autotelic automatons
You find the strangest stuff when googling around. I needed a definition of “autotelic,” which comes up all the time in the Kant that I’m reading now. And I found this, a paper from the Sony Computer Research Lab. Here’s the abstract:
The dominant motivational paradigm in embodied AI so far is based on the classical behaviorist approach of reward and punishment. The paper introduces a new principle based on ’flow theory’. This new, ‘autotelic’, principle proposes that agents can become self-motivated if their target is to balance challenges and skills. The paper presents an operational version of this principle and argues that it enables a developing robot to self-regulate his development.
I haven’t read the paper yet, and who knows if I ever will, but it sounds like an application of modern “human resource” management techniques to inanimate, yet incipiently thinking, things. Why offer a reward when the work is a reward in and of itself? Why “manage” when they can be taught - can be expected - to manage themselves?
