now, it’s safe to say that the humans squarely outperform machines, at least in places where the ‘intelligence’ really matters.