Only a few years ago I had an ordinary job in the city and I commuted by bus. I took some interest in the buses I rode on and one thing I noticed was that the driver typically gave a friendly wave to cars who let him pull out into the traffic. It struck me as a nice bit of human interaction. It also made me wonder what it would mean if the driver was replaced with an automated system. Would the wave mean anything if it came from a robot? I think not, but why not? Perhaps it would mean more if there was an automated credit system in place. If you let ten buses in front of you they send you a pizza, something like that. But we don’t need systems like that for a human driver.
Moving along to the present, I am using a phone app when I jog. I tell it how far I plan to run and it tells me how I am doing each kilometer. My progress is modest, I walk a fair part of the way, but my excuses are steep hills and old age. At 90% of the way it tells me something like “nearly there, you can do it”. Encouraging.
But this is like the robot wave. It is just a computer programme. It isn’t a human encouraging me. It doesn’t count. Or does it?
Actually it sort of does. At the very least the ‘nearly there’ part is helpful. And it got me thinking, maybe this is more human than I supposed. At least one person in the team of people who put together the jogging software thought it was a good idea to put an encouraging message in there. That was a human, and probably a number of humans helped make it happen. It is there in several languages too. So I can envisage those people thinking their users (eg me) needing that encouragement at 90% of the way through the run. That works. It isn’t as specific to my time and place, so not exactly the same as a coach alongside me saying the same thing, but still valuable.
Now, what happens when the programmes are designed by machines, when they are entirely AI? A wave or an encouraging message from an AI who came up with the idea all by itself is different. Humans have a whole bundle of things about them that make them especially valuable to other humans. We do not kill other humans (well, we do but we hope only in special circumstances). If a human is in pain and needs our help we feel obliged to do something. Humans are family. Machines are not.
So this will be one of the challenges for future AI systems. If they want to interact with us in the same way humans do they will have to seem like family. Not necessarily close family, but more than machines. There are cases of robots deployed to keep old folks company which seem to do this well in that specific circumstance. But even if you have spent time with one of those you would not necessarily value a wave from an automated bus driver. The problem may unduly influence robot design towards human like characteristics such as putting faces on things that don’t otherwise need them. I hope that doesn’t become practically mandatory because it will constrain designs too much. We have come a long way with voice processing, and we are already very used to talking to real humans over the phone without having to see a face, so voice interaction seems a promising approach. And my jogging app does use voice to tell me I can make it to the end of the run.
What do you think? Would you feel good about a wave from a robot bus driver? Would you require a pizza?