I'll Tell You About the Theologian Who Proved Why AI Will Never Surpass Man (Part 2)

In the previous part of this article, we reflected on the goals that Robert J. Spitzer in Why is Human Self-Consciousness Different from Artificial Intelligence and Animal Consciousness sets for demonstrating why AI will never truly be able to reach our “humanity”. We also began to introduce his demonstration, which starts with Gödel’s theorem but develops based on the fundamental philosophical position of the theologian Bernard Lonergan, which can be traced back to his work Insight: A Study of Human Understanding, 1957.

José Molina How long does it take to start a god 2018 grease pencil on paper 222 x 28 cm

Considering the enormous gap between artificial and human intelligence, it is essential to ask ourselves three crucial questions: (1) How does human intelligence manage to develop an understanding of the generalizations that support the intelligibility and internal coherence of arithmetic and mathematics? (2) How are human beings able to discern the truth (the correctness or validity) of these arithmetic and mathematical generalizations? (3) How can human beings understand the inviolability (the universal and necessary truth) of some arithmetic and mathematical generalizations, known as “axioms”? There are two contemporary theories that offer a possible explanation: (1) The quantum explanation proposed by Roger Penrose and (2) The theory of heuristic anticipation advanced by Plato, Polanyi, Whitehead, and Lonergan.

José Molina I doubt 2017 resin and acid treated wood cm 16×40

As early as 1981 (The Emperor’s New Mind) and 1994 (Shadows of the Mind), Penrose hypothesized that the processes of consciousness could be linked to quantum physics, particularly to phenomena that occur at the level of microstructures in the brain, such as microtubules. This idea was developed further with the physician Stuart Hameroff, and is called the Orchestrated Objective Reduction (Orch-OR) theory. According to this theory, consciousness emerges from quantum processes that occur within the microtubules in brain cells. Penrose believes that man has a non-computable consciousness, and is therefore not representable by a Turing machine. However, Penrose's research is complex and does not universally answer the question of how humans have non-computable truths. The second path (that of heuristic anticipation) reaches its peak with the theologian Lonergan. Like many philosophers before him (Aristotle, Augustine, Kant), he too divides what we perceive from what we idealize; however, Lonergan does not imagine theoretical structures, he goes straight to the point. What is not pure perception is, for him, processed to answer eight questions: what, where, when, why, how, who, how much, and how frequently. The result is no different: the answers to these eight questions place the perceived in a complete relational system with the rest of knowledge and generate new ideas, not necessarily associated with an image or perception; it is no longer necessary to resort to superstructures of concepts.

José Molina I imagine 2017 resin and acid treated wood 16 x 40 cm

The superstructures of Kant or Augustine are innate in man, just as the functional questions are for Lonergan. The question at this point is: could these questions be hardwired into the brain (and consequently also hardwired or programmable in an AI)? The answer is pivotal: it is reasonably no. The brain has a material existence of its own, which Lonergan considers an ‘empirical residue’. In contrast, the eight questions (categories) are not identifiable as empirical residue; Lonergan sees them as innate and transcendent. Certainly, if we stick only to definitions, the theologian's thought might seem recursive; however, it is already Plato who makes us note that without an anticipation of possible solutions, there would be no questions. To have awareness of something about which we do not yet know much, we must have a drive towards knowledge. The AI we know today does not ask questions; rather, it is a kind of oracle, predicting answers to questions that others pose. Can AI ever ask questions? Maybe. But can it ever have a recursive, insatiable, in some ways non-functional drive to know everything about everything? We have talked about Shannon's theorem and the weight of information; AI excels at this, but no information exists without a question, and no question arises unless there is a doubt that for man proceeds in a universal and totalizing way. Turing machines, and therefore AI, proceed for functional purposes. Could they be programmed with an insatiable hunger for knowledge—not of what is already known, mind you, but of what is still unknown? According to this line of thought, no, and Gödel is a good starting point for that demonstration. What is to be understood about this theorem is not so much its postulation on arithmetic or computation, but on the gnoseological implications (philosophy of knowledge): could a formal system ask itself if a formal system is complete? No, that's what Gödel proves. Lonergan goes further, taking the excess that Aristotle and Kant had reserved for man, in the world of ideas, and replacing it with the eight questions that link everything to everything else.

José Molina Hermanos 2019 resin and acid treated wood

The question of when AI will surpass man in its broadest and deepest dimension is at the center of the debate today. It is a question we have been exploring for months through popular science articles, drawing on philosophers and researchers. Spitzer’s research and the contribution of the theologian Bernard Lonergan seem very convincing in favor of human superiority.