Augustine’s Invention of the Inner Self: Legacy of a Christian Platonist by Phillip Cary

The Self-Invented Self and AI

Some thoughts from Saint Augustine…himself

I know I'm not the only person to feel rather disturbed by the last few months' progress with LLMs (Large Language Models) - from ChatGPT to the announcement of GPT-4 being able to describe the contents of pictures, I’m in fact quite freaked out. I'm equally certain I'm not the only person to search for solace (or at least serenity) in philosophy. Hell, I even doubt that I'm the only person to look to Saint Augustine, but I still suspect that (on that score at least) I'm reasonably unique.[3]

A little context might be helpful before we dive into the book so that you understand why I made such an unusual choice (though I won’t hide what it is: Augustine’s Invention of the Inner Self: Legacy of a Christian Platonist by Phillip Cary, a philosopher of religion at Eastern University in Pennsylvania).

I've always been a bit of an AI skeptic. Perhaps relatedly, I'm neither a functionalist nor any other variety of materialist when it comes to the philosophy of mind. Mental phenomena (thinking and experience) seem obviously something of a different kind than the sorts of phenomena we call physical.[4] Whether we're talking Newtonian-described billiard balls careening about, quantum weirdness, chemical reactions, or whatever the heck else, my reaction to the arguments of philosophers like Daniel Dennett or the Churchlands, that the mental reduces to that, has always been one of simple incredulity. How can someone believe it?

Progress in machine learning didn't do too much to trouble me, and what little trouble it did cause me was assuaged by books like The Myth of Artificial Intelligence, in which Erik J. Larson made cogent (albeit incomplete) arguments that at the very least some mental phenomena such as abductive inference weren't being touched by progress in AI.[5] In broad strokes, he asserted that current approaches to AI were using induction to try to brute-force emulate other human capacities in a way roughly analogous to trying to use a hammer as a screwdriver.

And then ChatGPT was released, with Bing Sydney and the official GPT-4 release following shortly thereafter. And my comfort that "inductive" approaches alone would be insufficient took a bit of a body blow, enough for me to wonder if I was using hammers wrong the whole time why I was so skeptical of the idea of AGI and why I found the materialist and functionalist approaches to the philosophy of mind so thoroughly unconvincing. And if these AIs were starting to approach a cognitive level similar to humans, but they were doing it through completely different means…well, that was a profoundly creepy thought in its own right.[6]

The obvious explanation of my discomfort is probably religiosity. I am a Christian, albeit one whose sympathies mostly lie with the sort of mainline Protestantism some deride as watered-down. But Christianity isn't necessarily non-materialistic. Certainly the religious practices in much of the Old Testament seem quite this-worldly, with notions of an afterlife generally being vague at best.[7] Even the view of the afterlife seen in the New Testament is resolutely physical. We’re talking a literal resurrection of a literal, yes-you-can-touch-it-Thomas-if-it-will-shut-you-up body.

I know that a lot of people ascribe Christian beliefs in the non-materialist vein to Greek influence, to it being a simplified Platonism suitable for the common run of humanity.[8] So I started to wonder - where exactly had I imbibed the intuitions that made dualism so plausible to me and AGI so much the opposite? Was I right to suspect that it came through my religious faith? Were those intuitions something I could jettison while holding onto other commitments like that faith? If I did jettison them, could I make my peace with AI?[9] And then I remembered a book I'd first read some time before, a volume on Saint Augustine.

Summary of the Book

As I said above, the book is Saint Augustine's Invention of the Inner Self: Legacy of a Christian Platonist. In this work, Phillip Cary argues that the notion of an "inner self", a private pseudo-spatial zone "within us" in which we can look for our deepest (notably, not "highest") thoughts and inclinations and yearnings, was an "invention" of Augustine. The ambiguity of the Latin invenire, which can mean something similar to discovery or finding of something already present as well as creation of a novel thing, provides Cary with a bit of a relief valve, but Cary does believe Augustine created something new[10] and distinct from previous conceptions of the mind in the Western philosophical corpus. In Saint Augustine’s Invention of the Inner Self, he tells the tale of how that creation happened and also touches on how that creation influenced later developments in Western Christianity (though that is something he covers more in two later books that make a trilogy with this one).

Saint Augustine was born in northern Africa, in a region where (per Cary) the Christians were fervent though perhaps unsophisticated - a "Bible Belt." Unsatisfied with his mother's putatively simple-minded faith Augustine was a Manichaean in his youth.[11] Augustine felt he could not conceive of things without some sort of spatial extension, thinking that for something to exist was for it to exist in space. The Manichaean dualists, whether you consider them heretics or members of a competing religion, seemed to the young Augustine to provide a view of God and spiritual reality that comported with that conviction. Per Cary, and contrary to the modern idea of Manichaeans as being more “spiritual” than orthodox Christians, the Manichaeans treated souls as a sort of light or rarefied matter intermingled with the unclean matter of the universe, which meant that these souls had spatial extent.[12] The young Augustine even understood God’s omnipresence by Him being present part by part or bit by bit in each place, rather than Him being fully present in every place.

However, under the influence of Neoplatonism delivered secondhand via writers such as Cicero and by skilled preachers such as Saint Ambrose of Milan, Augustine gradually investigated other schools of philosophy that unsettled his Manichaean beliefs. His conversion to Christianity was also largely a conversion to Neoplatonism. And, interestingly, Augustine was arguably still a heretic even after his conversion for some time, being a better Neoplatonist than he was an orthodox Christian.

The young Augustine spent years trying to incorporate Neoplatonism, specifically the more developed version he picked up, at least indirectly, from books[13] by writers like Plotinus (or at least his student Porphyry), with orthodox Christianity. Cary traced a few false starts the young man made. Plotinus (and others, following an Aristotelian-derived idea of knowledge as identity between knower and known) believed that the soul was itself divine in part and so the highest part of the soul was identical with God, or at least a god, so that when we looked inward in contemplation at our higher selves, we looked upon that god.

To Neoplatonists such as Plotinus, there was a triad of divine entities:

The Ineffable One that is the source of everything else in existence.
The Divine Mind that corresponds both to Plato’s Forms and to Aristotle’s self-contemplating god.
The World Soul that animates the universe, with all our human souls participating in it or maybe emanating from it.

Each of these entities could be referred to as a god, and each succeeding one was ontologically dependent on the higher. This doctrine bears obvious resemblances to the Christian doctrine of the Trinity, and the vocabulary and logical machinery of these thinkers was used to explicate the fully developed Trinitarian doctrine coming out of Nicaea in this period. In hindsight, the hierarchy between these three distinguishes this from the true doctrine of the Trinity in which Father, Son, and Holy Ghost are co-equal, among other differences, but it’s easy to see how Augustine could be confused.

Augustine attempted to give the idea of finding God in the soul a Christianized spin. He identified a part of the soul with Reason, hinted to be Christ.[14] He tried to sketch out a program by which that divinity could be approached through a program of study in the liberal arts,[15] but as he grew in understanding of the Christian faith he had embraced, first as a catechumen and then as a newly-baptized Christian, he eventually noticed the heresy in the idea of identifying the created soul, even in part, with the Creator God. Though Augustine changed up his program, he didn’t abandon the idea of using study to get closer to God.[16]

So in later attempts, Augustine shifted the metaphor a little bit. Where Plotinus spoke of looking inward to a sort of shared divine soul, Augustine talked about looking inward and then upward (in Cary's terminology) in order to see God, reflecting the ontological separation between Creator and created.

To Augustine (at least as Cary tells it) seeing God was rather less mystical, at least in the senses of that word it usually carries today such as hiddenness, deepness, or the unexplained. To Augustine, to see God was to see Truth; it was to have insight, in a similar sense to understanding a new concept or theorem. Augustine’s view was that salvation would be an eternity of perfect, expansive understanding, not mediated by language or discursive, step-by-step argumentation or deduction. In fact, Augustine seems to have considered language a consequence of original sin, with un-Fallen beings likely being able to simply see into each others’ souls.[17]

But because Augustine thought in terms of a 2-part movement, and the intellectual vision and (in this life) dazzling, momentary glimpses of God were what happened when you looked upward, that left the question of what you saw when you just looked inward still open.

Augustine identified that “place” (and what one would see there) with the capacity of memory and its contents. Because Augustine now believed in the Platonist/Neoplatonist idea of a sort of causal hierarchy where the lower could not affect the higher and further because of his belief (related to the causal hierarchy) that signs could not be properly understood without first understanding their referent, this made the individual human mind the site of meaning in a physical world.[18] He further emphasized the sheer spatiality of memory, that it extended near-infinitely far even inside a finite human being. Augustine conceptualized an inner self, a dimension of reality comparable to the external world in scope and magnitude, at least metaphorically.

This is what Phillip Cary meant by the title of his book. This is the invention, and also the foundation of later ideas like the Lockean camera obscura. When Daniel Dennett argues so vociferously against the idea of the Cartesian theater in his work, it is a development of this basic conception that he opposes.

Interestingly to me, I think this might be the source of the intuition behind the Chinese room thought experiment,[19] as well. If thinking “should” be happening in an inner self, then if an inner self is not present, the metaphor of an empty room actually makes a lot of sense (and since I believe in this sort of inner self, this also explains much of my disquiet with something that seems to think without having one). Though first we should probably deal with whether Cary was right as an historical matter. Is this a unique intellectual concept traceable to Saint Augustine?

Is the Case Persuasive?

My key historical objection (and one I’m sure that has occurred to several of you) is that memory, in the form of a memory palace, already had connotations of spatiality in Greco-Roman thought. Maybe this was new to Augustine, but he was trained in rhetoric, so that seems unlikely. Was Augustine simply re-inventing the ars memoriae?

But Cary anticipated that objection and dealt with it ably. Explanations of memory palaces usually remained very closely tied to the physical structures which the rhetoricians used, their lighting and other physical details. Cary criticized traditional translations into English of some treatises on the art of memory, pointing out that writers on that art contrast literal seeing with thinking about things seen in the past, with verbiage suggesting an inner self often read into the text by later translators. Their view of how the art worked is that they were thinking about what they’d seen, not locating those sights within an internal structure inside their minds. He dealt similarly with the art of finding - inventio or heuresis (obviously etymologically-related to Augustine’s “invention”). This involved the use of topoi or commonplaces as patterns to couch and structure arguments. To Cary, the “places” referred to in discussions of this art are still places you actively “look for” something, looking metaphorically outward, not necessarily components of an inner self.[20]

Going back to memory itself, philosophical writings on its actual nature[21] often invoked the idea of a wax tablet, with memories being physical impressions left in the mind analogous to the written forms pressed into those tablets by a stylus.[22] Which made sense given the popularity of schools such as Stoicism, which explained the mind as a physically-existent phenomenon made of "soul-stuff.”

At the end of my re-read, Cary really had convinced me that Augustine had invented (in the stronger sense of novel creation) a new conception of the mind, at least within the Latin/Western tradition.

The Lockean camera obscura or the Cartesian theater of physicalist philosophers' opprobrium can be traced to this separated inner self. And along with that concept came the Platonist-esque separation of meaning from physical phenomena.

Having an inner dimension apart from (maybe orthogonal to?) to the external world makes the idea of dualism, of non-physical capacities constituting the human mind a lot more plausible. So the split between induction as a statistical procedure and abductive inference seems intuitive, rather than abduction being a simple specialization of induction distilled over extremely large data sets. It seems intuitive that a computer, if it is not human and does not have an inner self, cannot really be thinking, because it does not have any place to think.[23]

But if the inner self is a mistaken or confused idea, then a lot of those intuitions lose their force.

Is this Actually Important? Or Just an Amusing Excursion?

Is this a productive vein to mine for insights about AI? On one level, the level of developing new AI capabilities or looking for ways to improve the interpretability or controllability of these models, it’s probably not. A Latin-speaking proto-Berber from over a millennium ago is unlikely to have any insights directly pertinent[24] to those efforts. But I still would say that it is productive in a broader sense. The AI itself is not the whole story. How humans react to it, how we use it and abuse it, how we trust it or don’t…all those are questions where our basic intuitions about what it means to be a person and what it means to think are very relevant.

There’s a common joke that the Reformation, and the centuries of bloodshed and strife that followed, was an argument in the mind of Augustine, that ideas he’d had centuries and centuries before resurfaced and duked it out. I think that “joke” isn’t. Instead, it has a lot of truth to it. And today I believe it is also true that very old ideas are animating our thinking and discussion whether we realize it or not.

I am not saying Saint Augustine of Hippo is the source of all those ideas. In fact, Plato is probably a more important source, but I do think Augustine is one source and is a means by which Plato’s influence was passed to later Christians in the West. And I think it is very much worth looking backward to determine how much of how we think and what we value is inherited from our predecessors.

Certainly, I have at least found it valuable personally. For those who are skeptics about AGI (or like me to a large extent still are), or for those who aren't, challenging why you find certain arguments more or less persuasive is a good thing and figuring out the historical context behind how you analyze events is also useful. Modern tools used in philosophy of mind, in our attempts to understand ourselves, have drawn on old ideas before.[25]

Looking around at people’s reactions to news on GPT and related models, I can see several places where we are drawing on something that Augustine either introduced or exemplified.

Looking at Scott Aaronson’s thoughts on why he is less scared of AI than many of his interlocutors,[26] he seems to pattern-match AI to a “nerd” in a “nerds vs bullies” dynamic,[27] its intelligence, even if greater-than-human, being something we should admire. He also rejects the Orthogonality Thesis, which I gather is the idea that any intelligence can be applied to any goal, no matter how mind-numbingly boring or weird or random the goal might be. Instead, he believes in “knowledge and enlightenment as instruments for moral betterment.” That this bears a very strong resemblance to Augustine and other thinkers’ views of education as a path to God I think is obvious.
There is a similar echo in Bret Devereaux’s critique of ChatGPT in education.[28] To Devereaux (and honestly, I think to any good teacher), the point of the essay is not the essay itself. It is the thought behind it. Outsourcing the writing of the essay is to some extent also to outsource the thinking, and learning how to think and organize one’s thoughts is precisely the point. Looking at Bret Devereaux’s other writing, I don’t think anyone would be surprised to see that he holds to an old-fashioned view of the liberal arts, one in which their study is important not just for the knowledge gained, but for forming the character of future leaders and citizens.
Augustine’s Platonist-derived privileging of the inner self above the world is in the same line as Zvi Mowshowitz’s worries[29] that an AI could potentially eradicate “all value in the universe.” Augustine and Platonists did not place meaning in the external world, and Augustine did a lot to make us instead think of meaning as residing inside human skulls. If all value in the universe exists in human skulls, then perhaps an AI extinction risk scenario would destroy all value. But if it is at least partly objective, partly inherent to the outer world, then presumably it would still be there for other creatures to enjoy in the long term.

At the end of this, I’d like to have a stronger conclusion (“and that’s why the Butlerian Jihad in Dune had the right idea”), but I really don’t. At least I don’t have any conclusion that wide-ranging. I probably will start using GPT-driven tools more in my work to write boilerplate code and wire up endpoints, but I’m not looking at it for any actual insight yet, certainly not ethically or spiritually.[30] I’ve found Cary’s reconstructed Augustine convincing enough that I’m not quite prepared to give up those inner self intuitions, which leads me to doubt GPT will really have anything all that interesting to say on those matters.[31] And if the practical matters of making a living and engaging with the physical world really are automated out of humans’ hands, those higher matters will probably be all we have left, so maybe we should call “dibs” in advance.