The Dream Machine, M. Michell Waldrop

Turing's Cathedral: The Origins of the Digital Universe, George Dyson

It wasn’t so terribly long ago the smartest people alive were trying to get a large machine to do basic math. Granted, it was a lot of math, but think of it: Scores of mathematicians stuffed in rooms full of thousands upon thousands of tiny little lightbulb-like devices, solving quirky logistical problems like how to prevent a primitive air conditioner running full tilt from breaking, all so a machine can do basic math.

Turing’s Cathedral and The Dream Machine are two overlapping but complementary narratives of why people did this, what potential they saw in computers, and how they faced—or, in many cases, didn’t—the awesome potential for harm their inventions wrought. In The Dream Machine, the protagonist, J.C.R. Licklider, is a complex but pivotal figure worthy of a book-length treatment. Turing’s Cathedral lacks any such focus or structure. Neither book is a masterpiece from a literary perspective, but I got a lot more out of The Dream Machine, reminding me of the passion for technology I once had.

Any sufficiently useful technology undergoes a gradual transition from invention to appliance. Today’s brilliant new thing is tomorrow’s car or washing machine. Computers completed that transition with the smartphone. So I don’t think of, say, AI, as an invention. It’s just another app.

But maybe that’s wrong. I use AI frequently for work. The best explanation of how I use it comes, ironically, from mathematician and engineer John Von Neumann in the late 1940s:

“The importance of accelerating approximating and computing mathematics [sic] by factors like 10,000 or more, lies not only in that one might thereby do in 10,000 times less time problems which one is now doing, or say 100 times more of them in 100 times less time—but rather in that one will be able to handle problems which are considered completely unassailable at present."

Von Neumann was an infinitely better mathematician than writer, so let me rephrase: Computers would be so much faster at math that it would expand the universe of problems mathematicians would even attempt to solve. This has been my experience with AI as well. At work, we call it “AI Intern,” because we often assign it tasks that would have previously required a small army of interns. You would never rely on interns to do work for you (I hope!) but a bunch of ideas that previously would have been far too time-consuming suddenly become possible. For our story on New York City’s megamansions, we had an AI program “read” thousands and thousands of documents to identify likely townhouse combinations. Keyword searches were insufficient. Each document was, in some cases, hundreds of pages. It would have been a PhD-thesis amount of work. With LLMs, it took us about a week. This gave us a much smaller pile of documents to look over ourselves.

Now, Von Neumann was talking about a particular type of “unassailable” problem: building the “Super” bomb, which would lead to the development of the hydrogen bomb, a different but no less horrifying implement of destruction. It seems whenever humans can do math faster we use this ability to kill each other better. In fact, as Von Neumann demonstrated, we sometimes invent ways to do math faster specifically for the purpose of killing each other.

People seem to have bought into two cultural narratives of what AI will do. On the one hand, some believe AI is a scam that has little more utility than writing bad poetry or generating fake nudes. Others believe it is so powerful it could destroy/enslave humanity. I won’t pretend to speak for either extreme view. Instead, I want to zoom out. We’re still talking about rooms full of calculators here. The rooms are bigger, the calculators are more powerful, the cooling technology more complex, but nevertheless, it’s math. The code bases are still ones and zeros. To talk about what AI is, or is not, or will be, begins and ends with what people will do with it. Will people use it to help drop bombs or parse an historical document?

John Lindsay, a scholar of military technology at the Georgia Institute of Technology, recently wrote about how AI is used by militaries as part of “decision support systems” which have been around for decades:

“As the name implies, decision support systems support human decision-making; AI does not replace people. Human personnel still play important roles in designing, managing, interpreting, validating, evaluating, repairing and protecting their systems and data flows. Commanders still command.”

For anyone who recalls the 1964 film Fail Safe, all of this will sound terribly familiar. The short plot summary is the film depicts a nuclear crisis due to a computer malfunction. But the movie is really about the failure of decision support systems ie; human malfunction. Early in the film, an older Air Force officer laments that the new recruits are too much like machines themselves. He goes on to describe something that sounds an awful lot like cognitive surrender, a much-discussed phenomenon in the age of LLMs. Again, this film was made more than 60 years ago.

Will we lose our humanity to AI? Will we lose our lives to AI? These questions are not posed to provoke an answer, but to stoke awe in the very existence of the question. In the face of this fear, the natural reaction is to wish none of this crap had ever been invented. I sometimes give in to this impulse and wish I lived in pre-internet days. Occasionally I think I even mean it. But I mostly don’t, especially since I read The Dream Machine. For the first time in years, when I think of “technology” I don’t picture the Big Tech Companies. I see Licklider and his cohort sweating over a table saw. This really is the product of a bunch of nerds tinkering with rocks and crystals. We humans turned the stuff of Earth and stardust into machines that can do math. It is, in the most literal sense, awesome.

Moreover, even if we wanted to wish all this away, it’s irrelevant. We cannot. We’re never going to stop humans from inventing. And invention will inevitably have terrifying ramifications. I don’t know where we go from here. I am equal parts excited and afraid. Before I read these books I thought I had to be one or the other. But even the people who invent this stuff don’t know where it will take us. Dreams, after all, are also the stuff of nightmares.