Evaluation under simulation


Evaluation under simulation

Hi! I'm Tim Tyler - and this video discusses the idea that evolution progressively develops techniques to perform evaluations in simulated environments - where failures are less costly.

I will be discussing the history of evaluation under simulation, and the future prospects for the technique. I will argue that there is a trend towards increase use of evaluation under simulation - and that in the future, this will result in the elimination of much real-world conflict.

To start with some basics: in order to evaluate designs, it is necessary to build and test them - and this is true both for the biological designs nature produces, and for ones produced by humans.

In the human sphere this process is known as research and development - and biological evolution employs a broadly analogous strategy - involving trial and error - in order to produce its natural designs.

In the modern era, evaluation under simulation is a familiar engineering technique. Bridges are constructed and tested in simulated environments before being deployed for real. Similarly with skyscrapers, dams - and other construction projects.

The technique is related to other common tools: rapid prototyping and model building.

Evaulation under simulation can be seen as a type of prototyping or modelling that is performed in a virtual world.

The benefits of evaulation under simulation are pretty clear - testing a virtual model helps reduce real-world failure rates - thereby reducing costs.

Bits of information provide an excellent and highly flexible modelling material, which can apparently be used to construct and test anything.

A brief history

For the origins of the technique of evaulation under simulation we must go back millions of years - and look at the origins of brains.

An important function of brains is to act as a simulator of the world, to allow an organism to predict the likely consequences of its actions and thereby choose between them.

Brains are virtual reality simulators. Here is Dawkins making the point:

[Dawkins footage]

Having a simulation of the world benefits animals in several ways.

One example is male combat. With a world simulator, males no longer need to actually fight each other to determine who is the fitter and stronger. They can fight in the simulated environment provided by their brains, decide who the winner would be if they did fight, and then avoid actual combat. That can benefit both parties: the loser of the fight is no longer dead, and the winner has avoided the risk of getting damaged in the battle.

Venomous animals are among those likely to engage in ritual combat - because the potential costs of battle are high - and so the benefits of avoiding real combat are large.

To illustrate this, here are some spiders sizing each other up:

[spider footage]

...and here are some snakes - engaged in mock-fighting behaviour:

[snake footage]

Another example of male combat is found among male elephant seals. These often roar at each other, and rear up to estimate each other's size before going into battle.

[male elephant seal footage]

If they are of equal size and strength then sometimes a real battle results. However, frequently the loser retreats without much of a fight - here a dominant male tramples on a baby seal in his effort to see off a smaller rival.

[footage of male elephant seal in retreat]

It is not just male combat where a simulation of the world is useful.

Female choice is another important area where understanding the consequences of your actions is useful.

Rather than actually mating with suitors and producing offspring for natural selection to evaluate, females who have brains can imagine the consequences of being impregnated by particular males. This allows them to select males with desirable traits, such as disease-resistance genes - or the ability to provide supplementary parental investment. If their brains are working properly, the females can thus avoid wasting their resources by investing in unsuitable offspring.

This strategy of evaluation under simulation seems to have been something that evolution has liked. Brains have enjoyed explosive growth on the planet - and they are now popular and widely distributed.

However, for a long time, the main development seems to have mostly involved making brains bigger.

However, in the modern era, evaluation under simulation has enjoyed further successes in the biosphere - as a result of intelligent design. Engineers have successfully constructed digital computing machines that out-perform biological brains in many areas. These new devices can be made more reliable than brains, can perform serial computations much faster, and have better memories. They also interface conveniently to long-distance communications technologies.

It seems likely that their capabilities will continue to improve - and that in the relatively near future, their functionally will surpass biological brains in most areas that matter.

A likely result of this seems to be even more extensive use of the technique of evaluation under simulation.

To recap, evaluation under simulation reduces the costs to the evolutionary process of performing tests - by executing those tests in a virtual environment, where building things, testing them, and discarding the ones that don't work is relatively inexpensive. The result is faster evolution and so more rapid progress.

The strategy generates benefits directly for the individuals that employ it. Having a high-quality simulator is useful - and will often lead to improved fitness.

The virtualisation of conflict

One possible conseqence of the rise of simulation environments is the virtualisation of much of the world's conflict.

Today, we can see that society has successfully chanelled male competitive energies into ritualised, mock battles in many cases.

Laws dictate that companies compete economically for resources in a marketplace - rather than engaging directly in attacking or fighting behaviour oriented towards acquiring resources by more direct means.

Similarly many nations play football matches with each other - rather than going to war.

Looking at what happens in nature, ritualised combat often breaks down when combatants are evenly matched - and neither side can calculate in advance who the likely winner will be. Then they have to perform their test in the real world - in order to evaluate their issue.

However, in one respect this can be seen as a consequence of the limitations of their simulation equipment. With better simulators, the cases where the outcome would appear to be uncertain to the combatants would be reduced - and so fewer real-world fights would happen. Also there are conventions for deciding the outcomes of combats between evenly matched competitors - rematches, tie-breaks, and so on.

It seems reasonable to expect that the combination of better simulators, and conventions for deciding outcomes in cases of ties will reduce the frequency of real-world combat to very low levels.

Already we are witnessing progress in the moral zeitgeist - and combat is being regarded is increasingly primitive and barbaric.

Warfare is bad for business - and these days, many powerful corporate and government interests are simply opposed to it. The virtualisation of warfare seems likely to result in a reduced frequency of real-world warfare.

It seems likely that - like slavery - the frequency of real-world combat situations will be drastically reduced in the future - and that eventually, most of the evaluations required for evolution to progress will be performed under simulation.

Enjoy,


Tim Tyler | Contact | http://alife.co.uk/