AI Puzzle-Piece #1
From time to time I plan to reflect on the implications of AI for theological anthropology, the Christian understanding of being human. The first puzzle-piece concerns AI’s capacity to eliminate jobs.
Among the jobs that (some say) AI will eliminate is the job of a coder, a person who writes code for computer programs. In businesses heretofore, a new coder hire would receive instructions to write a program (code) that does specified things. There would be a middle manager, overseeing the work of several programmers and seeing that everything in a particular project fit together to meet expectations. This obviously is a very complicated division of labor, often involving more lines of code than any one person could comprehend. In this complex situation, middle managers don’t write code themselves. Yet it is significant they could write any part of the code if necessary, since they were once coders themselves.
AI is poised (some say) to eliminate the jobs of coders. Now the middle manager will tell AI to produce code that meets specifications. And AI will do so, rapidly. No one will be able to understand why or how the code has been written as it has, given its complexity.
But what if the program turns out to have problems? AI will be asked to fix the problems, which it will, leaving us still with code no one understands. In the old scenario, the middle manager had the coding knowledge to dig (if necessary) into the guts of the program and to figure out where it was going wrong. This was knowledge born of experience and included, in the term that Michael Polanyi used, “tacit knowledge,” knowledge that cannot be reduced to words on a page.
In the AI future as some see it, no one will need to learn how to write code, and so no one will have the knowledge to “check out” the product that AI produces. This seems likely to create novel risks for businesses that have eliminated their coder-employees.
—
It was Jarett Malouf in The New Atlantis who first alerted me to this issue. He calls it “a de-skilling epidemic.” He writes: “How good will we be at our AI-supervision jobs if we lose the skills AI replaces? As every engineer knows, the quality of a manager’s work is largely limited by their own technical expertise. A trope in the software engineering world is that of the overly ambitious, borderline antagonistic manager who knows nothing about what he’s asking the engineers to do and expects it done in a miraculous timeline.”
Malouf continues: “There’s an argument to be made — a bet, more so — that the hard skills we were once required to master are now and forever obsolete, and thus that the withering away of low-level competencies is unproblematic. But I would venture that what has always been true in software engineering will remain true: he who understands the layer beneath will thrive.”
—
I have a very simple story here, a sort of reverse-image. Between college and seminary, I taught junior high mathematics for a couple of years on a Pueblo reservation in New Mexico. My students had not memorized the multiplication table. When they needed to multiply, say, 7 times 9, they would mark out nine slashes on their paper, then count over their slashes seven times. They would have saved a lot of time by memorizing (when they got to 63, I’d say with a smile, You know, it’s going to be 63 tomorrow too). They would also be more accurate (since any of us, counting over a line of slashes, is going to make mistakes). But, bless them, my students knew what multiplication meant!
If you, dear reader, were ever challenged as to why you believe 7 times 9 is 63, you could do what my students always did: you could go back to first principles and produce the answer.
The problem with AI (though on an incredibly more complex level) is that, when AI gives us the answer (the 63), we aren’t able to go back and produce the reasoning that led to the answer. And the people who used to be able to do so are in danger of being fired because no longer needed.
Malouf’s point is, they are needed.
—
The realistic upshot, then, might be: There will be concrete, bottom-line risks in the work that AI does for us. Those risks require we have people who can understand what it does and how it produces its product. A machine, it seems, ever remains a machine.
This seems realistic to me. But it is neither inevitable nor without its own puzzles. We need to keep thinking—and try to think better.
—
Good Books & Good Talk: Sunday, August 30, at St. Matthew’s Cathedral in Dallas, from 5 to 6:30 p.m. We will begin our series on Dostoyevsky’s The Brothers Karamazov, discussing Part One (about 150-200 pages). Any translation is fine, but I recommend finding a copy from a reputable publisher.
Save the Date: The Dallas Fall Theology Lecture has been set for Sunday, Oct. 4, at 5 p.m. at St. Matthew’s Cathedral. I will speak on “What’s Salvation Have to Do with It? Reflections after the Camino.” The Reverends Tom and Kate Smith (who have also walked the Camino) will respond.
—
On the Web. Malouf’s essay is called “Stack Underflow: Losing the Craft of Coding,” and you can find it here: https://www.thenewatlantis.com/publications/stack-underflow-losing-the-craft-of-coding