OBSERVABILITY OF THE INTELLECT
If one remains in the realm of science and does not move into the realm of faith, then the intellect should be detectable in observing its potential owner. The subject of observation, of course, is the behavior of the corresponding object.
From the beginning of the "artificial intelligence" epic, it was implied that one should observe how a potential intelligence owner reacts to external impacts. If it reacts reasonably, then the intelligence is there. If it responds indistinguishably from the human, then the level of intelligence is a human-like one.
However, the results of this approach in practice are somewhat unexpected. For example, if an automat for controlling automatic store doors is tested using the approach of a person, animal, or robot as a test action, the response looks entirely rational. Does this mean that automatic doors or a washbasin faucet that turns on when you raise your hands are the possessors of intellect? Most likely, no. It is also difficult to recognize the intelligence of a golf ball or pendulum that reacts quite "reasonably" to a blow/push or a doorbell that responds reasonably to pressing a button.
As you can see, the approach based on evaluating the reaction to the impact essentially leads to the wilds of scholasticism and subjectivism: it is required to determine which reaction should be considered "reasonably intelligent" and which should not. And since the diversity of both impacts and reactions is not limited by anything, a universal definition of an "intelligent reaction" is unattainable. The result is a totally subjective principle of individual assessment: for someone, an automatic door is an intelligent system, and for someone, ChatGPT is not an intelligent one. This is inconsistent with the scientific approach, which assumes the possibility of obtaining an answer by measurements/observations.
This means that the principle of conformity of the observed reaction of the object to the impact with the observer's expectations is not a scientific approach: they depend on the observer's expectations, which are obviously subjective in nature. This is true for both the automatic door and ChatGPT.
However, there is another aspect of the behavior of the systems under test that does not require subjective assessments or assumptions: observation of the system's behavior IN THE ABSENCE OF EXTERNAL INFLUENCES. Here a difference is immediately revealed; on one side is an automatic door and a pendulum, and on the other is a man and an ant. In one case, the absence of external influences means a complete lack of activity; in the second case, action may occur without external impacts. Intelligence is the ability of a system to act not only reactively in response to external impacts but also to act actively (demonstrate self-influence). A mouse in a maze does not require a command to take up the search for an exit.
For a more correct definition, it is necessary to include the passage of time in the number of external "impacts" to exclude devices and organisms that perform cyclic actions regardless of external impacts. The immutability of the cycle is, in this case, the dynamic equivalent of a stationary state.
Intelligence is a decision-making tool. Decision-making requires intention: the same decision can be good or bad, depending on the intention. Intentions in a non-intelligent system are set (imposed) explicitly or implicitly by external influences (ChatGPT), or they are immutable and "built into" its design, providing a predetermined reflective response to impacts.
In an intelligent system, intentions are generated based on the analysis of information collected by the system itself, which can be observed as some kind of non-reactive activity or be wholly hidden from the observer; in any case, observation reveals some action not initiated by external influences.
Of course, the presence of intelligence does not mean the sameness/equivalence of the intelligence of various systems. The presented intellect can be characterized by many parameters, depending on the environment in which it functions and for what purposes it was created (or how it arose in the natural environment). Accordingly, when comparing two intellects, A and B, it may turn out that in one (quite measurable) parameter, A surpasses B, and in others - precisely the opposite. That is, a single quantitative parameter cannot express the "power" of the intellect.
As always, "border options" are possible - for example, adaptive systems that react reflectively but can adjust the parameters based on the accumulated data (the criteria for adaptation are part of the design).
Information systems can be very complex, store vast amounts of information/knowledge, and be extremely useful - like the Internet with search engines or automated manufacturing systems - without intelligence.
The presence of many self-proclaimed "AI-based" systems that do not meet the above criterion is natural and understandable - just as the term "smart" is applied to the mass of products from a phone to coffee makers; however, when it comes to professional and scientific fields, it makes sense to use the term "information system" if there are no verifiable grounds for using the term "intelligence."