Saturday, 26 November 2016

My "Evolution of Human Intelligence" model predicts the risk of considerable political instability

Over the last couple of months I have been working on a comprehensive document on my model of the evolution of intelligence, which has been delayed by domestic, health and other events but should be ready by the New Year. Trump's victory comes at the time I was drafting the section on what happened to make us "so much more intelligent" than animals and also explains some of the dangers of developing a "speed learning" brain on an animal brain foundation.

So how does my model explain the rise of Donald Trump. ...
Over the last 5 million years, apart from the most recent times, all early humans lived in small groups and their advantage over other great apes was their ability to adapt their ways of living, and their tools, to different environmental niches at a time when the Ice Ages were causing changes in to sea levels and vegetation cover all over the world. Normal evolutionary pressures would apply to such groups - the groups who best adapted and had the best tools relevant to their preferred environment would be the ones which survived. As we know from recent DNA research our own species could interbreed with Neanderthals and Denisovans (and other human "species") after separations of perhaps half a million years.

Assuming that such crossbreeding has been occurring over the last 5 million years, with climate change driven by the Ice Ages stirring the pot, the result will be the intermingling of a diverse genetic pool - and a an exchange of tool making abilities - with hybrid groups combining the best genetic and tool making skills of the parent groups.

It is important to realize that as long as tool making was passed on by simple trial and error copying the key survival need was the ability of infants to accurately copy their parents, and the neural net learning time limit (which affects all animals) still applies. Groups which had slightly larger brains would be better able to combine the tool making skills of interbreeding groups - so there would be evolutionary pressures for the brain to get larger. This means that over the last five million years ago the brain got bigger - culminating with the Neanderthal species - which had the largest ever brain.

Something unexpected happened some 100,000 years ago. Our species started on a roller-coaster ride where our ability to make better tools started to increase at an accelerating rate. In a few thousand generations we have moved from making better flint knives to making robots which explore distant planets. The amount of information we use has clearly increased so one might expect the evolutionary pressure would be to evolve an even bigger brain would be enormous. In fact what has happened is that our brains are slowly getting smaller!

The neural net model I have developed, using CODIL, provides a possible explanation for "increasing knowledge" and a smaller brain.
  1. While humans may have had a simple language for several million years there comes a point where the it becomes the most effective way of transmitting culture from generation to generation. It is then more effective to switch off "common sense" trial and error learning (which has served animals for a billion years or so) and accept the truth of what you are told (by voice) by your parents. This allows much more tried and tested information can be passed in this way in the same "learning time."
  2. The more steps involved in making a tool  before "success" can be claimed limits the complexity of tools that can be passed from generation to generation using trial end error. Teaching supported by language (with verbal reinforcements on the way) allows more complex tools to be made.
  3. Language allows learning by generalizations - which require far less brain storage than having to learn very many individual cases by trial and error.  
  4. Language allows young children to be taught about dangers (for instance when hunting large animals) without exposing them to the dangers of watching hunts taking place -  which is what learning by watching the action would entail.
  5. Above all language is a self modifying tool - every generation refines it a bit - and with every refinement it becomes a more efficient teaching tool.
All these factors effectively mean that we are now all part of a cultural neural network, which adds layers of complexity at each generation. Survival of the individual depends more and more on the culture, and less and less on purely biological factors, As the amount of information increases it is shared out - with different individuals and groups cultivating different areas of specialist knowledge. Modern civilization depends on at least 3000 generations of "deep learning" and individually we do not need such big brains to as in evolutionary terms survival depend more on the "protection" provided by the group culture rather than our ability to survive as a hunter-gatherer threatened by carnivores.

The real problem is that our civilization is built on a vast and in places contradictory ocean of knowledge and manned by a motley crew who all have their own interests, and those of their local group, at heart. Normally this works well on the practical day-to-day side of living. Language has allowed us to morph the way we think so that we can work together to make ever more useful tools to make life more comfortable, and we all accept that we need to specialize and trade our skills because by working together we all benefit.We take it for granted that, for example, some people grow our food while others build houses, or design better computers, or fly planes.

However to make this work smoothly there must be some way of managing and sharing resources and the work load - and this leads to differences and disputes up to the scale of major wars. I am not going into the politics of this but rather concentrate on why our "animal brains" let us down

Difficulties arise because, hidden beneath a thick blanket of culture, our "animal brains" which did not evolve to work at this level of sophistication or in such large communities. The basic animal brain evolved to analyse the environment in which the animal was living and make decisions to help the animal survive, In effect this is equivalent to what we call "common sense". We look around and it is "obvious" we live on a solid immovable flat surface - and massive tree branches fall faster than autumn leaves. Such an approach is clearly appropriate when faced with a situation which you can observe - but where you have no understanding of the underlying causes. Because the human brain's capacity is limited any one individual can only have in depth knowledge of a tiny part of the total cultural network and is forced back onto common sense observations of areas of human activity he does not understand.

So what are the limitations of our "animal brain" which get us into trouble.

The way the neural net works means that it cannot directly handle negatives, but rather works on the principle that some observational patterns (for instance those that representing danger) have a higher priority to other patterns. This means that perceived threats are given prominence, and once learnt are hard to unlearn. The limitation concerning negativity also explains the universal human failing called "confirmation bias." People readily accept ideas which agree with what they have already learnt, and ignore, or sometimes actively reject, ideas which do not fit with their preconceptions.

The way that the brain learns by observation creates another problem. Early brains clearly did not have the capacity to collect large volumes of data and process it with sophisticated statistical algorithms in the way that big data AI research programs do. Instead each remembered pattern has some kind of satisfaction ranking which is tweaked up or down as the pattern proves useful. Because the human brain still uses this mechanism the history of the pattern is automatically lost, and this explains why human long term memory is unreliable. It also provides a clue to our poor understanding of probabilities.

Of course our success as a species depends on speed learning - which works because we learn faster if we automatically accept what is taught is true  - giving such . We naturally follow what appears to be the best teacher available - which will tend to be the one with the most charismatic approach and the most self-confidence. If they have a simple message (compatible with the limited capacity of human short term memory) which matches your own confirmation bias, so much the better. Of course it helps if you already see them as part of your own social group - and nowadays this undoubtedly includes one way contacts via the TV screen, social networking accounts.

So back to the present political situation which triggered this blog. Any cultural grouping works better if all the participants work together for a common good and for this to happen we have to accept that there are people who are experts in fields that we do not understand. There needs to be a framework for agreeing the rules of interaction and there also has to be a mechanism for sharing out resources (money is a commonly used measure). This means that some people are needed whose skill is in managing the system - and the bigger the system the harder it is for any one individual (or a committee) to understand it well enough to really be in control. As the world get smaller the conflicting cultures of different populations adds to the complexity and increases the opportunities for dispute.

For most people most of the time things work moderately well. Food appears in the shops  and they have a roof over their heads. They accept that you need a skilled pilot to fly the plane that takes them on a holiday, etc., etc. However the tools we have created to support communal living are based on a rule-oriented model of the world and to work smoothly society requires its members to agreed a set of rules as to how resources are shared. But remember - our brains are, underneath the cultural gloss, simple pattern-recognition networks working from the limited "common sense" view of the would that evolved when every brain had to work independently.

In any such rule-based system there are bound to be winners and losers and in practice the reasons for any particular individual being a winner or a loser are many and varied. For millions of years the evolutionary process will have involved groups splitting up. The breakaway group - led by a charismatic individual with a dream of a better world - will have gone off to look for a better cave or a better hunting grounds or a better island or in more modern times a better utopia.Alternately one group may decide that they can get more resources by in some way displacing or conquering a neighbouring group. Over perhaps 5 million years some groups survived and prospered while other failed and our species evolved because only the best of a large number of groups survived in the long term - and most "go it alone" groups would have vanished without trace. That's how evolution works.- by "killing off" the less satisfactory solutions.

As society developed there were increasing advantages of being in bigger groups, and groups of groups. However the rules to bind the group(s) together became more complex and more arbitrary - and often were bent to favour some groups with a disproportionate share of the resources. In addition some people try to exploit the community to get more than a "fair share" of the resources. It is therefore not surprising that some individuals and groups became disillusioned.

This is where the inherent weaknesses of our animal brain start to cause trouble. The simplicity of the underlying neural net leads to confirmation bias ans those who feel (other rightly) that they are unfair losers will naturally be receptive to ideas which sound as it they could benefit. The speed learning shortcuts that mean we are "more intelligent" than animals means that the charismatic individual who has a simple solution (a cleverly worded sound bite small enough to fit in our limited short term memories) will be eagerly listened to. Perhaps 100,000 years ago the disgruntled group would follow the charismatic leader to the greener pastures and better hunting ground of the Utopia he claims (rightly or wrongly) lies on the other side of the dangerous river.

But the modern world is already overcrowded, and modern life is so interlocked, that it is not possible to move to another part of the world and try and build a fairer environment completely independent of the old, admittedly often faulty one. Simply finding one idealized cause and solution for all the troubles and then trying to impose this on the whole community in situ is unfortunately a very dangerous solution There are untold cases where the downtrodden masses have agreed on "who is to blame" and then realized that the problem was far deeper that their newly appointed leader had claimed - and that the much hailed simple minded solution will not cure the problems.

Upsetting the apple-cart in order to get "a better system" is a risky business in the highly networked modern world as almost certainly the only this that the winners have in common is the desire to change the system. Look back in history to the French Revolution, the Russian Revolution, the rise of Hilter, the collapse of Yugoslavia after the death of a strong ruling dictator, and more recently the much celebrated (in the first year or two) Arab Spring. In every case the groups that got together behind a sound bite rallying call to replace the establishment started to fight among themselves because there was no workable agreed agenda.

What my research has shown is how the simple neural network in an animal's brain can morph into an "intelligent" cultural rule-based system. What the preceding discussion and the potentially catastrophic effects of Brexit and the American Presidential Election are demonstrating is that under certain condition the human brain can discard the cultural rules that make an integrated society possible and revert to animal brain thinking with all its evolutionary ragbag of "poorly designed" features. Our speed-learning brain readily accepts the words of a charismatic individual with a simple message who we feel we can trust, perhaps because he frequently "visits" our home (via the TV screen) and we bypass all the hard-learnt safety checks.

What is happening would come as no surprise to an alien scientist, researching the evolution and ultimate fateof intelligence throughout the Galaxy. "Brexi"t and "Make America Great Again" are just examples of "accidents" in the evolution of the species and the culture. Only a few random genetic changes are beneficial  and many prove lethal, and the same survival of the fittest applies to "random" changes in culture, and the more complex the organism the more likely a random change is to be lethal, as it seems unlikely that either change could result in any significant improvement while both are generating uncertainty and outright dissent which will spread like cancer through society.

To conclude, my neural net to intelligence model predicts many weaknesses in the way the human brain works, and if I was a truly independent observer, living many light years away I could sit back and say that both the growth of human culture and intelligence, and the vulnerability of the cultural edifice humans have built result from the the random activities of the Blind Watchmaker of Evolution. Unfortunately, however hard I try to be objective I cannot forget that I am trapped on a planet where the weaknesses caused by our inperfect animal brains could seriously affect my future, and the future of further generations.

No comments:

Post a Comment