My primary interest in robotics is modeling and exploring human intelligence. As I started on my study of Artificial Intelligence in the early 90’s, I decided that a box on a table couldn’t be intelligent because it is disconnected from physical reality. One can define a tree as a particular type of plant, or list its qualities such as shade production and carbon filtering, or one can list its uses such as lumbar and edible nuts. But all these qualities require their own further definition.\u00a0 At the end of the day, this is just symbol manipulation and is ungrounded from meaning.\u00a0 Early attempts at AI tried to build meaning from complex semantic dictionaries and failed to produce intelligence. Those approaches are used today for automated parsing of natural text, but the only “intelligence” they show is in the logic implicit in the complex set of relationships built into the symbol definitions.\u00a0 To bring shared meaning to the symbols we use, it helps to physically connect them to the shared reality we experience.\u00a0 If you want to know what a tree is, you have to be embodied in the world — you have to be able to hug it.<\/p>\n
Hugging trees might sound like a silly basis for intelligence, but it stems from the basic cycle of autonomy: perception — thought — action.\u00a0 The classic approach to AI that focuses on symbolic manipulation fails because it only expresses “thought.”\u00a0 Connecting to the external shared reality through perception and action gives thought meaning.<\/p>\n
We can define living beings by our ability to act — to create change in the world or ourselves. Stones do not act — they react — they are passive.\u00a0 In contrast, all living things perform actions in some way.\u00a0 We can easily see these actions in the higher mammals, but even the simplest bacteria or plant creates action.\u00a0 Plants grow and change, transport fluids, mature and die.\u00a0 These are all actions. Plants’ capabilities of perception and thought are more limited than ours — but not non-existent.\u00a0 Consider the Venus Fly-Trap for perception and trivial decision making in a plant.<\/p>\n
Just like thought<\/em>\u00a0by itself does not get one far in life, action<\/em> on its own without perception or thought is not very functional.\u00a0 Early industrial machines are a great example of action without perception — they do one thing very well and will keep doing it even to the point of destruction if the environment changes.\u00a0 Charlie Chaplin perfectly captured the essence of action without perception or thought in his movie “Modern Times,” where he depicts an automated Feeding Machine:<\/p>\n