In Part 1 I explored the definition of intelligence, and suggested that acceptable definition of intelligence is subject to interpretation; and the origin of intelligent behavior can sometime be wrongly attributed. In this post, I want to explore concept of artificial intelligence (AI), which is experiencing a renaissance of sort lately.

Today when one talks about AI, the conversation quickly turned into a discussion of Deep learning (DL) which, in my opinion, is largely a rebranding of neural network.  Having been a practitioner of both neural networks and expert systems in the early 90’s I have come to appreciate the strengths and weaknesses of different AI approaches, but despite of the progress made since then, research of system-level intelligence continue to be largely neglected in favor of simpler mapping problems.

Let me offer an analogy to illustrate what I meant:   If human brains are like laptop computers, then neurons are like transistors, and neural networks are like circuits making up the CPU.    Showing transistors working together to realize a CPU instructions is inadequate at describing the complex behaviors of an applications-running laptop.  As most programmers know, there are system-level concepts involved in creating complex behaviors of applications–concepts like program loops, function calls, multitasking, object-oriented design, just to name a few, and we turn to computer science for these system-level functionality.   Similarly, for AI, neural networks only implement circuit-level functionality; for system-level functionality we need to turn to psychology, neural science and brain studies.

In Part 3, I will share some knowledge gleaned from my review of relevant knowledge from psychology, neural science and brain studies, and how they may be applied to create intelligent systems.

(The above article is solely the expressed opinion of the author and does not necessarily reflect the position of his current and past associations)