The robots have arrived.
Introduction
So today we are going to talk about A.I.
The chances are very good that you experienced a medium-to-strong emotion at reading that sentence and the chances are also good that the emotion was not terribly positive. The arrival of A.I. has guzzled electricity and water as it delivered self-harm, manipulated images and slop. What's to talk about?
Well, the thing is that "A.I." is a broader category than we often appreciate. It includes the things above, and those drawbacks. But it also has an uncanny ability to slice through some of the Gordian knots that encumber our lives and the economy. Including ones centered on data.
When I decided on this week's theme, I reached out to my friend Christopher Mims, a technology columnist at the Wall Street Journal. The timing was fortuitous; he has a book coming out this week — How to AI — about how normal people can use A.I. effectively.
"For the vast majority of us, A.I. isn't about replacing us," he told me. "It's about having a tireless assistant and tutor we can delegate things to. If you're smart about how you do that, it's a powerful tool that can save you time and energy."
Here I will stipulate — forcefully! — that this involves some offshoring of time and energy consumption to data centers. But bear with us.
"Deep research tools can be really good—and fast—at fact checking," Mims said. "But unless you pay for one and get good at using it, Google's hallucination-filled generative search summaries would lead you to believe their AI is so unreliable as to be a menace to truth-seeking."
This mirrors my experience. I began using A.I. tools last year as I attempted to build a database of poll results. I tried to use OpenAI's ChatGPT and found it clunky to the point of being idiotic. Anthropic's Claude, on the other hand, is adept at extracting numbers from lengthy PDFs — and constructing tools that use that data. As I used it more, I got a sense of how powerful the tool actually is.
Again, I am not denying the downsides. (Nor does Mims, who noted that "A.I. doesn't have to be intelligent in order to really screw up the labor market.") I am this week simply going to present an upside of which you might not be aware.
That's a lot of writing and earnestness for this newsletter. I apologize.
Chapter 1
Hope you didn't pivot your career to coding
Regular readers will be aware that I have two kids in elementary school. True to the stereotype, this does mean that I am often asked unexpected questions. A common one, for example, is "where is that plane going?"
It's admittedly an interesting question to ask! Haven't we all, at some point, looked up and wondered where those people were headed? Well, if you haven't, my kids have made up for your lack of curiosity. I've gotten adept at opening the FlightAware app on my phone.
Earlier this week, though, I had an idea: What if I asked Claude to track what planes flew over my house? So I asked it to create an app that would sit in my laptop's menu bar and do that: take my location and, within a certain radius and elevation, check for the appearance of planes. Over the course of an hour or two, I had a self-contained application that did exactly that.
It was eye-opening. First, because it was trivial to create: I literally asked Claude to make an app that tracked planes, it asked a few questions and, in a couple of minutes, had something running on my machine.
But my eyes were also opened by the patterns is found in the overhead traffic. For example, a lot of flights from Boston pass overhead on their way to cities in the South. I also found that there were a lot of private planes over my house that were headed to or from the airport near White Plains. One flight from Montreal to Florida passed overhead as it sought to avoid a storm.