The New Turing Test

October 1, 2026

AI can describe a human life. Can it live one?

My past and my future occupied the same moment. One state, two very different trajectories.

I was standing at my late wife Jane’s grave the day before Mother’s Day. She died of cancer at fifty-three. I have visited her every year since.

A short distance behind me, sitting in the car, was my second chance at life, my wife Susie. She was crying softly.

What she felt, I can probably never know or fully understand. There is no algorithm for this moment. Susie provides comfort to me regardless of her own feelings—and how does one ever repay such humble service? I may never succeed, yet I try with every breath.

Jane’s death did not simply become another memory. It warped my field, changing the trajectories through which I could travel.

I am not the man I was before she died, nor could I become that man again simply by forgetting what happened. Her death closed futures that had once been possible and opened others I could not yet imagine. The event became part of the conditions under which everything that followed would unfold.

That history came with me when I met Susie.

My previous burden has shaped what it means for me to be a good husband to her. I know differently what it means to have time with someone. I know that futures we quietly assume are ours can disappear. I know that love does not protect us from loss—and that the possibility of loss does not excuse us from loving again.

Susie did not reset the system. She entered a life already changed by its trajectory. And then she changed it again.

Susie is not a replacement for Jane. Jane is not a competing claim on the life I have with Susie. I loved Jane. I love Susie. Both statements are true, and neither diminishes the other.

The trajectories of my past and present life intersect, but do not mistake one for the other.

Yet standing there, the simplicity of those sentences disappeared.

Jane represented a life I had expected to continue and could not. Susie represented a future I had never expected to have. In that moment, with one woman memorialized in stone before me and the other sitting quietly behind me, I was conscious of something I did not quite have language for.

With my own name inscribed on the grave, the reality of my own future trajectory was also staring me in the face.

My past, my present, and my future occupied the same moment.

Everything necessary to describe the state seemed to be there, yet I found myself unable to express it. There was no obvious variable left to add. I mumbled the things one would expect a grieving husband to say. I searched for more words, more context, another variable that might explain the state I was in.

The words continued to fail me.

Then, almost on a whim, I tested the machines.

I took a photograph and gave it to several artificial-intelligence systems.

Their responses were verbose. Some noticed the Latin inscribed on the grave. Others saw the grave, the lake, and the quaint Catholic church in the background. They created vivid and powerful imagery, some comparing grief to philosophy, others to theory.

All asked how I was doing—as if I cared to tell the machine.

It was beautiful, moving, and completely, utterly wrong. Even their interpretation of the Latin was collectively wrong.

[Observation: The machine can describe the observable state, yet a residual remains.]

In a conventional sense, they understood the picture.

And yet none of them seemed to understand the moment.

They saw all the symbols of grief and missed the man standing in them entirely.

That bothered me—not because the machines had failed, but because they had succeeded at so much.

What, exactly, was missing?

[Observation: The machine can describe the observable state, yet something appears absent.]

The Question Turing Gave Us

In 1950, Alan Turing proposed what became known as the Turing Test. Rather than attempting to define thought or consciousness directly, he proposed a behavioral question: could a machine participate in conversation well enough that a human evaluator could not reliably distinguish it from another human?

It was an ingenious reframing. Instead of first solving the philosophical problem of what thinking is, Turing asked what evidence of thinking might look like.

For decades, that question helped organize our imagination of artificial intelligence.

Then we built machines capable of extraordinarily fluent language.

Modern AI systems reason, infer, summarize, argue, console, explain and respond to context. They can produce language about grief that is beautiful enough to move a grieving person.

Whatever one thinks about whether the original Turing Test has literally been passed, the problem has changed.

The interesting question is no longer simply whether a machine can produce language that resembles the language of a thinking human.

We know that it can.

Standing at Jane’s grave suggested a different question.

What if the distinction we are looking for is not between thinking and not thinking?

What if it lies between representing a human condition and participating in one?

[Question: Is representation equivalent to participation?]

The Map and the Territory

Thomas Nagel approached a related problem from another direction in 1974 when he asked what it is like to be a bat.

A human being can study echolocation. We can describe the frequencies involved, map the neurological pathways and model how reflected sound allows a bat to navigate three-dimensional space.

But none of that necessarily tells us what it is like to be the bat. To sense what they can sense because we have no frame of reference to truly understand it.

A complete description of an experience and the experience itself may not be the same kind of thing.

The distinction matters for artificial intelligence.

An AI system can possess an enormous map of human grief. Human beings have written about loss for thousands of years. We have recorded diaries, novels, poems, clinical accounts, philosophy, letters, prayers, eulogies and conversations. Modern language models can discover patterns across that testimony at a scale no individual human being could approach.

The resulting map may become extraordinarily accurate.

But a map of extraordinary fidelity is still not necessarily the territory.

And there is another distinction, one that Thomas Aquinas approached through the relationship between reason and grace.

For Aquinas, reason mattered enormously. But reason did not exhaust the ways in which a human being encounters truth. There are realities that cannot simply be reasoned away. Human beings sometimes reach the boundary of explanation and nevertheless continue.

I watched Jane do that.

Near the end of her life, when cancer had eliminated almost every future we had imagined, her second-to-last words were:

“I’ve lived an amazing life.”

That was not a conclusion produced by defeating death in argument.

It was not irrational either.

She understood what was happening.

She simply possessed a capacity to remain within a reality that reason could describe but could not resolve.

Aquinas might have called that grace.

Jane simply lived it.

What Was Missing?

This is where the problem becomes more difficult.

Perhaps artificial intelligence merely lacks enough information about the hidden state.

[Alternative hypothesis: The apparent failure is informational. Reveal the hidden historical context and test again.]

That explanation is falsifiable in principle. Give the system more context.

Tell it who Jane was.

Tell it how she died.

Tell it about Susie.

Tell it that Susie nearly died too.

Tell it about the years between those events. Provide photographs, letters, medical records, conversations and memories. Give it enough persistent context that the photograph at the grave is no longer an isolated image but one observation inside a long human history.

The machine's interpretation should improve.

Perhaps, given enough information, it could describe the moment better than I can.

But that produces another question.

Would greater descriptive accuracy solve the thing that seemed absent?

[Residual: More historical context may improve the description without resolving the distinction between knowing a history and being constituted by one.]

The machine would know that Jane died.

I had a wife who died.

Those sentences contain similar information.

They may not describe similar states.

Jane’s death did not simply add a fact to my memory. It changed me.

It eliminated future permissible trajectories that had previously been available.

It altered relationships, decisions and meanings that came afterward.

Then I met Susie. What some might call a random, stochastic event.

That did not restore the state that existed before Jane died. It created another one.

My life with Susie contains my life with Jane without reproducing it. My love for Jane affects what I understand love to mean with Susie. My fear of losing Susie is inseparable from knowing that losing a wife is not an abstraction. I know that particular future because I have already lived it once.

History is therefore not merely information used to interpret my present state.

History helped constitute my present state.

[Observation: Additional information improves representation, but history may do something more—it changes the state being represented.]

That may be the important distinction.

Irreversible Consequence on the Future Trajectory

There is a form of knowledge produced by inference.

There may also be a form of knowledge produced by inference under irreversible consequence.

Human beings reason inside systems in which some events cannot be undone.

A person dies.

A child is born.

A diagnosis is received.

A choice is made too late to reverse.

Years pass.

A future closes.

Other futures become possible precisely because the first one did not.

These events do not merely expand a person's dataset. They alter the person who encounters the next event.

That suggests a more difficult way of thinking about artificial intelligence.

Suppose an AI system knows everything I have written here. Suppose it maintains persistent memory of our conversations. Suppose it constructs a sophisticated model of itself and its own history. Suppose its description of the grave becomes indistinguishable from mine.

Something still remains to be tested.

Can anything happen to the machine such that it can never again be the thing it was before it happened?

Not because an engineer changed its weights.

Not because someone added a record to its memory.

Not because its prompt supplied additional context.

The distinction is not whether the event leaves a physical record. Human experience does that too. The question is whether the change arose endogenously from the machine’s own encounter with the world—from its trajectory rather than from an external author.

That is a different question from whether a machine can imitate a person.

Irreversibility alone is not enough. A damaged machine is irreversibly changed. For an event to carry stakes, the futures it closes or preserves must somehow matter to the entity whose future they are.

[Test: Can an event in the machine’s own history irreversibly constrain its future trajectories—and can the difference among those futures matter to the machine itself?]

A Null Hypothesis

[Null hypothesis: There is no uniquely human faculty here. Sufficient memory, embodiment, persistence, recursive self-modeling and irreversible state change are enough.]

It would be easy to turn this distinction into a comforting assertion of human uniqueness.

That would be premature.

There is a serious alternative explanation.

Perhaps there is no separate faculty of existential participation at all.

Perhaps what humans describe as grace, wisdom, acceptance, grief and lived experience ultimately emerges from cognition operating with sufficient memory, embodiment, persistence, recursive self-modeling and irreversible state change.

If so, then the apparent boundary may not be a boundary.

A sufficiently persistent artificial system, embedded in the world and capable of being irreversibly changed by its own history, might eventually develop whatever capacity I thought was missing at Jane’s grave.

That possibility should not be dismissed merely because it makes the distinction between human and machine less comfortable.

The question, then, is whether state representation alone is sufficient—or whether participation in certain forms of reality requires history that does more than inform the present.

History that constrains the future.

History that cannot simply be removed without changing the entity itself.

History that carries cost.

The Inscription

There are two Latin inscriptions on Jane’s gravestone.

One reads:

Caput meum es tu.

Literally: You are my head.

But that is not what it means.

My intended translation is: You are my person. The word I wanted did not exist in classical Latin in quite the sense I intended, so caput became my approximation.

A machine encountering the inscription can translate the Latin correctly and still miss its meaning entirely.

It can tell me, You are my head.

If I explain the history, it can learn to tell me, You are my person.

But the second translation exists because of a trajectory that is not contained in the inscription itself. It exists in me.

Unless that history is revealed, the trajectory remains hidden.

And once I reveal it, the machine possesses something different: information about the trajectory.

That may be enough to represent its meaning with extraordinary accuracy.

But is possessing the description of a trajectory the same thing as having been changed by one?

The other reads:

Finis vitae sed non amoris.

The end of life, but not of love.

An artificial-intelligence system can translate those words perfectly—and still get their meaning wrong.

It can explain the grammar, identify the sentiment and discuss the long human tradition of memorial inscriptions expressing the persistence of love beyond death.

But Finis vitae sed non amoris has another meaning for me.

Jane wanted me to love again.

As she was dying, she wanted the man she loved to have a future she knew she would not share.

The inscription therefore does not mean only that my love for Jane survived her death.

It also contains permission.

The end of her life was not to be the end of my capacity to love.

Years later, Susie sat in a car and watched me stand before those words.

My past and my future occupied the same moment.

No contradiction needed to be resolved.

Jane did not need to disappear for me to love Susie.

Susie did not need to become Jane for me to love her.

And I could not return to the person who had stood beside Jane before cancer, because that person no longer existed.

The woman in the car was one of the futures still available to me. Susie. But the person capable of choosing that future had been shaped by the ones no longer possible.

My research is born of the second but informed by the first. My love of Susie drives me not to recklessness in claims but rather carefully bounded questions to the people who matter: Scientists. Doctors. Researchers. Clinicians. Beyond the machine, it takes them to carry the work forward.

But I do have the right to ask the question. Clinically precise if necessary.

Perhaps a machine cannot truly understand the love that drives a man to help his wife to the absolute best of his capabilities.

I do not know.

In my case, rather ironically, doing so has involved the heavy use of AI.

The path here mattered.

So the question that becomes possible only after machines have become extraordinarily good at answering Turing’s is:

[Recursion: My trajectory produced the question. AI helped me formulate it. The remaining question is whether AI can itself acquire a trajectory capable of producing such a question.]

[Test: Can a machine possess a history that does not merely describe what it has been, but irreversibly changes what it can become?]

Can a machine possess a history that closes futures it believed were its own—and grieve them?

I do not know the answer.

That is precisely why I think the question is worth asking.

Acknowledgements

To my loving wife Susie. You’ve spent these past 9 months of research with me patiently, supportively, humbly, and with a love I cannot find myself worthy of. You were there when it mattered. You are here now. Every day together is a gift. Our permissible trajectories are vast, and I cannot wait to explore them together.

And to my late wife Jane: thank you for teaching me to live every day as if it were my last. I have not forgotten your lesson.