2016 October 30
A long time ago Asimov predicted that humanoid robots would be used to automate all the boring/difficult jobs we don’t want to do. He argued that it would cost too much money to replace every piece of human operated equipment with an automated variant, it would be much more economical to build one smart robot that could operate multiple existing machines.
Clearly this isn’t how things are playing out. We’re building self-driving cars, self-landing rockets, automated stop lights, etc, and there’s a serious lack of humanoid robots walking around. Why? A few reasons: 1) Asimov himself lamented the difficulty in building a decent bipedal robot. The organic versatility is very hard to reproduce mechanically, let alone control with any amount of grace. 2) General purpose AI has proven very difficult to design, so instead we’re doing purpose built AI. E.g. driving is complicated, but it’s a bounded problem where you can hardcode many of the rules. The same AI does not also need to learn how to dance or sing. 3) Asimov assumed a short transition period, that one day humans would be doing the task and the next day robots would. Instead, these transitions are taking place slowly by incrementally improving the existing machines in the next model. Stop lights didn’t replace traffic cops overnight. The first traffic lights were manual tools traffic cops operated, and they took years to install around the world. Slowly they have become more automated and more intelligent. And cops still have to direct traffic sometimes.
A humanoid robot assistant would be nice in the kitchen, but the dishwasher’s much cheaper. Next I need one that can cook better than the microwave.