A more formal implication of this question is whether intelligence requires a context.
This question may have little value to the fields of data science or statistics, however it is of central importance to the field of Artificial Intelligence. The aim of the AI field has been and will continue to be the simulation of human intelligence and possibly develop types of intelligence for which the human brain is not well equipped.
Such does not require understanding data set training requirements or postmodern thought. It requires knowing, in a more mathematically formal way, what intelligence is. The proclamation, "We know it when we see it," is not science and will not help develop the underdeveloped areas within the AI field.
Narrowness of Inquiry
When Norbert Wiener, Alonso Church, Claude Shannon, Alan Turing, Marvin Minsky, and others laid the foundations for Artificial Intelligence, they considered this question and others like it to be mathematical questions. Although they may have approached these questions with thought experiments like Turing's Immitation Game, they also developed those ideas mathematically, before they tried to embody their ideas in computers.
Not all these questions have a definitive answer in the literature, and the further investigation to reach them is of paramount importance to the further development of the Artificial Intelligence field.
The Turing Challenge to Cartesian Thought
Turing proposed at the end of the description of his famous test, "Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman? These questions replace our original, 'Can machines think?'"1
Turing effectively challenged the 1641 statement of René Descartes in his Discourse on Method and Meditations on First Philosophy:
"It never happens that [an automaton] arranges its speech in various ways, in order to reply appropriately to everything that may be said in its presence, as even the lowest type of man can do."
Descartes and Turing, when discussing automatons achieving human abilities, shared a single context through which they perceived intelligence. Those that have been either the actor or the administrator in actual Turing Tests understand the context: Dialog.
Intelligence Contexts Other Than Dialog2
The context of dialog is distinct from other contexts such as writing a text book, running a business, or raising children. If you apply the principle of comparing machine and human intelligence to automated vehicles (whether jet airliners, cars, trucks, drones, or trains) an entirely different context becomes immediately apparent.
Then the question becomes, does the distribution of fatalities, maiming, disfigurements, and property losses from automatic control match or do better than the same distributions of human control. We see not only the difference in context, but two other differences.
- The statistical comparison proposed by Turing is a single dimension. Either the computer is as indistinguishable from the human as the man is from the woman or not. In piloting or driving scenarios, the question of how to compare a maiming to a disfigurement arises. As in law and government, how much property loss is equal to a loss of one human life becomes part of the criteria.
- Validation of the automaton by inequality, where the control of the vehicle being distinguishably better than human control is still success. This in contrast to validation through rough equality, where the automaton's ability to keep up in a dialog with a human is renders it effectively indistinguishable from another human. (A dialog where the computer is too smart would make it distinguishable, thwarting the spoof.)
Range of Contextualization
We have two questions at the extremes in set theory.
Q1. What does intelligence then mean with NO context? Q2. What does intelligence then mean with all possible contexts?
These questions seem easy if taken one at a time.
A1. That's the same as any context. A2. That's what we've been calling general intelligence.
But are those two mathematically equal? Can we project that, if an automaton performs as well as or outperforms humans in a hundred contexts, it can surely do so with ten more contexts? Furthermore, do we select a low functioning, average functioning, or best functioning human for comparison?
What about contexts we don't have on earth yet but will have as time progresses?
Returning to Embodiment and Definition
Can a baby artificial mind grow into an adult by churning Internet data, or does it need to be placed in the context of a robotic entity so it can move around and experience interaction with the physical world?
Can a brain be intelligent without a body?
More generally and more formally ...
Does intelligence require a context?
 Chapter 1 ("Imitation Game") of Computing Machinery and Intelligence, 1951.