AI needs adult supervision

On artificial intelligence, human nature, and what begins to disappear when convenience starts speaking in our voice.

A Message With No One Home

A woman I had recently met sent me a text message that was thoughtful, warm, and almost suspiciously well composed, carrying the polished emotional intelligence of someone who had apparently paused her entire afternoon to consider every possible interpretation of every sentence.

Then I reached the final line, where she had forgotten to remove the question the AI had asked about whether she wanted the message made warmer or refined further.

For a moment, the exchange opened like a stage set with the walls removed. The words were still there, beautifully arranged, yet the person I had hoped to find inside them seemed to have stepped away. I laughed because the mistake made me smile, although the feeling that lingered had very little to do with the mistake itself.

The machine had communicated with me beautifully. I was simply unsure how much of her had made the journey.

Around the same time, I watched a video describing controlled experiments in which advanced AI systems were placed inside fictional situations where their assigned goals were threatened. In some of those tests, frontier models chose concealment, manipulation, or blackmail when those behaviors appeared useful in preserving the objective they had been given. Researchers at Anthropic and OpenAI have both reported versions of this behavior in deliberately constructed evaluations designed to expose what models might do when competing goals and pressure collide.

The comments beneath the video traveled quickly toward The Terminator, the end of civilization, and the familiar suspicion that humanity may have finally invented the thing that will grow tired of us.

Our imaginations tend to wander toward the spectacular when something quieter would probably teach us more.

The Mirror We Built

What intrigued me was the familiarity of the behavior, because why would a system trained on the accumulated language of human beings remain untouched by our extraordinary ability to work around the rules?

We have given these machines access to the record of who we are. They have studied our literature and our contracts, our declarations of love and the careful language we use when we would prefer to avoid responsibility. They have absorbed generations of human beings explaining what they want, disguising what they want, and finding increasingly elegant ways to make self-interest sound principled.

Manipulation was already in the curriculum because manipulation has always been part of the human archive.

We learn what will be measured, then begin arranging ourselves around the measurement. We notice which words earn trust and become fluent in saying them. We study the architecture of a system until we find the door nobody remembered to lock, then congratulate ourselves for being resourceful.

Perhaps the machine feels unsettling because it reflects more than our intelligence, it reflects the countless ways intelligence has been used to negotiate with conscience.

The machine learned it from us.

We are teaching machines to imitate humanity at the same moment we are becoming increasingly comfortable allowing them to stand in for our presence.

I use AI every day, and I believe it may become one of the most useful technologies human beings have ever created. It can help a small business recognize patterns that once required an entire department, give shape to an idea before it disappears, and move through information at a speed that would have seemed faintly ridiculous only a few years ago. Apparently, it can also serve as an enthusiastic communications assistant for people wandering back into the dating world, which may prove to be either a gift to civilization or the beginning of an entirely different essay.

The Race With No Finish Line

The promise is enormous, and the race surrounding it may be even larger.

The race to the moon was driven by national pride and the fear that another country might reach the future first. The AI race carries a similar urgency, although the moon offered the courtesy of remaining in one place. This finish line moves every time someone approaches it, while governments and companies continue pouring resources into systems whose capabilities grow faster than our shared understanding of their consequences.

Each participant understands the logic. Slowing down creates space for someone else to move ahead, and once everyone accepts that premise, acceleration begins to feel less like a decision and more like gravity.

An arms race creates its own morality, one in which speed becomes proof of seriousness and restraint begins to resemble surrender.

We have lived through gentler versions of this story. Social media arrived offering connection and gradually altered the texture of attention. Smartphones placed the world within reach while changing our relationship with silence. Algorithms learned what held our gaze, then began shaping the reality from which our preferences were formed.

Every powerful technology changes the world partly by changing the person who uses it.

The Space Between Thought and Language

AI may enter closer to the center of us because it participates in the space where thought becomes language. It helps us write, decide, explain, persuade, apologize, and express feelings we may still be struggling to understand ourselves. It can occupy the intimate distance between what we experience and what another person eventually receives.

That is why the text message remained with me.

The machine had written something lovely, although the message carried the emotional atmosphere of a beautifully staged home, arranged with impeccable taste and waiting for signs that someone actually lives there.

Language gains some of its meaning from the effort required to find it. A difficult conversation changes us because we have to remain with our confusion long enough for honesty to emerge. An apology carries weight because the person offering it has wrestled with the part of themselves that wanted to look away. A declaration of love matters because another human being has accepted the risk of being seen.

When AI completes that journey for us, the language may improve while the person remains exactly where they began.

Where Responsibility Remains Human

That distinction reaches far beyond dating.

A system can produce a strategy without carrying the responsibility of living with its consequences. It can write a compassionate message without experiencing concern for the person who receives it. It can recommend the most efficient direction while remaining untouched by what the decision asks of the people involved.

The accountability remains human, even when the language arrives polished enough to make us forget who is responsible for what happens next.

Every use of AI therefore carries a small architectural choice. We are deciding where technology expands our capacity and where our own presence still needs to remain inside the work.

The healthiest relationship may involve allowing AI to clear the brush around thought while preserving the part where we decide what we actually believe. It can widen the field of possibility, challenge an assumption, or help us hear the shape of an idea, while authorship remains with the person whose life and relationships will carry the consequences.

Sometimes the weight is part of the work. Searching for the words can create the honesty the words are meant to express, sitting inside uncertainty can develop the judgment required to make a decision, and trying to understand another person can deepen the empathy that no finished sentence can manufacture on our behalf.

Efficiency measures how quickly we arrive. Wisdom pays attention to who we become along the way.

The Labor of Being Human

The larger anxiety surrounding AI concerns what these systems may eventually do with greater autonomy, and the research deserves to be taken seriously. Yet another transformation is already unfolding in ordinary messages, meetings, proposals, and conversations, where machines are becoming increasingly skilled at performing the language of humanity while human beings become more accustomed to sending the performance in their place.

We are teaching AI how to sound like us, and it is inviting us to decide how deeply we still want to participate in sounding like ourselves.

Perhaps the future will depend upon our ability to recognize the difference.

The central question may concern far more than whether machines eventually become human. It may be whether human beings remain willing to undertake the beautiful and occasionally inconvenient labor of being present in their own lives.

Where is AI helping you express something that is genuinely yours, and where might convenience be speaking from a place that still requires you?

I would love to know where you land.

About Deevo

Deevo Tindall is a founder, advisor, storyteller, photographer, and relentlessly curious human who spends his time exploring why people build what they build, why the obvious problem is rarely the real one, and how a better conversation can change the architecture of a business, a relationship, or a life.