What Is Artificial Intelligence?
What is artificial intelligence?
Intelligence is rather difficult to define, which makes defining artificial intelligence even more difficult. While there are various definitions of artificial intelligence in use, there are essentially four main definitions that you need to know.
Each of these four definitions has various pros and cons. Some are vague and ambiguous while others may be biased human-centric ways. However, it’s important to understand all four, since discussions involving AI will likely be using one of these four definitions. In many cases, all of the parties involved in these discussions may not be using the same definition.
Human Intelligence Approach
The first definition we encounter is one that defines artificial intelligence relative to human intelligence. According to this definition, artificial intelligence is the ability of a machine to replicate human intelligence. This would include tasks like visual pattern recognition, problem-solving, and walking on two legs.
However, this definition is biased in a human-centric way. It does not consider other forms of natural intelligence as types of artificial intelligence. In addition, it does not define what intelligence is in the first place, so it’s a rather vague definition.
Natural Intelligence Approach
The second definition we encounter defines artificial intelligence relative to all measures of natural intelligence. According to this definition, artificial intelligence is the ability of a machine to replicate any natural intelligence. In addition to human intelligence, this would include animal intelligence (e.g. a mouse navigating a maze) and swarm intelligence (e.g. an army of ants protecting their colony).
This definition eliminates the human-centric bias by expanding the definition to include all forms of natural intelligence. However, it still suffers from being rather vague as it does not actually define intelligence itself. Rather, it simply defers this problem to the definition of natural intelligence — which itself is not very clear.
Rational Agent Approach
The third definition we encounter defines artificial intelligence as any machine that acts as a rational agent. A rational agent is anything that perceives its environment and chooses actions that maximize the expected likelihood of achieving a goal. Being rational simply means: always choosing the best action given your goal. Thus, artificial intelligence is any man-made agent that acts in a rational way.
This definition is much more precise and well-defined than the two previous definitions. In addition, it avoids being biased in a human-centric way. However, there are still many concepts of intelligence that it avoids addressing directly like learning, knowledge, reasoning, and planning. Despite this shortcoming, this definition is the most common definition used by AI practitioners and academics.
Moving Target Approach
Finally, the general public tends to have a quite different definition of artificial intelligence than AI practitioners. Their definition is constantly changing as technology is improving. Once some cutting-edge AI technology becomes mainstream, the general public typically stops referring to it as “artificial intelligence” and simply calls it “plain-old technology”.
In short, when most people say “artificial intelligence” they typically mean “anything that a human can do but a machine cannot yet do”. As a result, the general public’s definition of AI is both vague and human-centric. In addition, it’s a moving target that changes from year to year. Despite these limitations, this is still the most common definition you will encounter when discussing AI with the general public.
Understanding the various definitions of AI will help you to better understand artificial intelligence and separate hype from reality. In addition, it will allow you to communicate more effectively with others who may be using a different definition of AI than you are.
To learn more about AI, please check out my latest course — Artificial Intelligence: The Big Picture.
Originally published at https://matthewrenze.com on February 1, 2020.