Defining Artificial Intelligence, Once and for All

Artificial intelligence (AI) has been around for some time and is sort of the technological hot-topic of the day. As it turns out, it may not be such an artificial issue… since no one can really define the much needed definition of it. From a philosopher’s perspective and the AI community, “intelligence” is an open sea made more treacherous with words like “mind”, “free will”, and “thought” that create a whirlwind of ambiguity.

Defining Artificial Intelligence, once and for all

The tendency to contrast artificial intelligence with human intelligence was, until the mid 2000s, a common thing in the AI community.  Seeking to settle “definition matters” once and for all, a Russian tech billionaire, Dmitry Volkov, convened a summit in 2014 (on a yacht), putting a number of leading philosophers to task over a true AI definition. Not surprising however, they failed to agree on any of them. One of the philosophers remarked that a consensus was unlikely to be reached even within the technological breakthroughs of the century.

The problem with defining AI

The problem and need for an official AI definition comes from a regulatory angle. The laws of war, advertising, licensing, and other areas where AI is applied will all need accurate governance defined.

Defining Artificial Intelligence, once and for all

The reason AI has been difficult to define, is the word “intelligence” is often dependent on the context it is being used in. Most traditionalists define it as a level of reasoning power, which is why games have been designed for man and machine to outthink each other.  But this fails to establish where man or machine begin and end.

Research is now directed in the effort to determine the boundary between man and machine. Creating the possibility of multiple definitions.  This piece is meant to define AI once and for all in the absence of an official definition, so let’s give it a try.

Defining artificial intelligence

To begin with, ‘artificial’ is relatively easy to define and may be taken to mean: not occurring in nature. It may also mean, not occurring in the same form as it does in nature, which quite frankly is confusing in itself.

Defining Artificial Intelligence, once and for all

Defining intelligence on the other hand is not an easy task either. Researchers Shane Legg and Marcus Hutter in “A Formal Definition of Intelligence for Artificial Systems” argue that intelligence includes:

  • Property of some entity or its agent that interacts with some form of natural environment
  • An entity’s ability to succeed at a particular task or achieve a certain goal given a set of criteria
  • Emphasis on learning, adaptation, and flexibility with a wide range of scenarios and environments considered “authentic intelligence”
Defining Artificial Intelligence, once and for all

Depending on who you ask, intelligence will always have a different meaning. In a bid to define artificial intelligence as a single term, I did my homework and agree with the following definition by TechEmergence that AI is:

“An entity or a collective set of them able to receive inputs from the environment, interpret and learn from such inputs and exhibit related and flexible behaviors and actions that help the entity achieve a particular goal or objective over a period of time.”

Legg and Marcus note that any definition of AI so far does not limit the inner workings of AI or the approaches used to create an artificial intelligence entity. I agree with them and so does the above definition. Approaches to AI that have been widely acclaimed include:

  • Machine learning
  • Multi-agent learning
  • Artificial neural networks
  • Reinforcement learning
  • Self-supervised learning

AI is a continuum

In the real world, artificial intelligence is a continuum. It is classified into three categories:

  • Assisted intelligence-AI has replaced standardized and repetitive human tasks.
  • Augmented intelligence-humans as well as machines learn from each other like in machine learning and regularly redefine boundaries and what they can do together.
  • Autonomous intelligence-systems that are adaptive over time learning on their own and eventually taking over whole processes.
Defining Artificial Intelligence, once and for all

How far we are willing to go from assisted to autonomous intelligence wholly depends on us and a number of factors like how much trust what we are putting into artificial intelligence.

Winding up

At the end of the day, AI is really a moving target no matter how hard we try nail it down. Any definition of artificial intelligence is subject to criticism from any quarters including you. The biggest worry however is not in the definition but in the fear that someday, we may not be able to keep up with AI’s processing powers; a product of our own creation.

Dan Mitchell President & Solution Architect at Rack Simply
Dan Mitchell

President & Solution Architect at Rack Simply

Rack Simply Linkedin
Share on linkedin
LinkedIn
Share on facebook
Facebook
Share on twitter
Twitter
Share on google
Google+