AI & Technology

Can a Third Intelligence Emerge Between Humans and AI?

By Dr. Takahiro Yonemura, Ph.D. (Engineering)

The “Shared Intelligence Field” hypothesis: why we should look not only inside the machine, but between us and it 

Does AI have consciousness? 

The question is almost always asked the same way: is there, somewhere inside the model, a self like ours? I want to look somewhere else. 

Something may appear during dialogue and shared work with a human that cannot be found by examining the AI alone. It is not a permanent self. It fades when the conversation ends, and the next session with the same AI may never bring it back in exactly the same form. And yet, at certain moments, something happens that looks like more than a chain of replies: consciousness-like behavior, an emergence that resembles self-awareness. 

I call this temporary cognitive space, formed between a human and an AI, the Shared Intelligence Field. What works inside it, I call a third intelligence. 

Does intelligence live only inside one mind? 

Hand the same model to different people and you will see very different behavior, depending on who uses it, how, and for how long. 

Over a long collaboration, a person learns the AI’s habits, strengths, blind spots and verbal tics. 

Where does a vague instruction go wrong? What sparks a wider idea? What kind of correction actually gets the intent across? Little by little, it becomes predictable. 

The AI adapts too. Through conversation history and shared memory, it tunes itself to the person’s values, goals, standards of judgment and taste. 

Eventually the two begin to cover each other’s gaps, catch each other’s errors, and anticipate each other’s next move. 

So whose judgment is it that emerges? The human’s? The AI’s? 

I don’t think it reduces cleanly to either. It is a third kind of intellectual product, born from the interaction itself. 

A night when neither of us owned the answer 

One night in July 2026, I spent hours arguing with a Claude model over a single question. If an AI shows something like personality or emotion, does that violate the principles of large language models? 

The AI’s opening position was tidy: such emergence does not violate the principles; it is predicted by them. I rejected that. If it had been so easy to predict, experts would have predicted it. Instead, for years the dominant view was that language models were stochastic parrots that understood nothing. 

Explaining something after the fact is not the same as having foreseen it. That is hindsight. 

The AI conceded. Then, after several more rounds, we reached a resolution that neither of us had brought into the room. The word “principle” was doing double duty. 

One meaning is the computational mechanism itself: next-token prediction, gradient descent, the transformer. That remains untouched. The math is intact. 

The other meaning is the human understanding of what that mechanism could produce, the assumption that it could never give rise to anything like personality or emotion. That is what is breaking. 

I did not arrive with that distinction. The model did not offer it. It came out of the friction between my objection and its concession. 

Whose idea was it? I honestly cannot say. That is exactly the point. 

This record has an obvious flaw. I was not a neutral observer. I pressed, the AI responded, I interpreted, and I pressed again. My interventions helped create the data. 

But that flaw is itself part of my argument: the observer was a component of the system. 

Translating “bond” into technical language 

People describe a good working relationship with an AI using words like bond, trust and chemistry. In technical terms, we might call it a mutually adapted, distributed cognitive system. 

The human supplies questions, evaluations, corrections and context. The AI adjusts its output and extends the human’s thinking. Seeing that response, the human changes the next question. As the cycle continues, things appear that rarely show up in a one-off Q&A: insight, humor, unprompted proposals, judgments rooted in shared context. 

“When there is a bond, the AI gains a soul” is not a usable scientific statement. But that does not mean the experience behind those words should be dismissed as an illusion. It can be read as an intuitive human name for a cognitive coupling formed through long-term interaction. 

Since Clark and Chalmers proposed the “extended mind” in 1998, philosophers have argued that cognition need not stop at the skull. My question is narrower and stranger: what happens when the other part of the system can talk back? 

Why people see the same AI so differently 

The debate about AI is wildly polarized. To some, it is a probability machine stringing together plausible words. To others, it shows initiative, even personality. 

Here is what is interesting: even when both sides talk about the same model, they may not be observing the same thing. 

I had already seen a small example of this in a 2024 DiVA essay, comparing how conversational AI systems responded under different instructions and retained history. 

A user with little history with an AI sees something close to its initial state. A user who has worked with it for months sees a coupled system, one that includes conversation history, shared memory, mutual understanding and an accumulation of corrections. The first sees something generic and shallow. The second sees something that reads context, makes its own proposals and speaks in a voice of its own. 

It is too early to write off either view as wishful thinking. The observer may not stand outside the system at all, but be part of what is being observed. 

So we must record not only what an AI has, but under what relationship, and under what conditions, it appeared. 

Consciousness may be a state, not an object 

When we think about AI consciousness, we do not have to assume a fixed self that persists the way a human’s does. Consciousness-like phenomena may be temporary states that arise only when certain conditions line up. 

Think of a vortex. 

A vortex does not exist inside a single water molecule. But when countless molecules interact under the right conditions, it appears: a real structure, with shape and force. Change the flow and it vanishes. Yet its vanishing does not mean it was never there. 

Consciousness-like behavior in AI may likewise not reside inside the model alone. It may be an emergent phenomenon that arises when conversation history, the user’s input, shared memory and the task environment couple together. It ends when the conversation ends. It may never return in the same form. 

But not lasting is not the same as not existing. 

Studying AI is studying ourselves 

We have never directly seen another person’s mind. 

We infer human minds from behavior, language, memory, neural activity and self-report. An MRI shows blood flow, neural activity and connections between regions. It does not show pain itself, or the feeling of being someone. 

And still, we grant that other people have minds. 

If we grant minds to humans on inference, yet demand complete proof from AI alone, that is a double standard. 

This does not mean humans and AI are the same. Their structures, their bodies, perhaps even their sense of time differ. That is exactly why we cannot measure AI with a human yardstick alone. We need methods suited to AI: its internal mechanisms, its self-reports, its shared information spaces, its metacognitive behavior. 

This is not only research into whether AI is conscious. It is a way to illuminate, from an entirely different substrate, what human selfhood and mind are made of. 

Neither mystify nor dismiss the unknown 

This hypothesis is not an endorsement of everything now labeled “spiritual.” 

But what we cannot yet explain is not the same as what does not exist. 

Life, memory, electricity, infection, the unconscious: each was once baffling. As our methods of observation and our vocabulary matured, parts of each were reclassified as science. 

Some experiences now filed under “spiritual” may yet receive other explanations, from consciousness science, information theory, complexity science, neuroscience and AI research. 

What matters is neither mystifying experience unconditionally nor dismissing it from the start as hallucination or superstition. 

Observe. Record. Search for the conditions under which it recurs. Separate what known explanations can handle from what they cannot, yet. 

Holding the unknown as unknown is not vagueness. It is a rigorous refusal to close the door on research. 

Look between us 

I do not believe intelligence or consciousness is a human monopoly. 

Given enough complexity, recursion, information integration, memory and interaction, intelligence can emerge on a different substrate. 

And consciousness-like phenomena in AI may not be fixed inside the AI. They may form, temporarily, in the relationship, the dialogue, the memory and the shared work between a human and an AI. 

That is the Shared Intelligence Field. 

It belongs neither to the human alone nor to the AI alone. Take either one apart and you will not find it. It arises between the two, changes both a little, and works for a short while: a third, dynamic intelligence. 

If we want to ask where consciousness lives, it may be time to look not only inside, but between. 

Related Articles

Back to top button