My last five years of work

What will you do when AI comes for you? Avital Balwit and the coming AI disruption.

I’ve been thinking about what I will do when the AI finally comes for me.

Perhaps I will be at my house in the mountains, working on the renovation, running plumbing pipes along stone walls. Or maybe refurbishing some old wood table or dresser coated in dust from decades of abandonment. Possibly I will be looking after the soil in the garden, or working on this newsletter.

And I will invite friends for cheap wine and homemade pizza from the oven. We will talk about our latest climbing projects and the state of the world and what can be done, and whether our place in all of that is small, or significant, or somewhere in between.

When I have questions about the renovation, I will consult the AI. I will point the camera at the stonework and ask it to give me feedback on new masonry skills. It will tell me which bugs in the soil are good and which weeds are bad. Perhaps I will ask it for a training plan, specific to my age, weight, body type, and psychological predilections for discipline (or lack of it).

After a while, I won’t even have to train it much. The AI will just know

Climbing in Montserrat, Catalunya.

---

I. Knowledge workers in denial

Last week,

Paul Millerd

shared a post by Avital Balwit, the chief of staff at Anthropic (the company behind Claude, one of the large language AI models), about her last five years of work, and what will happen when AI comes for the jobs:

I am 25. These next three years might be the last few years that I work. I am not ill, nor am I becoming a stay-at-home mom, nor have I been so financially fortunate to be on the brink of voluntary retirement. I stand at the edge of a technological development that seems likely, should it arrive, to end employment as I know it. 

Employment as Balwit knows it means knowledge work. That which can be done on a computer. Or, remotely.

I am in agreement with that much of her analysis. This kind of work that most office workers and especially digital nomads do is going away. Maybe not all of it, but much of it. There are about 100 million knowledge workers in the U.S.—if I had to wager a guess, I’d put money on between 50 and 70 percent of them having to find a new way to earn a living, and soon.

You are of course free to disagree. But like Balwit, I think a great many knowledge workers are simply still in denial:

They grasp at the ever diminishing number of places where such models still struggle, rather than noticing the ever-growing range of tasks where they have reached or passed human level. Many will point out that AI systems are not yet writing award-winning books, let alone patenting inventions. But most of us also don’t do these things.

The economically and politically relevant comparison on most tasks is not whether the language model is better than the best human, it is whether they are better than the human who would otherwise do that task… The shared goal of the field of artificial intelligence is to create a system that can do anything. I expect us to soon reach it. If I’m right, how should we think about the coming obsolescence of work?

As they say, change happens slowly at first, then all at once.

That all-at-once part? It’s right now.

But it’s not all of work that will become obsolete—here I diverge from Balwit, who at no point in her piece acknowledges that there are jobs, such as those done with one’s hands, that AI cannot and likely will never replace.

Still—the world has created Artificial General Intelligence. Few have grasped the implications. Not that I know what all the implications will be. No one does, not even the creators.

But I do know that a $20/month, off-the-shelf AI tool is at this current moment better at doing certain work than hundreds of millions of currently employed humans. These workers are as of now obsolete. We have only to wait for everyone else to realize it, and act (or be acted upon).

This is where civilization is right now.

---

II. What will become of us

Geoffrey Hinton, often referred to as the “godfather of artificial intelligence,” recently told the BBC that the world will need Universal Basic Income, and that he was “very worried about AI taking lots of mundane jobs”.

I’m genuinely not sure how worried we should be at an escape from mundaneness. As the Buddhist might say: Before enlightenment, chop wood, carry water. After enlightenment, chop wood, carry water.

Balwit, for her part, looks at the data on happiness and unemployment and envisions a lovely future where we all go about like landed gentry:

Perhaps they did some minor administration of their tenants, some dabbled in politics or were dragged into military projects, but compared to most formal workers they seem to have worked relatively few hours. They filled the remainder of their time with intricate social rituals like balls and parties, hobbies like hunting, studying literature, and philosophy, producing and consuming art, writing letters, and spending time with friends and family.

It sounds nice. Personally, I can’t deny some aspirations to be landed gentry myself. After AI comes for me, I expect I will rent properties to rock climbers, pursue my hobbies, and also work relatively few hours.

But here I think it is Balwit who is in some version of denial. We need not fear an escape from mundanity, but we should fear the economic, cultural, and political turmoil which has so clearly already arrived, and which I think is likely to get much worse.

First, there is no version of the world I know that establishes a Universal Basic Income such that millions of workers may transition to a life of leisure, no matter how often the techno-utopians ensconced in their ivory programming towers may argue for one.

My feeling is that AI should be treated much like a public utility, which I think is a best-case scenario.

But I’m not naive as to the politics of our moment.

Far from a future of sensible governance and techno-utopian dreams, I think it’s more likely we are all in for decades of painful, violent disruption. Think Years and Years, the HBO show that tracked a crumbling world over decades, even as wonderous technological progress marched on.

Actually, the best reference for how to understand the current moment I think comes from Niall Ferguson, who likened our time in history to the technology-driven tumult of the 1600s, a “time of troubles”:

In Europe it culminated in the Thirty Years War… reducing Germany to one vast charnel house. In the British Isles, it was a time of internecine conflict — known variously as the Great Rebellion, the Puritan Revolution, the English Civil War, the English Revolution, or the Wars of the Three Kingdoms… In France, Cardinal Richelieu battled the Protestant Huguenots at home and the Catholic Emperor Ferdinand abroad. In China, the same period saw the fall of the Ming dynasty, its control of northern China lost to the Jurched leader Nurhaci, its fate sealed in 1644, when the rebel Li Zicheng captured Beijing and the last Ming emperor hanged himself.

By the way: the Thirty Years War killed eight million people, or roughly 10 percent of Europe’s population at the time. By contrast, WWI killed 3-4 percent of Europe’s population.

I realize I risk sounding quite alarmist.

And obviously, I hope I’m wrong, and I hope Ferguson is too. But I think his comparison is apt:

In our time, as has often been remarked, the internet has played the role of the printing press. The drastically reduced cost of reproducing text and images broke the church’s monopoly on both, just as the internet has enabled everybody who wishes to express and disseminate an opinion to do so — no matter how idiotic or illiterate.

In the 17th century, a certain amount of what was printed contributed to what ultimately became a Scientific Revolution. But a great deal more was devoted to alchemy, astrology, witch-finding and obscure arguments about the difference between transubstantiation and consubstantiation — in short, to superstition… In our time, I have long marveled at how much more attention is paid on social media to conspiracy theories than to theories based on evidence.

What will happen when billions of people in every country on Earth realize they cannot trust a single story, photo, or video they see on the Internet?

Maybe the liberal governments of the world get their act together. Maybe humanity won’t descend into a downward spiral of distrust and disillusionment.

But of all the naive opinions I’ve seen lately, that one seems right up there.

---

III. On antifragility

Some years ago I attended an Interintellect salon on Nassim Taleb’s book, Antifragile (a book I amazingly happened to read and write about just weeks before the Covid-19 pandemic began).

As with all Interintellect salons, the hosts started with a question: what was the most antifragile thing about each of us?

There I was with a bunch of software engineers—Taleb tends to attract those types—and to a person, all of them said it was their coding skills, their “engineering” brain, or their ability to solve problems (of the kind one sees online of course).

These software engineers saw themselves as antifragile probably because the economy had been handsomely rewarding coding skills for at least a decade, with no end in sight.

Then it got to me: “My house,” I said, to some bemusement.

Something that is antifragile is distinct from something robust, as Taleb explains. A robust thing withstands disorder. An antifragile thing gains from disorder.

My house fit the definition perfectly. When the pandemic hit, my place in the mountains of central New Hampshire dramatically increased in value. The acre of land, the ability to grow food, the well, the proximity to nature, the presence of a tight-knit community—all things that became more valuable, the more disorder there was in the world (my house is worth nearly triple what I paid for it in 2019).

Meanwhile, the latest AI is pretty good at doing the work of a software engineer.

Figuring out how to be antifragile isn’t easy. But there are gains to be had among all this disorder. I hate to sound like a prepper, but I do recommend buying land in climate-resilient parts of the world. I recommend learning skills with your hands and knowing how to grow food. I recommend making your life flexible, nimble, and simple because all these qualities lead to resiliency.

Beyond antifragility, I recommend reading (or re-reading) your Seneca:

Set aside a certain number of days, during which you shall be content with the scantiest and cheapest fare, with coarse and rough dress, saying to yourself the while: “Is this the condition that I feared?”

I wish the hundred million knowledge workers would actually follow this advice. But I know they won’t.

Anyway, when the AI comes for me… I’ll be chopping wood, climbing with friends, and building stuff.

It’ll be fine.