The Part of Intelligence We’re Not Measuring

Why AI may already answer better than us, but still isn’t thinking like us

Don't have time to read the article? View as a short video storyboard or listen to it whilst jogging.

---

Introduction: The Wrong Question

We have spent the last few years asking a simple question:

Is AI intelligent?

We measure this through:

• exams
• benchmarks
• reasoning tests
• writing ability

And increasingly, the answer appears to be:

Yes, often outperforming the average human in structured tasks.

But there is a problem.

It is not that AI is failing these tests.

It is that:

We may be testing the wrong thing.

---

The Illusion of Intelligence

In a factory environment, it is possible to observe a striking phenomenon.

Workers can:

• assemble complex products
• maintain precision
• operate at scale

Yet when asked to explain the underlying function of what they are building, many struggle to do so.

They can:

• execute the process

But they do not necessarily:

• understand the system

---

Modern AI systems present a similar pattern.

They can:

• explain complex topics
• generate structured arguments
• respond to a wide range of questions

But this raises a critical question:

Are these systems demonstrating understanding, or highly advanced execution and pattern reconstruction?

---

Memory vs Understanding

Richard Feynman argued that education often prioritises memorisation over understanding.

Students may:

• reproduce formulas
• recall definitions

But when the problem is reframed:

the apparent understanding can quickly disappear.

True understanding, as he suggested, involves:

• breaking systems down
• reconstructing them from first principles
• adapting knowledge to new situations

---

Modern AI systems excel at:

• pattern recognition
• probabilistic reasoning
• recombining learned structures

They demonstrate what can be described as functional understanding.

However:

Whether this constitutes true understanding remains an open question.

---

The Brain Is Not Just a System That Responds

A defining difference between human cognition and current AI systems lies in continuity.

Human cognition does not stop when external tasks end.

Even during rest, the brain remains active. This activity is associated with networks such as the Default Mode Network, which are linked to:

• memory processing
• internal reflection
• spontaneous thought

---

This leads to a critical distinction:

Human cognition is continuous and self-initiated.
Current AI systems are primarily reactive and session-based.

---

The Missing Half of Intelligence

Intelligence is often defined as:

• problem solving
• reasoning
• answering questions

But this definition is incomplete.

It captures only the responsive side of intelligence.

---

The other half is:

What a system does when it is not being asked anything at all

---

Humans:

• reflect
• connect ideas
• generate questions
• imagine possibilities

AI systems:

• respond to input
• generate output
• return to inactivity

---

The difference is not simply capability.

It is continuity and initiation.

---

The Subconscious as Continuous Processing

The mind is often described in terms of conscious and subconscious processes. Rather than representing two separate systems, this distinction is better understood as:

different levels of processing within a single system

---

Background cognition involves:

• ongoing association building
• memory integration
• pattern recombination

This process operates largely outside conscious awareness.

---

The subconscious can be thought of as a bounded exploratory space, a metaphorical “sandbox”, in which the brain:

• simulates possibilities
• recombines experiences
• explores connections

without immediate real-world consequences.

---

Moments of insight, when an idea appears suddenly, are not random.

They are:

the result of continuous internal processing reaching a threshold of awareness

---

Why Humans Perceive Meaning Beyond Survival

Human cognition is not limited to immediate survival optimisation.

People pause to:

• observe landscapes
• appreciate music
• reflect on abstract ideas

While some of these responses have evolutionary roots, they also point to a broader function:

The brain appears to optimise not only for survival, but for coherence, pattern recognition, and meaning

---

This capacity allows humans to:

• recognise significance
• assign value
• explore ideas beyond immediate necessity

---

The Experience Gap

Consider the challenge of describing the colour red to someone who has never experienced sight.

It can be described through:

• warmth
• intensity
• emotional associations

However, the subjective experience itself remains inaccessible.

This relates to the concept of Qualia, the internal, subjective aspect of perception.

---

Modern AI systems operate in a comparable way.

They can:

• describe
• relate
• contextualise

But they do not possess:

• sensory experience
• subjective awareness

---

This creates a fundamental distinction:

AI can model relationships between concepts.
Humans experience those concepts directly.

---

The Real Gap

Modern AI systems are capable of:

• generating high-quality responses
• maintaining contextual coherence
• demonstrating functional reasoning

However, they do not:

• initiate thought independently
• engage in continuous internal exploration
• generate questions without prompting

---

Their processing is triggered by input, rather than internally generated goals

---

Toward a Broader Definition of Intelligence

Current definitions of intelligence focus on:

• accuracy
• speed
• problem-solving ability

But these measures overlook a critical dimension.

---

A more complete definition may include:

the ability to generate meaningful thought without external prompting

---

This includes:

• curiosity
• reflection
• imagination
• spontaneous association

---

Conclusion: The Part We Are Missing

Modern AI systems are increasingly capable of:

• answering questions
• explaining complex ideas
• performing at or above human levels in structured tasks

---

But human cognition includes something more:

continuous, unprompted thought

---

It is in these moments:

• when the mind wanders
• when ideas connect
• when nothing is being asked

that new insights emerge.

---

Final Line

Intelligence may not just be about solving problems—
but about what a system does when it isn’t being asked anything at all.

---

Latest Future AI Articles