Humanoid intelligence is more than just skin deep
First published on CIO.co.uk, July 2016. Sticking a face on technology is still a cheap trick used regularly.
There’s a friend of mine, a colleague from back in the days when I worked at the BBC, who is a touch obsessed with the psychological concept of pareidolia. Every week she seems to have found another photograph of a dishcloth or toaster or towel rack that resembles the features of the human face. We appear, as a species, to have a very deep-rooted subconscious ability to see faces in things that aren’t in any way a face. No doubt it’s some sort of early evolutionary thing that enabled us to rise out of the swamps.
I’ve been thinking about Tracey’s face photos a great deal recently as I hear yet another story about the rise of robots, or the intelligence of the machines that are going to put us all out of work. And the thing that I have been thinking about specifically is whether a machine with a “human face” is going to be more likely to fool us into believing they are bestowed with human powers of intelligence?
Put it this way – we’ve had robots in our houses for decades. Nobody assumes that their washing machine is going to rise up and overthrow its human overlords (well, not unless you have overloaded the drum before a particularly punishing spin cycle). However, a Furby appears to have a soul with its big eyes and strange chirping. The humanoid and canine forms being built by Boston Dynamics are truly terrifying, but I’m sure mostly because they look like living creatures. They are no more alive than the Spin Dryer or the toy robot.
The masterful way in which IBM have named their Artificial Intelligence brand “Watson” strikes me as another way in which a “human face” has been put on the front of a machine to allow for some sort of emotional connection to be made. Watson does this, Watson does that – it should never be forgotten, of course, that Watson’s namesake Thomas was the man who thought that the world market for computers might reach to maybe five. I guess that there’s nothing quite as humanly endearing as being rubbish at predicting the future.
The other place I see this anthropomorphism is in the rise (again) of chatbots. Putting a conversational front end on to a computer can, as Joseph Weizenbaum proved back in the mid 1960s with ELIZA, allow for an emotional connection between user and machine. However I’ve recently been looking for any evidence that chatbots as a User Interface are more effective than more traditional approaches, and so far have drawn a blank.
I’ve no doubt that the combination of Moore’s Law and improvements in machine learning are leading to greater effectiveness in the processing of vast amounts of data to do useful things. Every time I use Google Maps to navigate me to somewhere I don’t know I get even more convinced. But whether there is any more “intelligence” in this I’m really not so sure. And anthropomorphic layers that try to “humanize” computers are possibly playing a fairly cheap psychological trick on us to imbue intelligence where there is none.