Intelligence is any goal-based behavior informed by evidence.
The goal is what we give as observers, and we judge how well the behavior satisfies the goal.
For example, a fire’s flame dances without any purpose or goal. We don’t claim that it’s avoiding objects or dancing for our pleasure.
To be informed means that behavior satisfying the goal is not random and is not coincidental, and bares a direct relationship to the evidence.
An example of random behavior would be a fish swimming by moving its tail left or right in a random sequence such as left, left, left, right, left, left, right, left. This may or may not swim in a straight line sometimes, and has no relationship to evidence.
Coincidental behavior would be a predetermined repeating sequence of steps, but without baring a relationship to the evidence. An example would be playing four-in-a-row by dropping in pieces left to right in a repeated sequence until the game is over. It may or may not win sometimes, against a human.
The creation of intelligence requires an experiment to obtain the relationship between goal, behavior and evidence. Improvement of intelligence over time is learning. The capture and storage of the evidence affects the frequency of experiments and the process of learning.
For example, DNA is one way of storing evidence. A biological life form is born with a mutation to its DNA introducing a new behavior, this coincidental behavior is successful in some way and that behavior is passed on to its offspring. In computing terms this is “static” memory storage. This introduces some intelligence to that breed of the species. Each generation is an experiment, the lineage as a whole is learning.
A complex nervous system, such as a brain, or “dynamic” memory store, allows for behaviors to be parametrized by some data.
A Saharan desert ant calculates distance (called “path integration”) to their goal and then returns home the same number of steps. If one adds stilts or shortens their legs, they overshoot or undershoot the distance.