I absolutely despise the voice recognition phone systems employed by the likes of Fedex or Microsoft. Every time I have to call them, I feel like an idiot talking to the pre-recorded voice on the other end. Note, I have no problem with the kinds of answering systems that make you push buttons. It’s the talking that irritates me. I really prefer to hear “push 1 to place an order” rather than “please say ‘place an order'”.
This made me think a bit. What if we would have a semi-functional AI answering phones, or doing similar menial tasks. Would I feel less weird talking to something that possesses at least some degree of apparent intelligence? Would I be able to have an actual conversation with an android?
I was watching a video of Asimo walking around and dancing the other day, and he seemed kinda cute. I’m sure I would have fun ordering him to walk here or there, or trying to talk to him to see how he reacts. But that is a little different. Asimo is a cutsey robot designed to look more like a toy or a human shaped pet. You know he is a dumb piece of hardware that has limited responses, and so it is easy to anthropomorphize him.
But if you end up talking to something that tries to closely imitate a human being, things are more complicated. I think nothing of chatting with Asimo, but the nice female voice on the other line that requires me to speak voice commands irritates me. I think this might be an indication that a voice activated phone system like that already falls into the uncanny valley. Talking to a self aware AI or an android may be much worse…
Are technically knowledgeable people more sensitive to the uncanny valley effect? For some of the clueless people out there, technology is indistinguishable from magic. So perhaps they are more willing to converse with that little voice system on the phone?
On the other hand, if we develop good algorithms for parsing spoken language this problem might disappear. If the AI can intelligently respond to you, ask follow-up questions and give coherent responses to off-topic questions then perhaps I would not feel that dumb talking to it. Would the machine need to be self aware for that? Or perhaps it would be enough if it could fake self-awareness really well? I don’t know…
I know one thing. If we are ever able to develop an AI that could parse spoken english on the fly, process it, analyze it and compose coherent responses, it would most likely find great use in telemarketing and phone answering systems.
Think about it. We could replace the Indian tech support centers with bunch of AI boxes that would go through basic troubleshooting steps with the users. Because that’s what most of the call centers in India do – they listen to your complaint, identify important keyboards, then do a quick search in the troubleshooting manual and then read from it.
That will be done by AI one day, I’m sure of it.
