PHILOSOPHICAL EGGS

Introduction

Plato (427-347 BC), in his dialogue Timaeus, has his character Timaeus offer Socrates, Hemocrates, and Critias a “likely story” about how our cosmos, our orderly universe, came to be from a preexisting state of chaos. Near the beginning of the story we encounter this passage:

“Let me tell you then why the creator made this world of generation. He was good, and the good can never have any jealousy of anything. And being free from jealousy, he desired that all things should be as like himself as they could be. This is in the truest sense the origin of creation and of the world, as we shall do well in believing on the testimony of wise men: God desired that all things should be good and nothing bad, so far as this was attainable. Wherefore also finding the whole visible sphere not at rest, but moving in an irregular and disorderly fashion, out of disorder he brought order, considering that this was in every way better than the other.”

God, being all good, had no envy and so made the cosmos to be as good as possible. To achieve this goodness he created a world soul with intelligence to guide the world’s material body in the measured ways we experience. The creation of soul was necessary since intelligence can’t exist in anything which is devoid of soul:

“Now the deeds of the best could never be or have been other than the fairest; and the creator, reflecting on the things which are by nature visible, found that no unintelligent creature taken as a whole was fairer than the intelligent taken as a whole; and that intelligence could not be present in anything which was devoid of soul. For which reason, when he was framing the universe, he put intelligence in soul, and soul in body, that he might be the creator of a work which was by nature fairest and best. Wherefore, using the language of probability, we may say that the world became a living creature truly endowed with soul and intelligence by the providence of God.”

I’ve been fascinated by these passages since I read them when I was 18 back in 1988. But recently I reread them and immediately, and somewhat surprisingly, had some thoughts about artificial intelligence in light of the claim that intelligence requires soul. In this post I want to share those thoughts which offer an interesting and certainly controversial way to question the possibility of certain forms of AI.

AI

Let’s start with this helpful overview of AI from the Notre Dame Learning website:

“Artificial intelligence (AI), in its simplest definition, refers to any technology or machine that can perform complex tasks typically associated with human intelligence. These tasks can include problem-solving, planning, reasoning, and decision-making. As the field of AI continues to expand, the terminology and definitions used to describe it also continue to evolve. The three main categories of AI presented in this overview are artificial narrow intelligence, artificial general intelligence, and generative artificial intelligence.

Artificial Narrow Intelligence (ANI), also called weak AI or narrow AI, refers to systems that are designed to perform specific, often complex, tasks such as analyzing large data sets, making predictions, identifying patterns, or generating text and images. All existing AI systems in use today, including generative AI models, fall under this category. These systems operate within defined limits and do not possess human-like understanding, reasoning, or adaptability across multiple domains.

Artificial General Intelligence (AGI) refers to the hypothetical ability of a machine to demonstrate broad, human-level intelligence. This includes the capacity to learn new tasks independently, apply knowledge across a range of areas, and adapt to unfamiliar situations. AGI does not currently exist and remains a goal of ongoing research.

Generative Artificial Intelligence (GenAI) refers to a subset of artificial narrow intelligence that uses algorithms to create or generate new, realistic content such as text, images, audio, and video based on patterns found in training data. Examples of GenAI systems include ChatGPT, Claude, Gemini, and Midjourney. Although these tools can produce original-seeming results, they operate within the boundaries of their training and do not possess general understanding or self-awareness.

GenAI systems have become a topic of significant discussion because they are challenging long-standing practices in education, especially in the assessment of learning and knowledge creation. The rapid growth of these tools has also raised questions about how they may transform professional roles, such as those in journalism, computer programming, and creative industries.”

A Platonic Argument Against Certain Forms of AI

Notice that ANI and GenAI, despite their power to perform various tasks far more efficiently than humans, lack certain human capabilities. And AGI (also known as strong AI, full AI, human-level AI, human-level intelligent AI, or general intelligent action) remains hypothetical. For some, these limitations are temporary. For example, the late AI philosopher John Pollock claimed “Once [my intelligent system] OSCAR is fully functional, the argument from analogy will lead us inexorably to attribute thoughts and feelings to OSCAR with precisely the same credentials with which we attribute them to human beings. Philosophical arguments to the contrary will be passé” (see Stanford Encyclopedia of Philosophy “Artificial Intelligence” section 8.2). Others disagree such as Selmer Bringsjord (former chair of the Department of Cognitive Science at Rensselaer Polytechnic Institute) who believes “the human mind will forever be superior to such machines.” I currently lean toward towards the latter position for a variety of reasons. But my Timaeus-inspired argument for doing so goes like this:

Premise 1: Only soul can have intelligence.

Premise 2: Machines which are said to have artificial intelligence have no soul.

Conclusion: Therefore, AI machines are, despite their name, actually not intelligent.

Why accept premise 1? We can approach an answer to this question with the help of Donald Zeyl who offers a helpful summary of Timaeus’ account of God’s creation of both the world soul and the souls of individuals:

“The composition of the world’s soul out of a harmonically proportionate series of portions of a mixture of both divisible and indivisible Sameness, Difference and Being, and the division of these portions into two intersecting circles (called the circle of the Same and of the Different) explain the cognitive powers of the soul in relation to the different types of objects of cognition: those that are and those that become. When joined with the world’s body, they also explain the cosmological organization of the universe.” (see the Stanford Encyclopedia of Philosophy’s entry “Plato’s Timaeus” section 1).

Here we see that both the world soul and the human soul have a central capacity to make judgments of things being the same or different since Timaeus says we must “distinguish between that which always is and never becomes and which is apprehended by reason and reflection, and that which always becomes and never is and is conceived by opinion with the help of sense.” The things which are the same, those which have Being, are what Plato calls the Forms: the immaterial, eternal, and prefect paradigms which God used to guide the creation of the cosmos and which we can access to some extent with our reason. The things which are different, those which Become, are the many physical things we experience through our senses. James Royce, in his book Man and His Nature (McGraw-Hill, 1961), helps us see how these insights about judgment and immaterial objects relate to premise 1:

“Matter excludes other matter from occupying the same space; these acts [of making judgments] show a compenetration, a disregard for the limitations imposed by quantity, which can only be due to the fact that they are not spatial. My concept of a triangle cannot be measured. I can have larger and smaller images, but the idea of what a triangle is applies equally well to all triangles, large and small, which could not be possible if the idea had size itself. My idea of an elephant is no bigger than my idea of a flea, for neither is quantified. Ideas do not occupy space: not having parts, they cannot extend over quantified matter. Nor does a simple idea occupy many parts of the brain at once, for then we would have many ideas, not one, of any one thing. A judgment means recognition of identity or non-identity of two concepts; but if one concept is in one space, and the other in another part, I could never get the two together in a judgment. The only conclusion is that the ultimate subject of such simple operations is itself simple.” (313).

According to Royce, judgments that something is different or the same (for example, ‘a triangle is not a square’) and the immaterial ideas that go into them (for example, the universal ideas of triangularity and squareness which, as universal, cannot be located in a particular place) cannot be accounted for if we are just complex bodies in space. But if judgments take place in a simple and non-spatial soul then perhaps we can make sense of them. And, since intelligence appears to require judgment, we begin to see why intelligence might require soul or an immaterial and simple substance capable of, among other things, making judgments of the same and different which employ immaterial ideas akin in some ways to Plato’s Forms.

And what about the relations between the propositions that must exist if we are to construct arguments and make inferences? When we reason one mental state seems to cause another mental state by virtue of its propositional content or meaning. We grasp the meaning of the proposition ‘All humans are mortal’, grasp the meaning of ‘Socrates is a human’, and then, based on our mental apprehension of this propositional content, make a judgment and infer the conclusion ‘Socrates is mortal’. This judgment is a form of rational causation which depends on the mind and its apprehension of propositions and propositional content. But physical causation is only a matter of energy interacting in accordance with the basic forces and laws of nature. It has nothing to do with propositions and their content at all. Indeed it would seem, as Edward Feser nicely put it, “that the electrochemical properties of the neural processes with which the thoughts are associated are entirely sufficient to bring about whatever effects they do bring about. The meaning or content of the thoughts is irrelevant” (see his Philosophy of Mind p. 152). Tyler Burge elaborates:

“[R]eason is a constitutive structural feature of causation by propositional psychological states and events. According to the natural sciences, reason is not a structural feature of material composites. The causation by material parts of material composites, operating in their physical relations to one another, must suffice to alone compose causation by material composites. It is hard to see how the causal powers and causal structure of material components could alone compose the causal powers and causal structure of causal transactions that hinge on the rational, propositional structures of propositional states and events. So it appears that rational, propositional, psychological causation is not the causation of a material composite.”

And,

“The second concern about compositional materialism is similar, but does not feature causation. Here it is: the physical structure of material composites consists in physical bonds among the parts. According to modern natural science, there is no place in the physical structure of material composites for rational, propositional bonds. The structure of propositional psychological states and events constitutively includes propositional, rational structure. So propositional states and events are not material composites.” (see his article “Modest Dualism” in The Waning of Materialism, eds. Koons and Bealer).

If these insights are convincing then we can conclude that our souls, insofar as they think ideas, make judgments, and form arguments, must be simple, immaterial, and irreducible to the complex material structures of the brain. Edward Feser, in his book Five Proofs of the Existence of God (Ignatius, 2017), points out that these three interrelated soul capacities traditionally constitute intelligence in the philosophical tradition:

“Intelligence, as traditionally understood, involves three basic capacities. First, there is the capacity to grasp abstract concepts, such as the concept man, which is what you have when you not only know this or that particular man or this or that particular subset of men, but what it is to be a man in general. To have the concept man is to have a universal idea that applies to all possible men, not only those that do exist or have existed, but also all those that could exist. Second, there is the capacity to put these ideas together into complete thoughts, as when you combine the concept man and the concept mortal in the thought that all men are mortal. Third, there is the capacity to infer one thought from others, as when you reason from the premises that all men are mortal and Socrates is a man to the conclusion that Socrates is mortal. Obviously the capacity to grasp abstract or universal concepts is the most fundamental of these three. You couldn’t form complete thoughts or reason from one thought to another if you didn’t have the concepts that are the constituents of the thoughts.” (31)

But if this is the case then intelligence requires an immaterial and simple soul to exist and premise 1 is established. Now, it is obvious that the machines in which so-called artificial intelligence now occurs are not immaterial and simple things: they are material and very complex systems of parts. Therefore it seems clear that premise 2 is true as well. Since the conclusion follows validly from the premises, we can conclude that AI machines, despite the hopes of those involved in Artificial General Intelligence and/or Generative Artificial Intelligence, will never have certain capacities of human intelligence, namely, capacities to think universal ideas, make judgments, and form arguments. This is not to overlook the amazing things Artificial Narrow Intelligence has achieved and the ways in which various AI functions supersede our abilities. But if my argument is sound then we must say that any AI system which appears to manifest rational intelligence only does so because there exist intelligent humans who programed them to have such an appearance and interpret them as having things they don’t actually have.

A Few Objections

Here are just a few ways to object to the foregoing ideas which are obviously controversial.

The conception of intelligence offered above might be questioned and replaced with another. For example, John Dewey defined intelligence in his book Reconstruction in Philosophy (1920) as follows:

“Observation of the detailed makeup of the situation; analysis into its diverse factors; clarification of what is obscure; discounting of the more insistent and vivid traits; tracing the consequences of the various modes of action that present themselves; regarding the decision reached as hypothetical and tentative until the supposed consequences which led to its adoption have been squared with actual consequences. This inquiry is intelligence.”

And this more contemporary definition was presented in a report, endorsed by 52 experts, entitled “Mainstream Science on Intelligence” published in the Wall Street Journal on December 13, 1994:

“Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. It is not merely book learning, a narrow academic skill, or test-taking smarts. Rather, it reflects a broader and deeper capability for comprehending our surroundings – “catching on.” “making sense” of things, or “figuring out” what to do.” (see the WSJ report here).

These definitions seem different from Feser’s traditional philosophical account intelligence which is consistent with the Platonic insights I explored. But we might wonder whether or not Dewey’s intelligence as inquiry would require the ability to think universal ideas, make judgments of the same and the different, and form arguments. If it does, then Dewey’s formulation wouldn’t replace but would presuppose the more traditional account of intelligence. And “the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience” is bound to require the factors of the traditional account as well. In any case, the point is that plenty of different views of intelligence can be entertained. Indeed, Ian J. Deary, in his book Intelligence: A Very Short Introduction (Oxford, 2001) reports that the 1996 intelligence task force report from the American Psychological Association observed that “When two dozen prominent theorists were recently to define intelligence, they gave two dozen somewhat different definitions” (118). So with different views of intelligence on the table we would presumably have different views of what artificial intelligence is and what it can or cannot achieve. These different views might open up ways to get around my argument.

We might also consider an eliminative materialist perspective which holds that things like ideas, judgments, arguments, and so on are nothing but “folk psychology” that have no place in the brain and therefore no place in reality. The task would be to show how a neurological account of the aspects of rational intelligence discussed above could be given. Is our experience of thinking universals, drawing logical inferences, and constructing arguments really a set of brain functions? As Tyler Burge pointed out above, “According to the natural sciences, reason is not a structural feature of material composites.” And “According to modern natural science, there is no place in the physical structure of material composites for rational, propositional bonds.” Now, just because there is no place in for reason in science doesn’t mean it doesn’t exist. But it may strongly suggest it doesn’t and those who align themselves with science may find the eliminative materialist position attractive. To be sure, such materialists appear to be self-refuting insofar as their position appears to use ideas, judgments, and arguments. But many of them will deny the charge of self-refutation and, if they can do so successfully, it might open up the door to a vision of genuine AI that doesn’t require soul and immaterial things at all.

Finally, many questions arise if we find ourselves committed to souls. For example, where do these immaterial and simple substances come from? How does one soul pair up with one particular body? Do we really want to accept a theory that appears to be inconsistent with the prevailing naturalism of the sciences? Don’t various forms of brain damage show we are made up of material parts after all? And how can a non-physical soul interact with a physical body in the first place? These are tough questions for so-called substance dualists. But there are plenty of plausible responses to them (for a helpful overview go here) and those convinced by some of these responses might find the belief in soul to be justified.

Conclusion

I hope to have shown that Plato’s insights from his dialogue Timaeus that (1) intelligence requires soul or an immaterial, simple substance and (2) one of the soul’s primary functions is judgment and, in particular, judgment of immaterial things that remain the same, while certainly controversial and even bizarre given the dialogue’s metaphysical and mythical context, can, on the one hand, be defended with plausible insights about the nature of ideas, judgments, and arguments and, on the other hand, be connected in an interesting way to AI. My argument offers us a way to question the possibility of forms of AI that promise to do exactly what human intelligence can do. Of course, the argument only excludes forms of AI when these forms are thought to be a function of material and mechanical aggregates alone. Should such aggregates develop soul, whether as an emergent property or in some other way, then new possibilities might appear for artificial intelligence to engage in the very same rational activities we ourselves do.

Go here for my favorite argument for the soul which includes many insights from this post.

Go here for my other posts on the soul.

Go here for my many posts on Plato.