Summary
This book discusses different aspects about building a developmental robot. The author discusses all the concepts in a very easy to understand manner. In chapter 11 and 12 it discusses about the groundwork and developmental approach for growing a robot.
Ground work: There were some remarkable robots in early 1950s but they were single task specific. But single task specific robots cannot help with human like intelligence. Robots engineered for tasks will not be humanlike. The physical anatomy influences the thought and cognitive development. So, they should be humanlike. A developmental approach should not adopt fixed formalisms (such as symbolic or neural methods); rather, it should concentrate on the more general mechanisms.
The developmental approach: Developmental robotics takes inspiration from the remarkable growth of sensorimotor skill and cognitive ability seen in child development. He gave some developmental timeline for different stages of human/robot agent. And humanoid anatomy is important for at least two reasons: It has an influence on the shape of our experiences and mental models; and similarity with humans engenders similar world understanding, and therefore better social empathy. Finally, he argues that general, open-ended, goal-free intelligence cannot be designed as AI processes; rather, they should be approached through synthe- sis experiments using mechanisms such as intrinsic motivation and play generators. The key concepts are synthesis, constructivism, and enaction.
My thoughts: We can use many concepts from these chapter for our SEDRo. We can use the developmental milestones for evaluation and also removing constraints. It also provides some reasons for why we should focus on behaviour not on tasks for intelligence development. And also, why should we use a humanoid approach to develop humanlike intelligence.
