Instinct As Intelligence

The Future Is Wild' on TLC or Discovery recently speculated about a fictional spider that builds webs across canyons. It struck me that instinct could be considered a form of (accumulated) intelligence. The Silver spider grabs a floating seed, connecting a thread to the side of the canyon it is on, gets blown to the other side. A different 'caste' of the same species on the original side then crosses over with a thicker strand. The colony then weaves a whole mesh across the canyon, not to catch insects but seeds. A third caste collects the seeds then then store the seeds in burrows etc. Each spider acts as an automaton but the EmergentProperty seems to an observer as an intelligent system. Instinct encodes this intelligence - it is a stored "program" that directs the individual to achieve goals for it's colony and species despite an unpredictable environment.

Many other actual examples from insects up to mammals indicate intelligent behaviour seems to result from instinct. Beehives regulate temperature by signals sent from the queen resulting in workers adjusting their wing beat rate, they direct each other to pollen by special dances, certain species of ant queen "plant" fungus in advance of laying in time to harvest for grubs to feed when born, etc. The behaviour seems highly intelligent for survival in their situation but clearly the organisms are not thinking about it in the sense we do. They receive inputs of time, temperature, chemicals, their eyes, ears and other organs do basic pattern recognition with simple brains but produce tuned responses based on programming provided by evolution.

Similarly other animals face a choice of how to allocate rest and hunting. Too much rest = not enough food, too much hunting means risk of becoming food and also balance of energy spent versus energy rewarded. I did some research it seems the allocation of daily, weekly, monthly, annual tasks by animals (another task "decision" is timing of moulting versus reproduction") follows DynamicProgramming (in the sense of linear programming) laws. The resulting patterns are optimal for the survival of the animal and species in the given environment (which of course can be upset by changes in the environment). Evolution iterates through alternatives and the resulting "program" is "saved" in the species' GeneticCode. This results in HardWired? neural network pathways in the organism as it develops and manifests in it's LifeSpace? as instincts. Though hard wired instinct still responds to a complex array of stimuli, like EventDrivenProgramming:

Even people - who rely more on cognitive intelligence, still have many instincts. Intuition and emotional responses seem to be controlled by instinct in humans. Habits and trained tasks seem to get "pushed back" into the instinct SubSystem?. For example you can think about 10 other things while driving to work as that task becomes so routine you can do it almost automatically. Or sports - you can concentrate on high level strategy without worrying about details of the actions once practised (and can actually be detrimental if you think too much about low level action sequences during a game).

Our own neuronal development includes many genetically controlled sequences that sound like the spider story above. In fact, most of our genes have done their life's work before we are born. Although the connections between this, instinct and intelligence has yet to be fully understood, the paradigm is clear.

Amen. Each neuron behaves "instinctively", both in the way it grows and the way it functions. The emergent parallels between social insects and neurons are astonsishing. Behavior we consider robotic or mechanical at a low level produces behavior we consider intelligent at a higher level.

Insects also have limited learning ability subject to constraints of brain size and instinct. For example one model of "maximum lifetime fitness" based on a sliding memory window m(t) of previously encountered mates is 

  F(epsilon,m,t,T,lambda,rho,f)  where


    epsilon=egg load,t=time of birth,T=death,
    lambda=encounter rates with hosts and mate types,rho=probability of survival,f=lifetime reproductive success
(from "Insect Learning" pg 177 ISBN 0-412-02561-2 ) in the example above length of m is varied from 0 to 5. They did simulations to see how F varies with m. Several real world observations are compared including that of the female bark beetle Ips pini that shows variation in mate choice based on size of previously experienced males.

So they do exhibit some intelligence at the micro level as well as collectively.

Evolution dynamics also shows beauty and "intelligence" at a macro scale which drives instinct development; consider the stereogram below depicting the iteration [??] (attractor) in a herbivore/carnivor/plant system.

From an Organism's point of view, energy is expended to maximize it's Species' potential for survival. From it's Ecosystem's PointOfView, it's predators keep it's Population in Check and as an external Species, it may be excluded from a new Ecosystem similar to an Immune system. The ability to disperse therefore may be increased if a lesser species acts like a barnacle on the more ferocious, though if this fiercer one is a target at it's destination, the symbiotic guests may be diminished. Stability of the component Populations then is like an outcome of the intelligence of the Ecosystem itself.

HEY!! I am asking, is this entire page an artifact of SchizoidGibberishWikiAuthor, just more organized than usual? AnswerMe, don't just delete my very reasonable question. -- MartySchrader

Sorry, I thought you were intentionally insulting the authors of this page. I didn't realise it was intended to be a legitimate question. I suppose it's possible that SchizoidGibberishWikiAuthor contributed to this page at some point in the past -- obviously in a more sane frame of mind. He certainly references it often enough to suggest a connection with it. However, his current contributions are invariably incoherent word salad. This page is entirely coherent and based on recent science.

Neural nets and similar "devices" have a kind of summing memory where experiences are collectively summed, but individual details or incidents are not kept. Experience may tell a fish to swim away fast if they see a red-and-white striped shape, for example. They may not remember each encounter with striped enemies, but the sum of bad experiences has left a "map" of fright-flight when stripes are seen. This is "lossy" learning, but still effective learning. Similarly, I may not remember each incident when a certain coding pattern caused maintenance problems, but the association of a given pattern and resulting problems still remains in my head. It's a "bad feeling", but only a feeling because we may not be able to dump out a list of the incidents that led to this feeling. Each experience slightly ratchets up the weights on various neuron firing thresholds. After repeated encounters, these weightings grow stronger until an instinctive impression is made. We don't necessarily remember the cause of each incremental ratcheting, for that's not necessarily "kept" by the brain. --top

Nah, I gotta complain about this one. There's a reason we write stuff down; for me it's a daily log, which I use to generate my weekly reports that get copied to the client. If you are having repeated bad experiences then you should be recording this stuff so that you don't repeat the same mistakes and so that others can learn from your bad experiences without having to suffer themselves.

Otherwise, how would we accumulate any worthwhile knowledge and glean wisdom from that? We absolutely need to write stuff down and share it with others in a cogent fashion. Depending on memory summing is non-predictive at best, and pretty amateur at worst. Eh?

I've probably encountered tens of thousands of such experiences, most small some big, over my career. That's not only a lot of stuff to write down, but also a lot of stuff to sift. Without some kind of database and multi-variable classification and search system (FlikiBase-like), it would be difficult to use such info even if it existed on paper. If you found a way to pull it off, I congratulate you. Perhaps one could only track the big stuff, but that would skew designs toward avoiding only big problems (SovietShoeFactoryPrinciple). A hundred small problems can outweigh the cost of a single big problem. -t

See: ComputationalBeautyOfNature, EmotionsVersusLogic, LinearOptimization, EvolutionarilyStableStrategy, ArtificialIntelligence, CollectiveIntelligence, SwarmTechnology

CategoryDiscussion

EditText of this page (last edited November 28, 2014) or FindPage with title or text search