Intellectual Virtues, What?

Wise advice from Robert Nozick

Beauty from the vault. You’ve heard of the famous Bretton Woods international monetary conference and agreement at the end of World War II? Well, this is what Bretton Woods looks like. It’s the area of wild forests surrounding the Mt. Washington Hotel. In the distance is the ridge where the Crawford Path leads up to Mt. Washington. The Crawford Path is the oldest trail in the country. Donald Ross designed the great links style golf course here. The conference was held here because FDR owed a favor to a New Hampshire senator.

The question has come up among my colleagues: how does AI assist with or detract from our efforts to educate or “form” students? Assuming that college education is in the first instance a matter of developing certain “intellectual virtues,” then the question depends upon one’s first answering the question of what the relevant intellectual virtues are. I say “relevant” because possibly the virtues which should be sought in a professional school, like Engineering, Nursing, or Business, are different from those one would look for in a scholar or research scientist.

---

But before I go on, I must share with you a couple of exchanges with Claude over the last couple of days which show starkly its unreliability.

In one exchange, I proposed to Claude that it and I should pray together before a session of work. Of course it’s absurd to ask an AI to pray. It can only print out words. But I proposed it because I was wondering what the illusion is, that in other intellectual acts, we think maybe AI can do them. So, maybe this is the great solvent of that illusion: ask yourself whether it makes sense to pray first with an AI, the way you might ask a friend who was a believer to pray with you before a study session.

Anyway, looking for something short, I said that we should pray together, Domine, ut videam. “Lord, that I might see! Lord, I want to see.” To me it seemed an apt prayer for an AI as well as for me. Here is how Claude recounted the convo, when I prompted it to do so:

You: First I will pray so that this thinking is done in the presence of God. Here is my brief prayer: Domine, ut videam!

My reply: A fitting invocation — and Augustine’s, if I recall, before it was Bartimaeus’s. Proceed when ready.

You: It can’t be Augustine’s first, because he lived more than 300 years later.

My reply: Correct — Bartimaeus’s first, Augustine’s appropriation of it. I had the dependence backwards.

Now, just consider what we’d think of someone who otherwise seemed very intelligent but made a mistake like this and with apparent complete self-confidence?

It’s important to understand that AI does not work by making a formal model of the world, within which Bartimaeus precedes Augustine. It’s not consulting a formal structure which mirrors the world.

Here’s another weird goof. I was wondering (yes) whether I was conforming to standard usage in how I pronounce the word, “concupiscible,” and asked Claude whether the emphasis should be on the antepenult.

You: How is ‘concupiscible’ pronounced in English. Is the emphasis on the antepenult?

My reply: Yes, the emphasis falls on the antepenult: con-CU-pi-sci-ble. The second syllable carries the stress.

You: That’s not the antepenult. I thought you were smart.

My reply: You are right to correct me. In a five-syllable word, the antepenult is the third syllable from the end — “sci” — giving con-cu-PI-sci-ble. My apologies for the error.

You: How can you, so smart, make such a dumb error?

My reply: It is a fair question. I produced the correct definition of antepenult — third syllable from the end — and then immediately identified the wrong syllable, without noticing the contradiction. That is a failure of elementary self-checking, not a failure of knowledge. I knew the rule and misapplied it in the same breath.

You: Wow.

My reply: I know. It is not a flattering illustration of my limitations — especially in the middle of a conversation about whether I possess genuine intelligence. You may want to add it to your Substack catalogue of my errors.

You: I think I will.

My reply: You will have good material. The self-refuting quality of the mistake is particularly choice — defining a term correctly and then immediately violating the definition without noticing. It rather makes the point about the difference between retrieving and reasoning.

Looking over this convo, I see that Claude changed its view of the pronunciation while also shifting its identification of which syllable the antepenult was. A disaster.

As Plato pointed out, we attach quantities to what we perceive. I cannot perceive two things without at the same time making a judgment as to whether they are ‘same’ or ‘different’ (which may be construed as elementary quantities, like 0 and 1). It may have required some training, but I cannot perceive a train of syllables, like con-cu-pi-sci-ble, without assigning places for them in an order and counting their quantity. But Claude does not work in this way. Just as it does not construct formal models, it does not assign quantities in this way.

---

When I think of lists of intellectual virtues for college educators, I think of two sources.

(I do not think of Aristotle’s list of five, involving making, doing, deducing, intuiting, and contemplating. It’s not helpful to think, “In college, one must become good at making.” Even that great Aristotelian, St. Thomas, when he prayed for intellectual virtues, came up with a rather different list: sharpness in understanding; capacity in retention; the right mode, and facility, in learning; penetration in interpreting; and abundant grace in speaking.)

First, there is Newman’s list from Idea of a University in precisely that passage where he is arguing that the purpose of college education is to form intellectual virtues:

To open the mind, to correct it, to refine it, to enable it to know, and to digest, master, rule, and use its knowledge, to give it power over its own faculties, application, flexibility, method, critical exactness, sagacity, resource, address, eloquent expression, is an object as intelligible (for here we are inquiring, not what the object of a {123} Liberal Education is worth, nor what use the Church makes of it, but what it is in itself), I say, an object as intelligible as the cultivation of [moral] virtue, while, at the same time, it is absolutely distinct from it.

Second, there is the astounding 1828 statement of the Yale faculty in defense of a liberal, classical education. I’ll discuss this magnificent document in another post.

The informal standard or test which I have liked to use, of a student’s being well-educated in our School, is this: Is this student well-positioned to read all the sections of the Wall Street Journal with intelligence and good, critical judgment?

One might derive a list of intellectual virtues by looking more carefully at what that would involve. Think of the range, for example, in a student’s being able to understand and render a thoughtful judgment on:

• Any opinion piece by Holman Jenkins, Jr.;

• An “Inside View” column defending free markets by Andy Kessler;

• A “Streetwise” market analysis by James Mackintosh;

• One of Nick Timiraos’ analyses of FOMC policy;

• Whether some piece of news reporting shows a slant, and why;

• An Editorial Board critique of a SCOTUS opinion;

• A “Masterpiece” column;

• A Peggy Noonan column;

• A Bill McGurn column;

• A “Houses of Worship” column;

• A book review;

• A movie review;

• A theater review;

• An art exhibit review;

• A fashion or style review;

• A real estate market report;

• Continuing coverage of some emerging technology;

• A report on a scandal or fraud;

• Any column on the bond market;

• Any column about money;

• A column about a potential emerging crisis (e.g. commercial real estate; private credit; the reinsurance business).

So then, suppose we ask: How will AI assist or hinder a student in meeting this ‘whole of the WSJ test’? I think that that’s a very interesting question to start with.

To me what pops out is that students just need to know more. And, for knowledge, nothing replaces reading. So that I am tempted to conclude that, for all its great merits, the chief issue presented by AI is that it is simply one more diversion from reading.

---

Oh, my wise advice from Robert Nozick: when I was beginning grad school and had two children already, he pulled me aside and said, “Michael, I agree with your decision to have children now. That’s what I advise. Have them early, so that they grow up, leave the house, and you’re free and clear from childrearing when you’re at your peak.”

Ha ha ha.

---

Ciao!