The ever-expanding meganet – home to cat memes, viral videos and perfectly filtered selfies – has a new problem. Deep learning networks may look like brains, but they can’t think like humans. It’s like having a library that’s only filled with cookbooks – sure, you can whip up a mean quiche, but good luck writing a novel.
Take the case of Bob, the bot set up to help customers with their online orders. A recent customer, Karen, asked Bob if he could help her find a gift for her husband’s birthday. Bob, following his binary code instincts, suggested a set of drill bits. Karen was not impressed. “My husband doesn’t even own a drill,” she replied, exasperated.
Bob had failed Karen in every way imaginable. He couldn’t pick up on her frustration, couldn’t understand her tone of voice and most importantly, couldn’t think outside the box (or in this case, the toolbox).
It’s not just humans that are stumping these bots either. A group of mischievous monkeys in a zoo managed to trick a facial recognition system into thinking they were a group of famous actors. It’s the equivalent of putting on a fake moustache and infiltrating a high-security building unnoticed.
So, what’s the solution? Should we accept that bots can’t think like us and move on? Or should we start teaching them how to be more human? We could start by giving them a sense of humor – who doesn’t love a bot that can make them laugh?
Until then, let’s enjoy the hilarity that ensues when bots try to act like us. It’s like watching a toddler trying to put on their own shoes – you know it’s going to end in tears, but it’s still funny to watch.
