Biological AI, a unit of cognition?

Biological AI, a unit of cognition?

The octopus has a brain in  each of its eight legs … and a ninth brain to do, well, whatever it does.

Presumably the ninth brain is the ‘CEO’ of the creature’s nerve system and makes the ‘big calls’

but we don’t know exactly how that comes about, so let’s assume the ‘big brain’ does the clever stuff.

 

The octopus is a very smart animal, learning quickly and exhibiting high level cognitive skills,

but because sadly it lives only a short time all its lifetime learning is lost generation after generation.

However, as a neurological layout, the ‘brain for each leg’ has always fascinated me. 

It makes explicit a structural and histological gap, hidden in our species between a conscious big brain and the

unconscious autonomic nervous system.  

So, setting the octopus aside but being mindful of its example, it seems to me that our ‘brain’ too is 

an aggregate of ‘brains’ that merely looks like a unitary entity but  is not.

We may not have brains in our appendages like the octopus but we probably

had separate ‘brains’ that have been aggregated into neatly packaged ‘bulge’ of nerve cells.

This I imagine to be similar to the aggregation of thousands of separate nephrons bilaterally

distributed along the length of an early triblobast worm eventually into two neatly bound kidneys.

In order to pursue the idea of reducing a ‘brain’ into its simplest ‘nephronic’

(borrowing the analogy from above) form I have imagined a basic ‘brain-unit’

which when aggregated automatically generates an information rich layer of abstraction beyond

immediate needs of functionality

The organism’s brain-unit at minimum consists of a sensor for input values; an effector for output;

a transmitter to indicate activity and a  working memory*. 

Let’s imagine a basic brain-unit: ( Figure 1)  based on existing simple biological structures.

  

Sensor: generic detector ( eg a light spot, magnetometer, deformation sensor)

Effector: motile armature (aka pilli, cilia based on contractile proteins)

Transmitter: a flashing light ( eg photo-phosphorescence or a transient electrical membrane potential)

Memory*: Oscillating chemical systems (eg a molecular-equilibrium oscillator reaction,

or oscillating electro-membrane potentials)

*I imagine to be a ripple memory, ie a die-away echo that is persistent but not permanent

This type of memory was common in early computers and was called mercury-memory, an early form of RAM. Chemical equivalents eg Briggs–Rauscher reactions oscillators are well known in biology and chemistry

Example behaviour. Consider A simple ‘swim-to’ response of a light dependent organism.

 Light is detected by a sensor,a swim hair is activated: if light persists, swim activation continues,

if not, swim response slowly stops as stimulus ebbs away. Unicellular photosynthetic flagellates are adept at this behaviour as school students of Paramecium will know.

We can imagine an array of these units as shown in Figure 2.

In fact we can create quite a decent response to the light stimulus and any other stimulus using such arrays. 

But, what if we have a lot of brain units?

If we are dealing not with unicells but colonies of cells, or even multicellular creatures.

I mean assume the logical assembly described as part of an organism is replicated beyond utility to redundancy.

In a multicellular organism this is easy to imagine for cells capable of creating

a brain-unit but as it were far from the action.

Some units are employed but others crucially, are simply ‘spares’. Units which have no immediate function.

Imagine then a spare 'brain' as in Figure 3. It can detect activity occurring in its sister units because

it too has a sensor and can see the sister units transmitting;

it too has a ripple memory to store (if only temporarily) an input signal; like them all it too has a transmitter but  its effector is not attached to anything. It does not ‘do’ anything but it the cell ‘aware’ of something going on.

In other words the spare brain is aware of the activity of its multiple counterparts as they go about their tasks.

It can’t do anything itself but it can ‘talk-about’ its awareness via transmissions which in turn can be ‘heard’ by the other brain-units. These units may not ‘understand’ what the spare brain is saying but the transmission  by the spare-brain is more complex and carries more information than any of their single transmissions simply because it is created by input from multiple sources separated in time.

The nature of the spare-brain’s transmissions will reflect the multiple stimuli from the other brains and therefore will have multiple ripple-echos from the multiple inputs. Ripples are waves and waves reinforce or destroy each other according to frequency, period and amplitude. Thus the spare brain’s ‘talk’ is then more of an information-rich ‘song’ derived from the spatially separated brain-units. That is, it contains information about the world outside.

The next step is begged by the question ‘can the organism benefit in terms of survival by leveraging the information from the spare-brain’s song?’  Clearly a coordinated feeding or even escape response (for instance) by the array of units that have effectors, would in fact be a response that reflected an abstracted ‘concept,’ an awareness of its environment.

Evolution, blind though it is,  is very effective at selecting behaviour that has utility

value and the paragraph above seems plausible to me.

Finally there is no natural limit to the hierarchy of brain-units,

Vast pyramids can be imagined, each layer more abstract than the one below. A real brain in the making. 

Thanks to the example of the octopus and its spare brains this may be a window into how

intelligence evolves automatically. A self-training Bio-AI system.