PIE # 85 - Artificial Inteligence: Two Vastly Different Approaches

(01.02.2014) Teodor Przymusińskii

The science and engineering field of Artificial Intelligence (AI) aims to create intelligent machines. Accordingly, it is involved with “machine intelligence” as opposed to “human intelligence”. By historical standards, it is a very young and thus relatively immature area of research. It was founded in 1956 at a Dartmouth College conference by John McCarthy, Marvin Minsky, Allen Newell and Herbert Simon, who quickly became the leaders of a rapidly developing and promising field. Their initial expectations were so optimistic that from today’s perspective they appear very naïve: Simon, a future Nobel Prize winner, predicted that within twenty years machines will be capable of doing any work a man can do, while Minsky wrote that within a generation the problem of creating AI will be essentially solved.

By the seventies, however, this early optimism turned into disillusionment. Twenty years had passed and AI had not even come close to accomplishing any of these lofty goals. Fortunately, by the middle of the eighties, due to the success of expert systems, AI recovered and proceeded to rapidly develop, producing more and more sophisticated applications and machines. From IBM’s Deep Blue chess master, through speech recognition systems like Apple’s Siri, to autonomous Google Cars and even house robots like Roomba, we are literally surrounded by products originating from the field of AI. However, we are still nowhere close to reaching the goal that Herbert Simon almost sixty years ago expected to be solved within twenty years.

Approaches to AI can be divided into two broad categories: those based on formal foundations and those relying on mimicking natural phenomena. The former rely on the well understood and highly sophisticated machinery of mathematics and mathematical logic and thus are based on a firm scientific foundation. The latter attempt to make machines behave in ways resembling human behavior or the behavior of the nature. Their scientific foundations are usually weak and they tend to be somewhat ad hoc (which should by no means suggest that they are not successful). In my presentation, I will describe – necessarily at a very basic level – one sample approach from each category and give you some idea of how they work and what they can accomplish. The profound differences between these approaches should serve as a vivid illustration of the breadth of the scope and variety of the proposed solutions.