As far as Artificial Intelligence is concerned, what is "intelligence"? The definition I see on various sites like Wikipedia:

Intelligence has been defined in many different ways including as one's capacity for logic, understanding, self-awareness, learning, emotional knowledge, planning, creativity, and problem solving

Merriam Webster:

  1. The ability to learn or understand or to deal with new or trying situations : reason; also : the skilled use of reason.
  2. The ability to apply knowledge to manipulate one's environment or to think abstractly as measured by objective criteria (such as tests).

etc seem to be a bit broad and nebulous, and not necessarily what I would be thinking of if I wanted to build an AI, or evaluate the intelligence of non human life-forms.

The definition I currently go with is:

General problem solving ability.

However, I'm not sure if this is broad enough to encompass all we think of when we say "intelligence" in the context of AI, or what we would be looking for in "intelligent" life-forms. What's a useful definition of intelligence. Broad enough to encompass all the we consider when we think intelligence, yet narrow enough to exclude particular idiosyncrasies of specific intelligent agents? A universal definition of intelligence applicable to all intelligent agents.

  • $\begingroup$ I suppose there are many different interpretations of 'intelligence', as you already hint at yourself in your question. What else do you hope to get out of this question than a list of many different interpretations/theories of intelligence? A "universal definition of intelligence applicable to all intelligent agents" will be necessarily broad, no? $\endgroup$
    – Steven Jeuris
    Commented Jul 20, 2017 at 9:41
  • $\begingroup$ Broad yes, "but not including unnecessary things like emotions and such. A universal definition of intelligence applicable to all intelligent agents" is what I seek. Maybe broad is not what I mean. $\endgroup$ Commented Jul 20, 2017 at 12:15
  • $\begingroup$ What is "unnecessary" can also be a major discussion, making the question rather opinion based. Why are you interested in knowing? If you are trying to build an AI- or multi-agent system you can also combine different definitions to formulate your own definition. What makes your agents intelligent is their your own interpretation. $\endgroup$ Commented Jul 20, 2017 at 15:43
  • $\begingroup$ I feel that without knowing what intelligence means, it is ill advised for me to even think of creating intelligence. I believe it is imperative for me to define intelligence, before I seek to explore it. I want to develop an axiomatic system for intelligence as my life work, so this question is very important to me. (I plan to throw no less than 2 decades into AI research), after which I'll most likely give up if I fail. I can not afford to start off on a wrong foot. $\endgroup$ Commented Jul 22, 2017 at 15:58

2 Answers 2


This is a very broad and not well defined issue, but two theories that may help you to get a better understanding are the following :

Triarchic theory of intelligence by Robert J. Sternberg

Sternberg's definition of human intelligence is "(a) mental activity directed toward purposive adaptation to, selection and shaping of, real-world environments relevant to one's life" (Sternberg, 1985, p. 45).

Sternberg divides intelligence into three major subtheories:

Componential – analytical subtheory

Analytical giftedness is influential in being able to take apart problems and being able to see solutions not often seen. Unfortunately, individuals with only this type are not as adept at creating unique ideas of their own.

Experiential – creative subtheory

Sternberg splits the role of experience into two parts: novelty and automation. A novel situation is one that you have never experienced before. People that are adept at managing a novel situation can take the task and find new ways of solving it that the majority of people would not notice (Sternberg, 1997). A process that has been automated has been performed multiple times and can now be done with little or no extra thought. Once a process is automatized, it can be run in parallel with the same or other processes.

Practical – contextual subtheory

Sternberg's third subtheory of intelligence, called practical or contextual, "deals with the mental activity involved in attaining fit to context" (Sternberg, 1985, p. 45). Through the three processes of adaptation, shaping, and selection, individuals create an ideal fit between themselves and their environment. This type of intelligence is often referred to as "street smarts." Adaptation occurs when one makes a change within oneself in order to better adjust to one's surroundings (Sternberg, 1985). For example, when the weather changes and temperatures drop, people adapt by wearing extra layers of clothing to remain warm.

Theory of multiple intelligences by Howard Gardner

The theory of multiple intelligences differentiates intelligence into specific 'modalities', rather than seeing intelligence as dominated by a single general ability. Howard Gardner proposed this model in his 1983 book Frames of Mind: The Theory of Multiple Intelligences. According to Gardner, an intelligence must fulfill eight criteria: musical-rhythmic, visual-spatial, verbal-linguistic, logical-mathematical, bodily-kinesthetic, interpersonal, intrapersonal, and naturalistic. He later suggested that existential and moral intelligence may also be worthy of inclusion.


In the domain of Artificial Intelligence, especially when considering the plausibility of "Superintelligence", the definition is of "intelligence" is under active debate. For an overview of some of these arguments, see the provocatively named "Superintelligence is a free lunch, and there are no free lunches" by Erik Hoel.

But clearly rating a superintelligence using an IQ test is useless. We wouldn’t use Einstein as a metric for a real superintelligence of the variety Bostrom is worried about any more than we’d use the mouse as a metric for Einstein. So clearly the danger isn’t just from an AI with an exceptionally high IQ (since it’s not like high IQ people run the world anyways). Rather, the danger comes from the possibility of a runaway process of a learning algorithm that creates a god-like AI. To examine that we need a more abstract and highly generalizable notion of intelligence.

A definition of intelligence is actually given by Legg and Hutter in their paper “Universal intelligence: a definition of machine intelligence.” Taking their point broadly, the intelligence of an agent is the sum of the performance of that agent on all possible problems, weighted by the simplicity of those problems (simple problems are worth more). A superintelligent entity would then be something that scores extremely high on this scale.

So universal intelligence is at least somewhat describable. Interestingly, Einstein scores pretty low on this metric. In fact, every human would score pretty low on this metric. This is because the space of all problems include things that human beings are really bad at, like picking out the same two color pixels on a TV screen (and three pixels triads, and so on).

We can also define intelligence broadly as: given a particular goal, the agent who needs to achieve that goal learns everything that’s necessary to achieve that goal with a high probability (which might mean achieving other goals en route to the main goal, etc).


Not the answer you're looking for? Browse other questions tagged or ask your own question.