The Motivation of Genius, Part I

2

January 2017
Ivan Obolensky

What drives us to excel? What motivates our thoughts and why does discovery move in certain directions, while not in others? How much of what we do is the result of what we think we decide, and how much is instinctual? Put in another context, how much of our behavior is hardwired and how much is the result of our personal learning software? These questions will be answered over the next few articles. By trying to answer them, we can gain insights about our lives, the world we live in, and our future as a species.

To begin, let’s look at certain behaviors in nature.

Locusts belong to the order Orthoptera (from the Greek ortho, straight and ptera, winglet). Grasshoppers, katydids, and crickets are part of this order. They have three main stages of development: egg, nymph, and adult. Female locusts drill holes in the ground and deposit ‘pods’ of eggs surrounded by a protective froth. As nymphs grow, they add segments. After each segment is made, the nymphs shed their skins eventually reaching their adult stage.1

Locusts are a species of short-horned grasshoppers.2 They exist in two different behavioral states: solitary and gregarious. When population densities are low, their color is brown or greenish. When densities increase beyond a critical threshold, crowding causes their hind legs to bump against each other and triggers the gregarious phase. Serotonin levels spike by a factor of three resulting in color changes (they turn black and yellow). Individuals undergo a 30% increase in brain size and release pheromones that cause the individuals to be more attracted to each other. Gregarious behavior is induced by visual, tactile, and olfactory cues. If an individual is removed from the swarm, he/she will revert to the solitary phase and lose the distinctive gregarious black and yellow markings.3

We know that our human ancestors experienced a similarly rapid increase in brain size over successive generations when our species was in its infancy. (See “Why Language?”) Locusts require enhanced visual and olfactory processing capabilities when operating in a swarm, hence the need for larger brains. Whether cooperation in humans required similar changes in our brain structures, and why it happened are interesting questions. As with locusts, density factors may have played a part.

Density tipping points appear elsewhere in nature. Diseases require minimum host population sizes in order to exist. Should population size decrease sufficiently, a disease will no longer be able to sustain itself. In addition, host populations often develop immunities, reducing the number of potential hosts.4

Looking at population density levels, there appear to be two tipping points. One at the lower band and one at the upper band. The values that determine these tipping points and the behavioral changes that result are different population to population, due to many factors.

Female Artic brown squirrels of the boreal forests of Northern Canada alter their reproductive output when population densities reach a critical level. At very high densities, females shut down their reproductive systems and only reactivate them when densities have subsided and environmental conditions are better able to sustain the population. One of the determining factors is the mother’s food availability. Hungry mothers lead to fewer youngsters.

In the above instance, the size of the population and the proximity of others of the same species induce changes in individual behavior, not the other way around, as is often thought.5

Group size, the interaction of individual elements, and the phenomena associated with them, are part of the study of complexity.

Complexity theory postulates the existence of emergent phenomena or behavior. Emergent behavior arises through the interaction of smaller or simpler parts that when combined, exhibit properties that the smaller parts do not.

An example is how oxygen and hydrogen can form water or hydrogen peroxide. Both are made from the same elements, but each compound has very different characteristics. Table salt is made of sodium, a metal, that is so reactive it must be stored in oil to prevent it coming in contact with water. Should it do so, it bursts into flame. Chlorine is a gas that is so corrosive it was banned after World War I as a weapon. The combination of the two is the inert compound, sodium chloride, that we use every day.

Emergent behavior has several characteristics. One of which is novelty: a new attribute that is unpredictable as an outcome given the elements that make it up. Many commercial products such as Nylon, Mylar, and other household products are the result of a great deal of experimentation to discover the emergent properties that we might find useful. Many were discovered by accidental combinations of elements and processes.6

Coherence is another characteristic of emergence. The emergent property maintains itself over a period of time. It exhibits a wholeness with its own characteristics.

Hurricanes and crystal formation have emergent properties. Temperature is also considered an emergent phenomenon, the result of collisions and oscillations of millions of molecules.

Emergence also exists in nature in the form of group behavior.

Swarming, flocking, and schooling, as well as crowd patterns in humans, such as traffic jams, are examples of emergence in biological systems.7

In the case of locusts, how do they learn to swarm?

We do know that conspicuous coloring, smell, and better vision allow the swarm to hold together. Other than as a means of survival, the need to form such a complex structure is not clear. It appears to be instinctual.

Instinct is a behavior exhibited by most living organisms that have to survive in changing environments. Organisms must learn to cope as individuals and as a species. Higher order animals require skill sets that are either hardwired or learned. Learning, too, is probably instinctual.

Instincts are another area that is not fully understood, although we know certain things about them.

An instinct acts as an inductive bias. It is also called a learning bias. It is the set of assumptions that the learner uses to judge whether an action is good or bad. A simple one for an infant might be to feel full. The infant might not know how that is done at the outset, but when it connects moving its mouth and swallowing with feeling full, it gets the idea. Inductive biases are an essential part of all learning processes.

To learn fast, a system must have more biases.

What does this mean?

Suppose we create an artificial intelligence system that contains billions of neural networks. We then hook it up to a supercomputer. In fact, we go all out and hook it up to three supercomputers working in parallel. Spending over a billion dollars on computation alone, we have potentially the most sophisticated AI system ever devised. We plug it in and press ENTER. What happens? Nothing.

How come?

We did not tell it to do anything. So we program it and tell it to learn. We hit the ENTER key again and wait. We wait and wait and wait. Still nothing. We examine our process and discover that we needed to give it a hint, a direction. We have to give it a bias. We are of two minds about this. We want it to learn everything but if we give it a bias, we immediately restrict what it will learn. It is a paradox. By saying what it is we want it to do, we automatically bias the computer. It no longer learns in a direction other than in the direction we decided.

For example, we have a robot. It just sits there so we tell it to move around and explore the maze it’s in. It moves. By moving, it no longer learns in the direction of being still. It has a bias. It moves and gets feedback as it crashes into walls. It never considered learning about colors or poetry. It couldn’t. It might be able to if we have it wirelessly connected to our billion-dollar machine, but we’d have to bias it to learn in that direction, and we didn’t. We gave it a hint to move instead.

This is the fundamental paradox of all learning. To begin, we have to set limits or boundaries.

Over time, a system can explore other areas, but learning is experiential, and no matter what we think about it, our initial jog in a specific direction is a bias that has been introduced. This is not necessarily a defect.

All babies, regardless of species, come with a large number of biases. If they didn’t, they couldn’t breathe, feed, move, or even sense what moving is. This we call instinct, and what hardwired-instincts do is extraordinary. Instincts give us a purchase on our world — a genetic leg-up so that we can eventually find ourselves in the saddle of our own lives. Without them, we’d be motionless, and chances are, we’d be dead.

We live in a world that requires energy. The fewer the biases, the longer it takes for a living entity to make sense of the world and act, and the easier it is to be eaten by another with more inbred biases that allow it to learn faster and act more rapidly. It is also prey to those creatures that have lived longer and learned more.

Make no mistake — biases, instincts, and hardwiring are required for survival.

Nature has used eggs as a means of allowing the young of a species to form sufficiently to survive, but once they hatch, what should the new entity do, and how will it do it, given that the environment presented is unknown? It must have a hardwired action plan while the processing part, what we consider the mind, goes along for the ride and learns as much as it can.

In this, we have the inherent difference between hardware and software. Hardware will always give direction, even if it is only designed to initially load the software. Yet even software needs an operating system to function. In the past, when operating systems failed to load, the user was prompted to use a boot disc. Operating systems today are more sophisticated and have their own internal boot disks, but without that initial nudge, all our computers would be ornaments, rather than functioning machines.

Hardware — or in the case of humans: bodies — introduce biases because they are designed that way and with good reason. Survival demands it.

In an ideal world, it might be possible to have no biases, but it is likely that any learning machine must have them. Even neural networks require learning periods and data to practice on. The nature of biases applies just as much to artificial intelligence.8

Humans are concerned that AI might one day control the world. With an understanding of biases, it seems unlikely. Control implies connection, and to control the world, any AI, or superintelligence, must be connected and be able to interact with it. It is therefore biased in that direction, and its weakness becomes obvious: disconnection and isolation.

It is possible to build self-sufficient killer AI robots, but they too will eventually encounter the same problems we do. Hunter-killer robots have a built-in bias, or instinct, to destroy humans or other robots. If they are actually intelligent, what do they think will happen when they can no longer find humans or other robots to destroy? Now what? It is also worth noting that population densities would determine the asking of that question.

Given our fears of AI, why are we motivated to channel our genius to create such machines in the first place?

Perhaps, that too is instinctual.

By definition, complexity requires large numbers of elements.

In the first half of the 21st century, the human population now numbers in the billions and lives in urban environments, packed close together, in an interconnected economy that spans the globe.

Given that we have sufficient numbers and our behavior is complex, what emergent behavior can we observe today?

Traffic and high volume of interactions needed are two.

It is an annoying aspect of modern life. Whether on the road, in an elevator, in a convenience store, in the market place, at work, at play, even in the virtual world, we are surrounded and jostled by others of our kind in extraordinary numbers.

In the dim past, humans grouped together for survival in bands of around fifty. To cooperate in that environment required communication skills, a close connection to our surroundings, and to each other. We also needed more of us. Survival depended on having enough humans around to survive at all.

In that early environment, the Internet, Facebook, social media, selfies, Pokemon Go, would have been dangerous. Looking at our phones for long periods of time rather than being aware of our surroundings would have resulted in us being lunch for other species. Survival required vigilance.

Humans banded together back then because the choice of not doing so would have led to extinction.

Today we are banded together not by choice, but because we must. There is a difference.

One phenomenon we see today is declining birthrates in developed countries. Is this an emergent phenomenon?

Most likely it is.

In locusts and Artic brown squirrels we have seen two methods that nature has evolved to deal with population densities. The species either migrates to find new areas to inhabit or controls its numbers internally. Either solution solves the problem.

The question is how does a particular species go about it?

In the case of humans, I would like to offer an observation:

The widespread proliferation and direction of the technologies we are using and developing, as well as the social phenomena we see around us today, form our species’ answers to the question asked above. We are moving from the real to the virtual with giddy abandon. The question is why?

The title of this article is the motivation of genius. It took genius to create the electronic infrastructure, devices, and systems necessary to sustain it. What prompted this direction? More fundamentally, what prompts genius to create in the first place? Does induced behavior (instinct) play a part? Although difficult to answer, there is some evidence to support that instinct does.

The reasoning behind this idea will be outlined in the next article.

 


  1. A. (2015). About Locusts. Retrieved January 9, 2017 from http://www.agriculture.gov.au/pests-diseases-weeds/locusts/about/about_locusts
  2. Britt, R. R. (2009). Grasshoppers vs. Locusts: What Makes a Swarm. Live Science. Retrieved January 9, 2017 from http://www.livescience.com/7782-grasshoppers-locusts-swarm.html
  3. Ott, S. R. and Rogers, S. M. (2010) Gregarious desert locusts have substantially larger brains with altered proportions compared with solitary phase. Biol Sci. Retrieved January 9, 2017 from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2982065/
  4. Desowitz, R. S. (2015) New Guinea Tapeworms and Jewish Grandmothers, Tales of Parasites and People. New York, NY: W. W. Norton & Company.
  5. A. (2000) Animals Regulate Their Numbers By Own Population Density. Science News. Retrieved January 9, 2017 from https://www.sciencedaily.com/releases/2000/11/001128070536.htm.
  6. Georgescu-Roegen, N. (1971). The Entropy Law and the Economic Process. Cambridge, MA: Harvard University Press.
  7. Surowiecki, J. (2004). The Wisdom of Crowds: Why the Many are Smarter than the Few and How Collective Wisdom Shapes Business, Economies, Societies, and Nations. New York, NY: Doubleday.
  8. Hall, J. S. (2007). Beyond AI: Creating the Conscience of the Machine. Amherst, NY: Prometheus Books.

Please click here to sign up for our monthly updates or notifications of articles.

Interested in reprinting our articles? Please see our reprint requirements.

© 2017 Ivan Obolensky. All rights reserved. No part of this publication can be reproduced without the written permission from the author.

  1. Silvia
    Silvia01-13-2017

    Ivan, as always very interesting information and ideas I was not aware of.

    Real thanks for sharing your knowledge and conclusions, it is very valuable.

    Keep it on.

    Silvia Llorens

Leave a Reply