The Witness

By Jessica Lake

The Anatomical Boundary

This started with something I remember reading.

I have no idea how literally true it is.

The argument was that increasing human brain size eventually ran into an anatomical problem.

We walk upright.

That constrains the pelvis.

Babies with increasingly large heads still have to get through it.

Maybe.

What interested me was the constraint.

It reminded me of Stein’s Law, attributed to economist Herbert Stein:

“If something cann0ot go on forever, it will stop.”

I love that.

Somewhere, somehow, continued expansion of the individual brain runs into a constraint.

Fine.

What happens next?

The Network Strategy

If increasing cognitive capacity remains useful, but simply making the individual brain larger becomes increasingly expensive or impossible, evolution has another option.

Add brains.

One individual does not have to know everything if another individual knows something useful and can tell them.

Now information can be distributed across a population.

• One brain knows where the water is.

• Another knows which plant will kill you.

• Another knows where the animals migrate.

• Another knows what happened the last time the river rose.

The individual biological node remains finite. The network does not. Or at least it becomes considerably larger.

That sounds suspiciously familiar to a systems engineer. When vertical scaling becomes expensive, scale horizontally.

But now I have created another problem. The nodes need an interconnect. If the information in your head is useful to me, I need some way to get at it.

Now communication stops being merely social behavior.

It becomes part of the information architecture.

The Physical Sieve: Constraints on Expressive Mediums

So how do two biological nodes communicate?

Reality gets another vote.

The signaling medium has to survive the environment in which the organism actually lives.

Visual communication is wonderful.

• Gesture.

• Expression.

• Posture.

• Direction of gaze.

• Pointing.

But visual communication has limitations.

Turn off the lights.

Put something between us.

Turn your back.

Give both of my hands something more important to do.

Oops.

Sound has a different collection of advantages.

I do not have to be looking at you.

You do not have to be looking at me.

My hands remain available.

It works in the dark.

It can travel around some obstacles.

That does not make sound universally superior.

It makes it useful under a particularly interesting collection of constraints.

So perhaps vocal communication becomes increasingly important.

Fine.

We have an interconnect.

Unfortunately, it is terrible.

The Channel Constraint

Inside my head is an enormously complicated parallel state.

• Sensory information.

• Relationships.

• Objects.

• Goals.

• Emotions.

• Expectations.

• Context.

• Things I know about you.

• Things I know about me.

• Things I know about the world.

And now I get to transmit all of that through...

One utterance after another.

Good luck with that.

• Speech is not perfectly serial, of course.

• Prosody carries information.

• Timing carries information.

• Volume carries information.

• Gesture and expression can ride alongside it.

But words themselves arrive in sequence.

• One—

• After—

• Another.

So I have a transformation problem.

I have to take something enormously complicated and substantially parallel and squeeze enough of it through a narrow temporal channel that another mind can reconstruct something sufficiently close to what I mean.

Not everything I mean.

That would take forever.

Enough.

Language begins to look like an encoding system.

Parallel internal state goes in.

A serial approximation comes out.

But solving that problem immediately creates another one.

The receiver.

The Decoding Optimization

I know what I mean.

• You don't.

• Worse, you are not a passive decoder.

• You have your own knowledge.

• Your own assumptions.

• Your own expectations.

• Your own fears.

• Your own goals.

• Your own interpretation of every miserable word I choose.

So successful communication cannot depend only upon what I say. It depends upon what you reconstruct when you hear it. That suggests another useful piece of machinery.

A model of you.

Before I say something consequential, I can ask:

• What will she think I mean?

• What will he object to?

• What does she already know?

• What assumption am I making that he does not share?

• Where will she become confused?

• Where will he become angry?

• What am I leaving out?

• What order makes this easiest to understand?

Now I have a Devil’s Advocate.

Not necessarily because I evolved some magnificent abstract reasoning engine.

Perhaps because saying the wrong thing to another human being can be expensive.

The model of the receiver becomes an internal adversary.

I produce something.

• It attacks.

• I modify.

• It attacks again.

Eventually the message survives well enough to risk sending.

But communication is not merely words.

I can rehearse the rest of me too.

• Where do I pause?

• When do I slow down?

• When do I smile?

• When do I look away?

• When do I stare directly at you?

• What does my face do while I say the dangerous part?

I can rehearse the gesture.

• The posture.

• The timing.

• The movement.

And if I have visual imagery, perhaps I can rehearse your side as well.

• Your face.

• Your expression.

• Your reaction.

Now I am no longer merely constructing a sentence.

I am rehearsing an interaction.

• My side.

• Your side.

• My response to your response.

• And perhaps yours to mine.

That can become quite a loop.

And This Is Where I Stopped

That was enough.

I had begun with a birth canal.

Somehow I had ended up with a system that takes an enormously complicated parallel internal state, squeezes it through a miserable little serial channel, models the person at the other end, rehearses what it intends to do, tests the result, modifies it, and tries again.

Fine.

I went to bed.

The next morning I woke up with another thought.

Hey.

That pesky little voice sounds suspiciously like consciousness.

I had not been trying to explain consciousness.

I had been trying to explain persuasion.

But suppose I want to know whether something I am about to say works.

What better way to test it than to hear it?

I already possess an extraordinarily sophisticated system for receiving speech.

Why invent another one?

Say it internally.

Listen to it.

Output becomes input.

The same principle can extend beyond words.

I can feel myself making the gesture.

I can rehearse the pause.

• The smile.

• The stare.

• The movement.

For someone with visual imagery, perhaps the proposed scene can be seen as well.

Run the proposed output back through whatever machinery is already available to evaluate incoming information.

• Hear it.

• See it.

• Feel it.

• Generate.

• Receive.

• Evaluate.

• Constrain.

• Modify.

• Again.

Now I have an internal voice.

Perhaps an internal scene.

A model of another person.

A continuing serial stream.

And apparently someone listening to the whole thing.

Me—

Oh—

My.

I seem to have wandered into consciousness.