I played on the soccer team [1] at Northern Valley-Old Tappan, could hold my own in a scuffle and got a lot of detention [2], but my classmates still found the time I spent reading baffling. Less a spur to mockery than puzzlement.
Freshman orientation at Brown was new and overwhelming — many different people loudly saying many different things. After a day-and-a-half of this I retreated into my room, laid down on my bed and picked up whatever book I had been working my way through [3] to counter-balance the overstimulation.
I heard a rustling at the end of the bed and looked up. Another freshman sat there watching me read. “Here it comes,” I thought, “He’s going to ask me why I’m in here when everyone else is out there.”
I was wrong. He said, “I liked that book very much. What do you think of it?” I had found my people. That’s why I (like many others) stay so attached to the school — it was formative more than credentialing.
Sometimes people chuckle at the juxtaposition of Brown University and McKinsey. [4] I see it as two parts of a whole — because some of the things I learned on College Hill have proved intellectually and practically essential as I have sought to redesign business processes or technology organizations. AI will force universities to change no less than the private sector must -- and just like businesses universities need to innovate ruthlessly while preserving what’s special and human.
At their best universities form individuals by teaching them to interrogate reality carefully and allowing them responsibility to act within it. AI can deepen that -- not diminish it.
Academia and business collaborate on knowledge creation and talent formation
The academic and business worlds are more intertwined than they admit: different structures and economic models, but overlapping missions. They succeed when they are truth-seeking institutions, and they both have a joint charge to form the next generation of professionals.
I like to tell Brown’s Computer Science professors, “You get them for grades 13 through 17, and then you send them to me and I try to look after them for grades 18 through 28 or so.” Students may shiver at the thought of 28th grade, but universities and corporations have a joint mission and joint responsibility for the intellectual (and ethical) development of people in the nascent stages of their professional careers.
No less than universities, corporations in a market economy are epistemological institutions. As Adam Braff and I discussed, every business decision you make depends on an understanding of reality and how it might evolve in the future -- that’s why you see so many PhDs running around trading floors and Pharma companies. [5] American corporations now fund four times as much research and development as the federal government does, and spend almost as much on basic research as Washington.
As a CIO or CTO, you might ask yourself where academic research converges with your business or technology strategy, especially now that so many companies want to reach the innovation frontier. And: we will need early-tenure software engineers. We will just need them to know and do different things than they would have in the past.
Why do academics fear AI?
I was up in Rhode Island last week, and had the privilege of meeting with about a half-dozen faculty and administrators.
Those I spoke to were quite thoughtful about how some of the same political and technology forces buffeting the business world are reshaping higher education [6]. They wanted to exploit AI for research and teaching, recommit higher education to epistemological rigor, break down the silos between disciplines, reduce the cultural gap between the academy and the American public and invest in the ethical (as well as intellectual) formation of young professionals. And I just might have urged one or two of the humanities professors to think of history as a graph.
Good! If you care even a whit for the creation and dissemination of knowledge, you are racing to figure out how AI can help you collect data and improve learning outcomes. Academics wrestle with entropic data no less than business managers do.
But not all academics are so broad-minded -- Bob Pasker pointed out using AI in any form of research unnerves some historians. That doesn’t make any more sense in the academic world than it does in the private sector. Yes, AI creates the opportunity for more slop -- but how many tendentious books and newspaper articles do you see every year -- some of them even from Brown professors? Yes, students can use AI to cheat -- they also used their Mac SEs to cheat when I was an undergrad TA for CS1 in 1990.
AI should reinforce the academy’s epistemological mission
AI doesn’t have to degrade anything important. Brown’s history department used to teach us to question everything. My junior seminar on “Problems in Historiography” assigned books by respected historians, so professors could demolish each one’s arguments as biased and distortive.
When a frustrated student finally asked how undergraduates could possibly write senior theses if eminent historians couldn’t escape bias, the professor answered, “Very good. Now we come to the point of the class,” and started to explain how we had to interrogate every source, every assumption and every conclusion ruthlessly. I remember that exchange with every slide I write and every recommendation I make to a client.
AI wouldn’t degrade that lesson. You could argue that intellectual fads that dispense with empiricism pose a far bigger threat than anything emerging from a frontier lab. Today I use AI to challenge my own thinking and writing for sloppy reasoning -- maybe professors in the current era could use AI-enabled semantic analysis to demonstrate bias and unsupported claims with far more precision and specificity than they could in the 1990s.
And it might even enhance practical formation as well
Some of the most important and most practical learning at universities derives from doing things outside the classroom.
Just as leopard cubs learn to hunt by pouncing on their litter-mates, we learned about leadership by running little institutions — with deadlines, budgets, staffs, uncooperative technology platforms and consequences.
Getting a daily newspaper to bed (or running a similar student organization) is an exacting schoolmaster, because a student newspaper isn’t a simulation of an organization. It is one. We couldn’t unring the bell if we had fouled up the reporting on how the disciplinary system suspended or expelled a student accused of wrongdoing. AI changes none of that -- and may even enhance it.
Before decades at McKinsey, what was the very first process I ever fixed? Making “Project Midnight” a reality at the Brown Daily Herald in 1991. [7]
It was a good lesson — getting there required synchronizing product and operations with the technical realities of the day. Maybe it was my first experience with systems thinking, even if I hadn’t read “The Goal” yet and had never heard of “The Theory of Constraints.”
The paper published a print edition every day of the academic year, so dozens of students stayed at 195 Angell Street until the wee hours of the morning from Sunday to Thursday. We used Microsoft Word and Aldus Pagemaker to get the paper pasted down and ready for Cliff to take it to the printer.
The Herald was very advanced in desktop publishing in that era, thanks to David Temkin, but still was quite an analog affair in some respects. A tabloid-format, 600 dpi printer would have been USD 17,000, so we had to use X-acto knives and a waxer to paste printouts and photostats onto flats. I spent a year with an X-acto knife tucked behind my right ear. Adam Braff (yes, the same one -- small world) was scared of the waxer. I can show you how to toss an X-acto knife across the room safely if you like.
Under the name Project Midnight, years of editorial boards had sought to put the paper to bed before the early hours of the morning. They never succeeded. To my knowledge, not one issue had ever gone to bed before midnight.
When the outgoing board named me editor-in-chief, I reasoned we had to get Project Midnight right. Getting people back to their dorms earlier would reduce the risk that parents would pressure their offspring to quit the paper. And it would reduce the risk that I would fail any classes.
The very first night of the semester we finished the paper at 12:13 am (Amy likes to point out I can remember anything except people’s names). How did we do this?
Design for operations. Most of you have probably never upgraded the memory on a Mac II or Mac IIcx. Adding SIMM cards to a Mac II was a giant pain in the neck — you had to remove the hard drive to get to the slots. Adding memory to the Mac IIcx was a snap — somebody sat down and asked: how do we make this easier?
What if we applied the same discipline to the “product” — the design of the newspaper itself? So we removed any design flourishes that didn’t make the paper more readable and we tried to pre-build as much as we could in the PageMaker templates we used. For example, we dramatically simplified the little element that says “continued on page x,” using a text box property rather than a separate element to display a little line. This eliminated a mouse action. Small in itself, but a dozen of these simplifications, multiplied by dozens of times per issue, added up.
Assign property rights. We used Mac Pluses with 16-bit processors and tiny screens to write and edit stories. We used our three Mac IIs with 32-bit processors and big screens to do layout in PageMaker. The Mac IIs were the bottleneck in the whole process — the night editor, the wire editor, the sports editor, the opinions editor and so forth were always jostling for time on one of the Mac IIs. There was never anybody using them early in the evening, and a line of editors waiting around by 11 pm.
Having read about Ronald Coase’s paper “The Problem of Social Cost” in a “Law and Economics” class, I decided to create a schedule. National news (off the AP wire) and the features section got 8 pm slots. Opinions got a 9 pm slot. News and sports got the 10 pm slots, given that currency was more important for those sections.
Even more importantly, we decided
Editors could trade slots. If the Opinions editor had a study group at 9 pm, he could trade his slot to the Features editor.
And if you missed your slot, you went to the back of the queue.
The bottleneck disappeared.
Cheat like hell. In part we never got the paper out by midnight because nobody believed we could get it out by midnight. Most days production started around 5 pm because people had classes and the like. But there were no classes on the day before the first issue of the semester. So I asked a couple of trusted folks to start doing the layout early in the afternoon. We could only do that once, but I wanted the staff to believe that midnight was possible, and they did. The extra afternoon of layout time wasn’t really the fix. Changing the staff’s belief about what the paper could do — once, on purpose — was.
We had a network crash the second night of production, so we didn’t get the paper to bed until about 2 in the morning. But on the third night, we got there at 12:08 am. We very quickly got into a groove of completing the paper around midnight. A few times we finished up around 10 pm.
I visited the Herald’s new, much smaller offices a couple of years ago. They use a modern content management system to publish online and only put out a print edition (for appearance’s sake) once a week. They don’t have the avalanche of advertising revenues that recruiting ads used to provide, so I’m sure they don’t spend USD 8,000 per semester on alcohol. There’s no AP printer clattering away in the corner. No flats. No wax machine. No light tables. No X-acto knives. No room for the weekend magazine staff to pass out drunk on the floor. They put out a great newspaper, but from the looks of it, they could be selling insurance. Yes, long before AI, the internet wrung some of the raffishness and excess out of the Herald.
AI changes nothing about the challenge of getting a newspaper out every day and every story right. In the early 1990s desktop publishing allowed a student newspaper to do things major dailies using ATEX couldn’t. What could AI enable today’s editorial board to achieve? I bet it wouldn’t be prosaic.
Footnotes
[1] I was a goalie. My role on the varsity team: target practice. Being useful is more important than being glamorous.
[2] Just before I graduated NVOT’s principal told me, “Kaplan, every year we have students who receive academic honors. Every year we have students who rack up lots of disciplinary points. You are the first student I know of to do both.”
[3] Mrs. Prosaic Times swears I remember every irrelevant detail at the distance of decades. I cannot remember what book I read. This mortifies me.
[4] Even the legendary former Managing Director Marvin Bower was Brown, class of 1925.
[5] Corporate life is also a bit less regimented than academics suspect. They say: “you can’t just tell people what to do at a university.” If only they knew how many corporate organization charts resembled Jackson Pollock paintings and how yawning the gap between executive intention and front-line actions can be. Many executives feel like Lt. Weinberg in A Few Good Men when he asks “What’s the secret? What are the magic words? I give orders every day; nobody obeys.” Of course universities lack the bottom-line economic number — share price, profits — that corporations rally people around.
[6] The college wage premium has been stagnant since 2000 and has recently started to decline. AI may outperform classroom instruction, not just supplement it — a randomized trial at Harvard found students using an AI tutor learned about twice as much, in less time, than students in an active-learning class covering the identical material. And public confidence in higher education has declined. Barely 40 percent of Americans say they trust universities — that puts public funding at risk.
[7] I’m including here one of the first pieces from Prosaic Times for a few reasons. It ran before I built a list of nearly 6,000 subs (across Substack and LinkedIn). It’s summer re-run season, and I wanted to spend more time this weekend thinking about how to use the Corsair harness I built to decompose a project into hypotheses, propositions, analyses and facts. And because I hope, even with all the changes coming to the academy, students don’t lose the opportunity to learn outside the classroom.
