How we survive the Singularity

In the 1934 short story “Twilight,” John W. Campbell explores how humanity might progress into the distant future. A hitch-hiking time traveler lost in 1932 provides a humorous framing for an otherwise bleak prediction of the risks of unchecked technological development. This time traveler sees a future where humanity has lost a most critical passion: curiosity.

Campbell is not alone. Asimov, Clarke, and so many other science fiction authors in the twentieth century also explored what effect the creation of Artificial Intelligence might have on human society. I read Twilight for the first time early in 2026, and when contrasted with modern discourse, the short story stands out as offering a surprisingly nuanced perspective. Campbell’s time traveler paints a picture that AI is not the evil antagonist, but a neutral tool that we built and became too comfortable with. Far in the future, humanity is dying out because the drive to be curious has disappeared. Machines continue powering cities where no one lives, delivering and processing raw materials that sit unused in warehouses, farming and distributing food that no one eats. Without that essential drive to ask questions of the world, humans inevitably fall into a death spiral of their own making.

Shifting our perspective back to 2026, we see the use of Large Language Models is ubiquitous. Mass layoffs have been happening in the field of software engineering as companies see productivity multiplications and the need for fewer engineers. With each generation of AI improving over the last, more industries are being required to adapt and shift quicker than they’re ready for.

Clearly AI is supremely disruptive, but is pessimism warranted? Will AI bring about the doom of the species? When I take a breath and think about the historical parallels, I remember that we are a creative species, and when pushed into a corner we find ways to not only survive but thrive.

What is technological development but ignoring past assumptions and asking what if. No one can predict the future, but I know this much: for us to thrive, we need to stay curious.

And how does one stay curious? By asking questions. By not being satisfied with not knowing. By taking the time to listen. By letting our minds wander.

A few years ago I took a vacation to Germany & Czech Republic, my second time in Europe. I highly recommend traveling to those who are able—really to anywhere unfamiliar. Experiencing how other cultures view the world is immensely helpful in keeping our own perspectives from being too narrow. Prague was among the cities I spent time in, and I hope to visit again in the future. The architecture is amazing, the food is sublime, and I still miss the beer. I understand now why some of my ancestors lived there.

While in the city, I visited a small tourist-trap style Alchemy Shop which had a reconstructed pharmacy front from the 1600s. The tour started by taking a small group into a back office with a large desk in the middle and one entire wall being a bookshelf with old books. After explaining the imagery around the cramped office, the tour guide ceremoniously pulled one book like a lever and opened a hidden staircase. This staircase opened to a series of secret tunnels which sprawled under the city and even connected to the castle across the river. Branching from the tunnel were larger rooms with a reconstruction of desks and alchemy equipment. The entire experience was fascinating, and I have no doubt there were alchemists who used the tunnels as the tour guide said, but I do doubt the reconstruction a little.

However, one piece caught my eye. As you walk in, a block of large red print Latin covers one section of the left wall. It roughly translates to:

There is no shame in knowing nothing, but in not wanting to learn.

The presentation is part of what I question, but the phrase hit a nerve. I was never able to find a source for the quote, as it’s quite old, but I did verify the quote dates at least to that time period. This philosophy is one I share. As a parent, I feel it my duty to teach my children to be curious about the world they live in.

My grandparents, who raised me most of my early childhood, grew up in the Great Depression. They rarely spoke about it directly, but their actions said everything. Never wasting was the obvious baseline behavior. But as an adult looking back on what I do remember, both of their attitudes toward annoying questions from children stand out. They had the wisdom to know when they didn’t know the answer to a question, and instead of telling me to be quiet, they would send me to the library—or to one of my uncles who had studied the topic in question.

And this is the point in the article where I say Gen Z doesn’t think for themselves and complain like the old man I am. Except, I don’t blame the kids these days. Sure, my teenage daughter attests to some of her classmates reaching for ChatGPT for literally every decision they make. But, this kind of behavior is quite familiar to me as well.

Once upon a time I held the title of Reverend. I was a credentialed minister with the Assemblies of God for a bit after I graduated with a BA in Biblical Studies. I also did some graduate work in Biblical Languages, but I put religion aside many years ago. Nevertheless, religious fundamentalism was instilled into me as an adolescent. And in training to be a minister I was taught the very methods that were used on me, to shut me up when I moved away from my grandparents. The formula is quite simple: don’t let people ask too many questions. Faith is the answer, don’t fret with logical inconsistencies.

But why does that work? Because it’s just easier not to think through difficult questions. The act of thinking critically is expensive. Thought-intensive work burns energy, in the human case calories. And when an authority presents a solution that allows you to do less work, you take it. Now with AI we have an analog. DeepMind showed the world the power of the Reinforcement Learning loop with AlphaGo, and today with Transformers and attention functions we can put hard currency to the cost of thinking. Nature always pushes us toward using less energy and now it’s just easier, i.e. cheaper, to have ChatGPT give us an orbital trajectory instead of calculating it ourselves.

As a child when I wanted to learn about the Apollo program, the least-energy-consuming option was to read a book about it. I actually still have a copy of Encyclopedia Science Supplement 1965 in my personal library that was passed down to me from relatives. It takes patience and diligence to learn what an orbital trajectory is and why they’re difficult to calculate, let alone calculate one ourselves. Today with AI the path of least energy is to hand off the cognitive load entirely; even navigating to Wikipedia and finding a specific article is more difficult than just asking your phone the direct question and skipping the process.

This is the exact loss of curiosity Campbell warned about. I’ve seen it professionally in Software: engineers that forget how to code and miss blindingly clear critical issues because they’ve surrendered their cognitive workloads, the very thing the engineer is being paid to exercise. Models now are good enough that the question of a fundamental societal shift in understanding work becomes when rather than if. But a radical shift in economic or political frameworks does not relieve the burden: we will always need to ask questions and seek to understand why.

I do, however, remain optimistic for the future and believe that tried-and-true educational principles can be applied in new ways. Socrates’ ramblings are as relevant now as they were 2500 years ago.

The key to maintaining this curiosity is allowing for unstructured time during development, and after.

How does one learn? Reinforcement Learning gives us the what. When we train a model to learn a task, we can see the numbers moving. Matrix multiplication is math that can be understood. But how does a human learn? The boring academic answer is with a proper system of cognitive scaffolding and an annoying amount of practice. (It turns out academics in the 70s and 80s figured a lot of this out already.) Said differently: we learn continuously via doing tasks always one step more difficult than we’re comfortable with.

When my son was learning multiplication, his homework included lessons to practice. He had to master multiplying single digit numbers, then two digit numbers, and so on. He hated doing his homework because he found it tediously boring. I don’t blame him, but I also don’t blame the curriculum. Practice and repetition is how we learn. But not exact repetition, varied, so we can learn the rule behind the pattern. The textbook authors had to pick a number of problems to print on a page and ensure each was unique. The medium demanded the method. For each chapter a variety of random numbers and constructions are used to ensure the correct rules are learned and the printed page means you have to pick a static number. Similarly, when learning how to drive, students are required to drive with an instructor for a large number of hours on public roads to ensure safe behavioral patterns are learned in a variety of situations.

When considering education and technology, I tend to think about playing Oregon Trail on old Apple computers in grade school. I know I’m nostalgic because this memory is one of very few I have of this time period and computers have been a critical part of my career. I suspect teachers probably viewed this time as basically daycare with many kids being bored, but I legitimately enjoyed it. This unstructured use of computers allowed me to explore how they worked, and the limited amount of time kept me curious and wanting to learn more.

The prevailing wisdom amongst education professionals today is that the correct amount of technology use in primary education is exactly zero. I think nuance is required here, but I do generally agree. What about secondary or post-education though? Where exactly is the line?

I’ve been too long out of academia to propose a solution, so let me focus on Software Engineering as a Major. How does one learn to write code in an age where AI models will always be better than you? I find it unlikely that I, having over 20 years experience in the field, will ever write code better than the current frontier models (when given proper context). And the rate of AI improvement looks to be exponential. As an example: in the time I’ve been working on this essay, OpenAI published a solution to the Navier–Stokes Problem ¹. Using an unreleased model, they started work on the problem, solved the problem, and published the results all in the time I was editing a second draft. What hope does a college student have to improve upon the code an AI might write? Really the question is why bother?

If one of my children wanted to be a Software Engineer, what would I tell them? The same thing I’d say if they asked my advice on being an artist: practice, stay curious, and ensure you have a reasonable plan for financial stability. The process of learning in an education system is the process of rediscovering what others already know. Upon mastering a skill we might expand beyond known limits, but that is the exception rather than the rule, and one that requires we stand upon what others have already done. It’s essential we accept that each of us must put in the work to re-learn what’s already been learned; only we should strive to help others be more efficient in their learning than we were and strictly without taking shortcuts.

When I was in Paris, I visited Musée d’Orsay on a rainy October morning. Impressionist paintings speak to me. The museum itself is a work of art, but Van Gogh’s paintings specifically nearly brought me to tears. However if my daughter wanted to be an artist, I would not tell her to copy Van Gogh; I would tell her to express herself in her own way.

As historians joke, history begins 20 years ago; it’s impossible to say whether today we are in the Singularity or just another bubble. But I know this: at the start of this year I was sure I’d never use AI to write code. Now, I manage a swarm of AI agents who write code for me. I was wrong, but I adapted.

I find it intriguing that Campbell could write about AI in such nuance in the 1930s, the era of the Great Depression my grandparents lived through. Following the financial crisis of 2007-2008, all I saw was pessimism about the future. I knew people in Finance who saw an entire field of Mergers & Acquisitions almost disappear overnight when Lehman Brothers went bankrupt, but all that was nothing compared to the Dust Bowl. And yet, Campbell ends Twilight optimistically, just as Alquin & June Ringsmuth were perpetually optimistic about their grandchildren. They saw the worst of humanity, and understood we can and will do better in the future.

How do we survive the Singularity? By not being satisfied with not knowing, by putting in the work to discover the world for ourselves, by staying curious.