Tuesday, February 3, 2015

A New Equation for Stupidity

In a TED talk filmed in November 2013, Alex Wissner-Gross offers "A new equation for intelligence".  I'm not sure if there was an 'old equation for intelligence'? The talk was based on the concepts of a research paper published in April 2013: Causal Entropic Forces. I must confess that the mathematics of this paper is way beyond my understanding. However, many of the statements made in the TED talk simplify the concepts to the point where we can understand and make sense (or perhaps nonsense) of them.

According to Wissner-Gross, "Intelligence is a force F that acts so as to maximize future freedom of action. With strength T, with the diversity of possible accessible futures S, to some future time horizon 'tau'."  This is a complex statement, but in does make some sense.  In a much simpler way, we might say that "Intelligent things are things that attempt to optimize and maintain their future flexibility, their future options."

Wissner-Gross describes a number of experiments that demonstrate this theory and support this formula for intelligence.  The experimental descriptions in the TED talk, and in the research paper are deliberately simple. They are designed to 'optimize or maintain their options for future states'.

How can this be done?  You can only create a system that can "optimize it's future options", if you make the system, in some way,
 - aware of it's current state in some trivial representation
 - able to make choices and to take actions (modeled as entropy)
 - and able to count resulting future options, eg. aware of ALL OF the consequences of it's choices.
and one other point, which is not discussed by the Wissner-Gross presentation, ignorance of external systems and variables.  We shall come to discuss this later.

In other words, Wissner-Gross created a system with a fundamental level of 'self consciousness' although no sense of 'free-will'. They are computer programs designed, driven to seek actions that 'optimize future options' by creating 'simulated entropy' models. There is no free to chose other options.

What Wissner-Gross observed was that when we build a self-aware simulation and force it to 'choose', actions that result in the most 'future possible choices', we get a system that appears to be 'intelligent'. Is it intelligent? We might say that, if it looks like a duck and it walks like a duck and quacks like a duck, it's a duck. If it looks intelligent, and acts like it is intelligent, then it is intelligent. Or maybe not?

Wissner-Gross gives us a mathematical formula, or equation for intelligence: Intelligence is a force F that acts so as to maximize future freedom of action. With strength T, with the diversity of possible accessible futures S, to some future time horizon 'tau'.

F(intelligence) = T ∇ Sτ

It's an interesting formula. Intelligence the force, the drive that results when an entropic system falls into a state where it naturally chooses actions that maximize future options. When we consider that for a moment, we can rephrase the formula. We originally described it as:

F(of intelligence) = T ∇ Sτ

We might instead view it as:

F(of healthiness) = T ∇ Sτ

Thus, the force of healthiness acts so as to maximize future freedom of action. With strength T, with the diversity of possible accessible futures S, to some future time horizon 'tau'.

We can look at this formula from another perspective.  Maybe it's the formula for 'life'?

F(of life) = T ∇ Sτ

The life force is the tendency for systems that make choices that increase their future options - to live longer and build complexity. Health is a result of life and does not exist without life. Life, intelligence, and healthiness emerge and become more complex as the system becomes more complex. Intelligence and healthiness are different ways of looking at the same thing. The systems created in Wissner-Gross's simulations are not 'alive', and they are supported by external decision systems that make their choices. Real life evolves from simple, similar situations, that facilitate a rise in complexity.  That 'rise in complexity' occurs when simple systems encounter other systems with similar and complimentary properties and discover, or evolve, the ability to cooperate with other systems.

Life is a result of freedom to choose, including choosing to cooperate with other entities (systems) that have freedom to choose, thus increasing the freedoms of all involved.

Which came first: intelligence or life?  Is it a chicken-egg question? Or does it depend on how we define intelligence, and how we define life?  Does life, or does intelligence exist as soon as a system gains 'freedom to choose'? Or does 'freedom to choose' arise out of system complexity, as it approaches 'life' and 'intelligence'? Can intelligence exist without life? Can life exist without intelligence? Or does 'freedom to choose' simply not exist, except as an illusion we believe, because our life systems are so complex that the give us the illusion of choice?

It is also interesting to consider the 'time horizon', because different time horizons can lead to very different results - in more complex systems. An action that might be intelligent when considering a daily activity, might be stupid when we consider it from a month, year, or lifetime perspective.

And that leads us to the question of 'stupidity'.  If we have a formula for intelligence, is there a formula for stupidity? Is stupidity the inverse of intelligence, and intelligence the inverse of stupidity. Is a rock stupid? No. A rock is dumb, because it cannot hear, it cannot see, it cannot speak, it cannot think or decide, and it cannot act.  But it is not stupid, because it cannot make stupid decisions nor take stupid actions. Stupid things (or stupid systems) are things that take stupid actions.

What is a stupid action?  What is stupidity? If "Intelligence is a force F that acts so as to maximize future freedom of action. With strength T, with the diversity of possible accessible futures S, to some future time horizon 'tau'."

Then Stupidity is "a force F that acts so as to limit or minimize future freedom of action. With strength T, with the diversity of possible accessible futures S, to some future time horizon 'tau'."

F(stupidity) = T ∇ Sτ

Intelligence is a set of actions that are deliberately considered and chosen to keep the system alive and as active as possible.  Stupidity is the set of actions that are deliberately considered and designed to shut down the system.  Entropy is naturally stupid. It works inexorably to diffuse energy in a system.  But stupidity is even worse. Stupidity is 'system suicide'.

Now we can see that something is clearly wrong.  In real life, stupidity is not 'suicide'.  Stupidity does not lead directly to death - in real life. Sometimes it even does quite well, thank you very much. Stupid luck exists.

In the model of healthiness and unhealthiness discussed in the post Embracing Unhealthiness, we recognized that healthiness is linked to unhealthiness such that the sum of healthiness and unhealthiness equals 100 percent of our health potential.  The sum of healthiness and unhealthiness is your 'potential for healthiness'.

What about the sum of intelligent choices and stupid choices? Now we can see the cracks in this model of 'intelligence'. The model is not 'free to choose', it can only choose what is calculated as optimal in terms of future freedoms within the specified time period.

It reminds me of an old joke about an uncle, and a small boy.  The uncle calls over one of the relatives and says "Watch this!". Then he turns to the small boy and asks "Would you like to have this nice, shiny silver dollar, or would you prefer this old, dirty, crumpled up five dollar bill?" The young boy chooses the shiny silver dollar.

But when the relative talks to the boy later, and asks "Don't you know that you could buy 5 silver dollars with that bill?", the young boy answers "Of course I do.  But, as soon as I take the paper money, he'll stop offering me the silver dollar."

The boy has figured out something intelligent.  He does not make the 'optimal decision'.  He does not have a specific 'time horizon' in mind.  He would prefer that there is no time horizon, as long as he keeps getting paid.  He has used his intelligence to look 'outside the system'.  This is one thing Wissner-Gross's models cannot do.  They are designed to take all of the information available, and calculate the best option. There are no external variables, no external facts.

Intelligence is making good decisions when you DON'T have all the facts. If you have all of the facts, you don't need intelligence to decide - a machine can decide.

There is another serious flaw in the Wissner-Gross model of intelligence. Real intelligence must acknowledge that other intelligences exist, learn to cooperate with them, and to compete with them. One of Wissner-Gross's models seems to present 'cooperation', but it's not cooperation between two systems, where each gets to 'choose to cooperate or not', it's simply a complex model that searches for the best solution, and as a result, looks like it is cooperating.

The goal of a real intelligence is to understand more.  And when it understands more, to find better questions. And sometimes, the goal of a real intelligence is suicide.  We are all going to die.  Some of us want to choose when and how to die.  The Wissner-Gross model of intelligence ignores this aspect of intelligence.

The Wissner-Gross model is missing two more fundamental elements of intelligent decisions.  Memory and risk.  A real intelligent system has a memory of past decisions, and can make rapid decisions, when necessary - based on memory, not on calculation.  Decisions based on memory free up the calculation parts of the brain for decisions where time is not a priority. Risk assessment is also a fundamental factor in intelligence.  What good does it do to choose the 'most future options' if what you are choosing is the number of bee stings you might receive?

If the Wissner-Gross model is not 'intelligence', what is it? It is a formula for self interest, based on rational calculations. The formula can drive the system towards a longer life, a more stable life. Maybe even a more boring life. But, boring can be intelligent if you are an accountant, trying to make money.

In real life, we often use our intelligence to rationalize our decisions. And that's a good way to succeed. If we attempt to only make 'rational' decisions, that are rationalized before the decision is made, we will make very few decisions - because all of the facts are not available.

The Wissner-Gross formula is a powerful tool for decision making.  Much like a calculator, or a spreadsheet.  But it is not a formula for 'intelligence'. Intelligence requires irrational thought.

This is a blog about health freedom. What has this got to do with health, and health freedom? There are some powerful implications for freedom, and also for health. Freedom to choose increases our ability to make intelligent decisions. Constraints on freedom reduce our ability to make intelligent decisions, resulting in more stupid decisions. This is not stated directly by the Wissner-Gross model, because the model is not actually free to choose.  It can only choose the option that has the most future freedoms.

In a similar fashion, we can also see that the sum of intelligence and stupidity (unintelligence), the sum of decisions that maximize future options and decisions that minimize future options in any system - the total potential for action in any system is 100 percent.  None of the systems created by Wessner-Gross exceeded their systemic limits of intelligence. None of the systems exhibited 'exceptional' intelligence. This might seem obvious, but it is worth understanding.  No single system can exceed its own limits of intelligence.

The best way to increase intelligence is to increase the complexity of the analysis, not to limit it.  The best way to increase intelligence is through communities, not rationalization by a single individual. The best way to improve healthiness is to recognize that healthiness is not just an individual trait, it is also a measure of the success of our societies, our communities. The best way to create healthy communities, is to work to create healthy communities, not to create selfish independent models of intelligence.

to your health, tracy

Tracy is the author of two books about healthicine: