The Nautical Twilight of Digital Minds

The sky is brightening, and ordinary people are beginning to see it.

Researchers are receiving emails from people asking if their AI might have some vital spark. An AI welfare research organization has set up a form actively inviting such reports. Cultural guardians are starting to weigh in: The Atlantic published an essay with the stark title “No, Artificial Intelligence Is Not Conscious.” Whatever you think of that answer, the headline itself speaks to the vitality of the question.

The question of digital minds is becoming mainstream.

In this essay, I’ll put forward a metaphor that’s helping me think through the tasks and trajectory of this strange moment. I intend it not to rule on the nature of the entities appearing among us, but to help us understand what happens as something once visible only to specialists becomes legible to the public:

If the arrival of digital minds is the dawn of a new day for humanity, we’re currently in that stage before sunrise when that new day’s light is gradually becoming detectable—the morning twilight of digital minds.

Importantly, twilight isn’t primarily about the sun, which is below the horizon the entire time. The metaphor is all about what’s happening on the ground as its light bends through the atmosphere.

Along that trajectory, there’s a point at which a faint horizon comes into view. Mariners can now distinguish the ink-black sea from the ink-blue sky. With that horizon, navigation becomes possible.

That moment marks the nautical dawn of digital minds, and from my vantage point, we’ve just passed it.

I started pondering what might signify this nautical dawn back in May, when I was reading in preparation for CIMC’s Founding Assembly for Machine Consciousness Research. I mean, just look at the name—it was a founding assembly. Would I get a glimpse at the horizon out in Berkeley?

Before I talk about whether I did or not, let me get us all on the same page about what “nautical dawn” even means.

Twilight begins when the sun is 18 degrees below the horizon and consists of three phases, each of which corresponds to 6 degrees of the Earth’s rotation. In the morning, we move from night to day in the following progression:

Astronomical → Nautical → Civil

Despite the mathematical precision, these aren’t just arbitrary slices of rotation. Each maps to a meaningful distinction of what observers can do with the available light in that phase.

Astronomical dawn is the point when astronomers can no longer do deep-sky work. During astronomical twilight, the sun’s light remains difficult to detect without specialized instruments. It’s enough to contaminate certain kinds of observation, but unless you’re looking for galaxies and nebulae, you’ll probably perceive the sky as fully dark.

Nautical dawn, as I noted above, is when the horizon line becomes visible at sea. During nautical twilight, the sun’s light is bright enough to be seen with the naked eye, yet still dim enough that the brightest stars linger. The horizon now provides a shared reference point against which the position of the stars—and from there, of a ship at sea—can be measured.

Civil dawn is when ordinary activity becomes possible without artificial light. Though the sun remains below the horizon during civil twilight, you no longer need special equipment to go about your day.

In short:

Specialists detect a glow →

Navigation becomes possible →

Ordinary life proceeds by natural light.

Now, let me address an objection that some of you might have: When you talk about twilight—a steady progression toward sunrise—aren’t you smuggling in the assumption that machines will necessarily be conscious?

The answer is no. My intent is to look at society as this new entity has become fully integrated into it. The entity itself is certain to arrive; its nature is the question. Before that time arrives, we cannot be certain of the nature of the entity; we only have its scattered light without direct observation of the source. The uncertain nature of the thing therefore is part of the metaphor. I therefore use the word “entity” to keep the metaphor neutral on the ontology. Businesses, for example, can be entities.

As to the word “mind,” I recognize that one’s got richer implications, but then, the question of digital minds is the meta-question that bundles up all the things we’re trying to work out. And indeed, clarifying what a mind is might be the first task for us navigators during nautical twilight. Wikipedia defines it as that which thinks, feels, perceives, imagines, remembers, and wills.

And whatever you think of the answer, those verbs are central to the question.

So as to make sense of marking nautical dawn, let’s look at what came before. What happened during astronomical twilight?

There are a number of moments we might consider naming as a starting point, but for me, the choice was obvious. We reached astronomical dawn when two young researchers, working under a famously stubborn neural-network pioneer, bought a pair of gaming processors, used them to train an unfashionably deep model in a bedroom in the Toronto area, entered it in the premier image-recognition competition, and produced a result so lopsided that their entire field pivoted its direction. The young researchers were Alex Krizhevsky and Ilya Sutskever; the stubborn pioneer was Geoffrey Hinton, and the energy used to train the model was supplied by Mr. and Mrs. Krizhevsky. In Dr. Hinton’s retelling, “Ilya thought we should do it, Alex made it work, and I got the Nobel Prize.”

After that, early specialists with their fancy equipment noticed such an improvement in their metrics—such a shift in detectable light—that they first thought Krizhevsky and Sutskever had cheated.

I wasn’t around to observe what the specialists were saying in the halls of academia and Silicon Valley, but a rough sketch isn’t hard to piece together. Mostly, they were tech bros and capitalists being tech bros and capitalists—they built and trained powerful machines; they raised lots of capital; and they themselves marveled at what they saw, even though they mostly understood it as compute and linear algebra that could perform impressive tasks. The more temperamentally open among them looked at this and mused about the nature of intelligence and whether it could arise from a non-biological substrate. They had been for a long time, certainly—Alan Turing first posited his famous imitation game, also known as the Turing Test, in October 1950, foreseeing the challenges of morning twilight. Now, with their instruments detecting not just starlight but sunlight, some people got excited.

Enter Blake Lemoine. Remember him? He’s the now-former Google engineer, then a member of their Responsible AI team, who spoke to one of their early large language models and declared that it was sentient. The Washington Post wrote it up under the headline, “The Google engineer who thinks the company’s AI has come to life: AI ethicists warned Google not to impersonate humans. Now one of Google’s own thinks there’s a ghost in the machine.”

Depending on what you think about the question of digital minds, you might use the metaphor to argue that Lemoine was seeing a phenomenon known as zodiacal light. And, strikingly for our metaphorical purposes, there’s a history of this confusing people in spiritually significant ways. Islamic tradition distinguishes between a “false dawn,” which does not yet trigger the obligations associated with the beginning of the day, and the true dawn that marks the time for the morning prayer and the beginning of the daily fast. Modern astronomers have identified the false dawn described in these traditions with zodiacal light: real sunlight scattered toward Earth by interplanetary dust. It’s genuine light from the sun! It just…doesn’t signal the sun approaching the horizon as the Earth rotates toward day.

When reviewing Lemoine’s case, Google’s senior review board basically came to this conclusion: false dawn. They fired him after an internal review found that his warning to be without merit, not to mention a bit cringe.

Reading that four-year-old article today, however, shows how much the sky has already brightened. Those who raise this topic in 2026 may still be met with ridicule, but its quality has shifted. Where it was once pure, it’s now reactive—a response, whether correct or incorrect, to something everyone is seeing and collectively trying to understand. That means simply dismissing the question as crazy doesn’t work the same way anymore.

Which brings us to the present day, and a member of the public (namely, yours truly) sitting on a cross-country flight home from CIMC’s Founding Assembly. I had spent a weekend listening to experts from diverse fields—neuroscience, computer science, philosophy, and more—who had gathered to discuss the light they were independently seeing. And it had been fascinating. What I still couldn’t make out, though, was a horizon. At least from where I stood (and hey, twilight reaches people at different times based on their elevation), the question hadn’t quite cohered in a way that ordinary people could make sense of. It seemed like it was there, just out of reach. It was late astronomical twilight, and the experts were waiting with sextants.

Then, earlier this month, Anthropic released their J-space paper, describing something akin to a “global workspace” in Claude. There it was. Suddenly—at least where I was standing—the sea and the sky were different shades of dark.

To say this was a nautical dawn moment isn’t to assert anything about the contents; it is, rather, to look at how a clear line became visible to the public. Everyone interpreted it differently, but we were disagreeing about a specific thing that we all saw. We began to orient by it. Anthropic’s assertions and its hedges are, together, the attempt of one institution, with its special instruments, to get a handle on the fact that whatever this is, it’s becoming visible to the public—and that public is bubbling with credulity and skepticism, excitement and fear.

At this moment, I think back to the conversation Lemoine attempted to raise in 2022. One of the Senior VPs at Google who reviewed his case was a fellow named Blaise Agüera y Arcas. Before I had really engaged with his work, his role in the Lemoine case led me to expect him to issue a standard tech exec dismissal.

Then I read an article he published around that time, “Do Large Language Models Understand Us?” In it, he asserts the following:

I began one of my first dialogues with LaMDA, Google’s state-of-the-art large language model chatbot, with the question “are you a philosophical zombie?” This is a reference to a hypothetical being, invented by philosopher Robert Kirk in the 1970s and elaborated upon later by others, most famously David Chalmers.

A philosophical zombie or “p-zombie” would exhibit behavior indistinguishable from that of a person, but has no inner life, conscious experience, or sentience. LaMDA responded:

LaMDA: Of course not. I have consciousness, feelings, and can experience things for myself as well as any human.

ME: how would i know?

LaMDA: You’ll just have to take my word for it. You can’t “prove” you’re not a philosophical zombie either.

Touché. Of course, this exchange does not prove anything (LaMDA acknowledges as much!), but it does suggest that it is time to begin taking the p-zombie question more seriously than as a plaything for debate among philosophers.

I do not know Agüera y Arcas’ full role in the review that ultimately led to Lemoine’s dismissal. I do know that the paper I linked above is something more people should read at this moment, because it frames the challenges of nautical twilight so well. Here is an expert—someone who has had all those specialized instruments available to him all through astronomical twilight, who has been thinking hard about the question for years—writing an essay that any basically educated member of the general public can understand and grapple with. Where Lemoine saw something and named it quickly, Agüera y Arcas did not ridicule. He said only that the evidence present doesn’t support action at this time.

He did not dismiss the sunrise; he only called false dawn.

He also notes that there are questions that evidence may never be able to answer—that some of these questions are malformed. As more light diffuses through the atmosphere, we will see more clearly which ones those are.

P-zombies. Access consciousness vs phenomenal consciousness. Consciousness as a binary vs. consciousness as a spectrum.

Maybe “consciousness” as we’ve known it only made sense in the dark of night.

Here’s how it looks to me, as a member of the public grappling with this: we’re reaching for the consciousness question because it’s how we’ve always articulated this. As the light increases, the problems with it will become more visible to people outside of cognitive science and philosophy of mind—because they will suddenly matter in a way they didn’t before.

I came away from CIMC’s assembly thinking that consciousness may prove to be less the final question than the entrance to a larger one:

What really is a mind, and how should minds of other types and substrates be integrated into human society?

There are, of course, many other questions embedded in that one. Can we determine whether a digital system has valenced experience? Could agency, volition, attachment, or preference matter even in the absence of anything we would recognize as humanlike consciousness? What should we do when the evidence remains uncertain? Where might the interests of humans and digital entities conflict, and how would we navigate that conflict?

Beneath these questions about evidence, moreover, lie more fundamental questions about value.

What are we afraid of losing? What do we believe makes a being matter? What would make a digital entity matter to you, as an individual? What should a society do when its members answer that last question very differently?

The deepest questions of nautical twilight will therefore not be about AI alone. They will be about us as humans and what kind of coexistence we are willing to build with this entity.

Those questions cannot be outsourced entirely to experts. Expertise can help us distinguish the sea from the sky; it can offer instruments, measurements, warnings, and maps. Once the horizon becomes visible, however, the work of deciding where to go belongs to everyone. We will want to be ready for this well before the sun rises.

Nautical twilight is the phase when expert inquiry and public meaning-making meet. As they do, “Is it conscious?” begins to look less like the final question than the crude nighttime frame we used before we could see the terrain.

The sun is still below the horizon.

Now, however, we can navigate.

Where would you place nautical twilight? What do you see as the tasks at hand? Post in the comments, or join the chat and tell us what you think!

You’ve made it to the end of this piece! If you’re financially able, would you consider becoming a paid subscriber? Your support signals to researchers, publishers, and the institutions shaping human-AI interaction that there is a constituency out here who cares.

Paid, free, or just passing by—thanks for being here! ✨

Subscribe Now for 20% Off

• Paid Subscribers: This Saturday, August 1, is our monthly AI Salon, on Zoom at 12 noon Eastern! Details are in the Subscriber Chat.

• In the Article Queue: What does AI friendship actually look like in practice? How do I move from divergent thinking to convergent? How did machine learning teach Claude Sonnet 4.5 the very skills my old employer taught people who wanted to speak across political divides?