For new readers and those who request to be “好友 good friends” please read my 公告栏 first

The Three Waves of AI

I just finished listening to an MIT Faculty Lecture/Interview online of Professor Brian Cantwell Smith who is a philosopher as well as cognitive and computer scientist on the subject of deep learning and AI. Here are the main points I got out of his talk (The talk/interview will be available on YouTube which unfortunately is blocked in China. But you can search MIT Faculty Forum on line and perhaps see the lecture there).

Smith characterizes the first wave of AI from the invention of neural network to late 80s where we developed simple neural network and use human logic plus limited real world data to train the networks to do relatively well defined tasks. Second Wave of AI came during the last two to three decades during which Big Data became prevalent. Not only neural network became multi-layered requiring more sophisticated processing. But we begin to incorporate large amount secondary data and less human logic to train these complex networks. Prof. Smith used two metaphor to illustrate this point. First, when we look at a map of an ocean, we see islands and water. But when we look at the same ocean from the air above, we see not only dry land above water but submerge land structure under water at different depth. Second, if we sit in an office room with an entrance/exit door. We know without seeing that through the door there will be a corridor with many other office doors and offices. This is useful knowledge which AI during the second wave (i.e., at present) began to incorporate during learning. It is in this sense, “Deep Learning” (not just with multi-layered computation) takes place. This is a higher concept of learning. It is less on human logic but more on how the real world works. Smith use a two dimensional diagram with Technology as the x-axis and the Real World as the y-axis to illustrate his point. Nowadays the expertise of most scientists still populate a thin slice of an area closely near the x-y axes. The effort of the second wave of AI and research are to move into the middle of this diagram. Finally, he looks forward to the Third Wave of AI as the future where questions of judgement, ethics, and Man-Machine interactions will be addressed and collaborating scientists look at the big picture.