I am currently drafting a paper on the Evolution of Human Intelligence which will bring together three interacting models, each representing different levels of activity and abstraction. The jumping off model I am calling the “Brainwave Model” which looks at simple decisions at the human short term memory level. Above this is an “Intelligent Pattern Recognition Model” which examines the relevant CODIL research and its relevance to culture, natural language and intelligence – and in effect defined the brain’s “Symbolic Language”. Below the Brainwave Model” there is the “Ideal Brain Model” (early draft to be rewritten) which looks at what the neurons need to be able to do in order to support the two higher models. The paper will continue looking at the evolution of the brain and human intelligence, using the models as a guide, starting with the requirements of a simple animal and looking at how the brain’s power increases as culture evolves.
Draft Section: The "Brainwave Model
The Brainwave Model forms a short term memory bridge which links the complex high level mental activities which we associate with human intelligence, with electrical and chemical activities at the neuron level, and the objects in the real world we are thinking about. It is best described by a simple example.
Imagine the brain as a sea of interconnected neurons and into this sea we drop pebbles of information. This creates ripples of activity which spread out across the sea, and eventually die away. For instance our eyes see a rabbit and result in a “rabbit” ripple becoming active. This process could well involve many hundreds or thousands of neurons becoming active as the ripple develops and this activity can only pass between neurons which are linked. Each ripple can be considered as an active thought in the short term memory and at this level of modelling we are not interested in the fine detail within a wave of activity.
At the same time the body becomes hungry and a “food” ripple becomes active. The two ripples spread and meet and combine to generate a new brainwave – “rabbit pie”. At the point at which they coalesce there will be a neuron which is linked in such a way that it can be activated by either the “rabbit”, the “food” or “rabbit pie” ripples. What has happened at the thought/concept level could be represented as:
rabbit, food à rabbit pie
We can generalize this to model the human short term memory. At any one time there are a maximum number of ripples (about seven) which can be active at any one time. Each ripple can be given a concept name, which for convenience in this text will be shown in bold brown font. Where two or more ripples intersect there will be a neuron (or a group of neurons) which can be activated by the relevant concept and this can “take a decision” by activating a new ripple.
Of course it is important to realise that a concept name, such as rabbit, is not a precisely defined entity as the ripple through the neurons would be different depending on the colour of the rabbit, or whether it was a real rabbit or a rabbit in a children’s story book. It may well be that in some situations the sight of a wild rabbit, or a carcase in a butcher’s shop will trigger the rabbit pie decision, while the sight of a domesticated rabbit, or a picture of Beatrix Potter’s Peter Rabbit will not. At the same time the strength of the food concept will vary depending on how hungry you are. Other minor factors might affect whether you imagine the pie to be topped with short crust or puff pastry – or whether instead you think of rabbit stew. Such differences are an essential features of how the brain works.
The brain is a dynamic learning (and forgetting) system which is not concerned with any externally defined global models. In a relation such as
rabbit, food à rabbit pie
the “meaning” of rabbit is defined only by the ripples active at the time the rabbit pie decision is made. As part of the brain’s learning process the mental activity involved will have a feedback effect which could modify the way the activated neurons are linked – so the meaning of rabbit could be slightly different the next time a similar rabbit is seen in a similar situation.
When we come to the exchange of information between people using natural language we need to agree stable long term concept names for objects, such as rabbits, but our individual brains will associate the concept rabbit with different memories which will develop over time. Later in the paper, when I look at CODIL as a model of brain activity, the same situation arises. CODIL was conceived as a practical working tool which did what its human user wanted – and by default items (the equivalent of concepts) have to be stable – although options were built in to allow CODIL to dynamically alter its behaviour over time.
Before moving on to discuss higher level models of brain activities it is necessary to understand how this simple brainwave model provides a basis for further research. The brain contains billions of neurons, and each neuron has direct (and indirect) links to many other neurons. This would represent a massive array involving tremendous computer power if one tried to look at the problem globally. The wave of activity associated with the current thoughts (concepts) in our short term memory act as an filter on the vast array selecting a minute number of entries – and in some cases none at all. When considering how the brain processes higher level “intelligent” ideas all neurons and links can be ignored apart from the tiny number which are activated in the current context.