Many claim that today’s AI can solve 90% of the problems put to it. That may be so, although I don’t have the empirical evidence needed to make such a claim. In any case, if the above were true, 10% of the problems put to AI would be solved “incorrectly” (according to the user’s criteria). I don’t want to dwell too much on the meaning of “error” or “success.” It is undoubtedly a matter of great philosophical depth, but it would seem, judging by the complaints, that this 10% of “errors” has to do with a failure to take into account the user’s values, beliefs, attitudes and/or goals.
If we turn to Piaget’s Theory of Cognitive Development (yes, I know this theory has been superseded in many respects, but its heuristic quality remains unbeatable), we could conclude that AI has reached the fourth and final stage of cognitive development, the stage of “formal operations.” Indeed, for Piaget, at around the age of 12 the human being begins to be able to detach from the concrete (what can be seen and touched) in order to direct their thinking toward the abstract. Concepts, symbols and ideas begin to be used, and complex problems can be solved whose statements are compressed by means of symbols (for example, the mind does not need to work with the concrete details of the concept “house” —walls, floor slabs, roof, fittings, etc.—; it is enough to compress all these attributes into a verbal symbol such as the word “house.” The same would happen with more complex mathematical, physical concepts, and so on). This capacity is refined in adulthood and, for many, is “the ceiling” of cognitive development. AI seems to be at the very peak of this capacity.
However, researchers such as William Perry realized that certain individuals reached a higher level of abstraction (one that Piaget did not describe) that allowed them to solve ambiguous, contradictory and context-dependent problems. Problems in which “two plus two does not equal four.” Perry called that superior capacity for thinking “postformal thought.” AI does not yet have this capacity, and that may be the reason why the user feels that 10% of the problems posed are not solved satisfactorily.
What is tremendously paradoxical is that reaching that higher stage (which not everyone reaches, as Perry warns us) requires going through years of academic training (formal thought) and of “intellectual tension” (real experience, over years, of using formal thought in ambiguous contexts). In short, it is a kind of “wisdom” gained through direct contact with problems, one that has demanded the loosening of formal knowledge.
When people ask me what will survive AI, I try to explain this idea. What will survive is what has more to do with “wisdom” and less with “knowledge.” What will survive is what has to do with judgment, prudence, the consideration of variables unrelated to formal logic (a great mistake many make is to assume that the only valid thinking is that which follows Aristotelian logic, and it is a fact that other forms of thought exist). Information and knowledge allow us to explore problems, but “history,” and social, motivational and attitudinal variables, and the goals we pursue, are what allow us to reach “solutions.”
AI has no biography (lived experience)… for now.