Artificial intelligence is not a term anyone should take lightly. By a strict definition, it means a computer system is able to perform tasks that normally require human intervention.
As companies like HP, Staples, CNN, and Expedia start to roll out Facebook chatbots to help you with mundane tasks, don’t expect anything like Roy Batty from Blade Runner. It’s more like one of those robotic dogs that can perform a limited set of tricks (stand up and sit but not roll over) and are designed not to imitate a human or even replicate a human function but only to steer humans toward online information, making a purchase, or...making a purchase.
In other words, we’re a long way from the Microsoft Tay chatbot. While the programming for that chatbot became a puppet for Internet trolls, at least it could pretend to talk like a human. And, while I found the chatbot to be offensive and even creepy, Microsoft deserves some credit for trying to imitate human speech.
I tested out several of the Facebook chatbots and felt a bit like a mouse being led through a very tightly controlled maze. The entire purpose is to lead you to the cheese.
My first attempt was with Poncho, a "chatcat" that tells you the weather. It was underwhelming and a little too safe. What’s supposed to be slightly cheeky and sarcastic turned out to be totally dull. The cat made one slightly cheeky comment about preferring warmer weather to the rain and cold in my area today. When I made a joke myself, it told me it only chats about the weather.
The Spring shopping bot seemed promising at first. It’s designed to guide you quickly into a discussion about specific products. You select whether you are interested in men’s clothing or women’s clothing. Eventually, I drilled down and the bot suggested some running shoes. When I asked for more details, a human stepped in and took over.
The CNN bot is even worse. It won’t even engage with you at all, merely sending news listings. When you start asking questions, it just sits there and doesn’t respond.
The only chatbot that impressed me a little was the one for HP. It didn’t engage in any banter at all, but it did guide me through the printing process, asking for my HP ePrint email address. I loaded up some photos and the bot happily printed the docs. Sadly, when my printer asked for a paper size I don’t have and I asked the bot for help, it prompted me to print more stuff.
Likewise, with the 1-800-Flowers bot, it kept asking for my shipping address. When I said I didn’t want to give out my address again, it asked me to type in my address. It felt like I was in an eternal loop with an AI bot that is only marginally helpful and maybe even too assertive.
I doubt we’ll see any news reports about Facebook chatbots going rogue. If anything, it made me wonder if this is all going to work out as planned. If the chatbots can’t go off script at all, they are not that useful -- as humans, we love to go off script. The main takeaway from testing the bots so far is that they all seem rather dull and lifeless.
It almost made me miss Tay.