Monday, February 8, 2010

Intelligence is evolution & evolution is intelligence

Evolution === Intelligence.

I don't mean this in any sense but the most absolute. The two have an absolute equivalency. Intelligence is an evolutionary algorithm and only an evolutionary algorithm. Evolution is intelligence at the very core of its essence.

I think it's strange that this idea hasn't found any traction or have been fully realized as a fundamental equivalence. I also find it to be completely understandable that it has not, because it seems to throw up red flags.

I figured it out 6/29/07 while blogging and narrowly managed to not dismiss it out of hand. It's actually a fundamentally profound idea and a seemingly obvious one that suffers a long history of near misses.

A lot of people seem to understand evolution is intelligence at a cursory level. I've watched as people come tantalizingly close over the years, while others have waxed poetically about obvious absurdities.

Gerald Edelman's neural Darwinism goes to great lengths to show how the seeming learning of the immune system via a microcosm could be expanded to explain how neurons work. And for as much promise as that idea has it seems to get bogged down in explaining neurons rather than piercing the fundamental equivalencies.

People point out how intelligently designing things requires a sort of tinker, tweak, and try methodology which sort of works like evolution, or how more generally technology evolves over time.

Jeff Hawkins goes on about how prediction is fundamentally what intelligence is about but fails to see what this means about the underlying algorithm. I read his book and looked at the source code for the work it spawned. A lot of promise and its ends up a basic Bayesian engine.

There's a 1995 paper from 15 years ago. That seems to get it (which I discovered a few days ago) but suggests using it with regard to neural networks and fuzzy logic.
http://65.44.200.132/Library/1995/Evolution_CI.pdf

In principle the truth is all of these ideas and none of them. Intelligence and evolution are two sides of the exact same coin. And to actually program a machine intelligence you need to evolve a prediction biosphere. How would a well functioning evolutionary algorithm whose population's successes and failures depend on the efficacy of their predictions about the outside world differ from an intelligent mind? I don't think they would.

If you were to codify predictions about anything and everything and evolve organisms whose success depends upon making accurate predictions, even if you could not see them as directly and obviously as equal to an intelligent mind, you should at the very least be given the jaw dropping impression of the power of such an idea.

One of the obvious problems with using genetic algorithms for anything more than making a really awesome antenna or critters to walk in a virtual world is that even the best solutions are still profoundly limited to their selection metrics and the sparsity of the environment. However, what couldn't you do with evolved predictions? You would have a set of understandings that could apply at a given time. In example of illusions two or more could apply but rarely at the same time.

With evolved predictions you could figure out the nature of the planets. You could identify a briefcase. You could use sound to help recognize something visual. You could go from the tiniest predictions about the distance of objects to the nature of the universe. After all, there's really no fundamental difference between a child playing and a scientist trying to figure out the nature of reality. There's also no real difference in going from a thoughtless baby to having figured out the most basic elements of the real world. One can go through most of Jean Piaget's developmental steps and understand why and how those understandings build.

You should also note that the idea answers a number of fundamental questions about how the brain works. Such a system would necessarily reverse the generally assumed information flow of senses and the brain. It isn't light hits your retina goes through some neurons and into the brain and then gets processed by the brain. Rather the mind is constantly making predictions and this light either confirms or denies those predictions. You don't need the image at all. There is no Cartesian Theater to watch the outside world through your eyes. You have already predicted what it should look like, and the only thing "seeing it" does is check if that prediction is correct or not. This is after all how evolution works. It isn't the DNA that contacts and interacts with the outside world but rather the body built by the DNA that either survives or does not.

This also demands that the brain is very plastic. That you should entirely be able to see again via electrical impulses on your tongue. You should be able to learn echolocation to see. You could quickly learn how different magnetic and electrical fields feel with a tiny magnet surgically implanted into your finger. If we allow for new predictions to be made, even if they seem rather absurd, if they are right they thrive and expand if they are wrong they die off. The first law of psychology is the second law of thermodynamics. We can learn skills quickly and effectively, even if we don't know we're learning them. We evolve a robust complete intelligence. We have a good "picture" of what the world looks like even without needing to really see anything in the world. There's no need to program anything about this internal landscape because evolution is a fantastic programmer (as I've seen many times first hand).

Much to my dismay, there is a lot of popularity for the Computational Theory of the Mind. That we have evolved programming and simply re-adapt it to a more modern context. But, there's no need to take the Pleistocene programming to throw spears and re-adapted it via fuzzy laws to do Euclidean geometry. After all, if we're already going to suggest that our brains have programs in them and these programs are evolved, why not evolve them again in real time? The idea ultimately relies on evolution making programs, and in fact there's no debate that evolution is at the core of human intelligence (regardless of what people think it is).

Why are we to conclude that this programing is somehow stored in our genes? After all, we don't have that many genes to make these programs. And why would these programs largely need a massive amount of development in order to perform? Why must children learn the basics in order to adapt the basics to other tasks. Why is childhood development so damned long before children engage these magical eye focusing programs? And why do they seem to actively learn so much? If so much of this is adapted from other bits of mentalese then why can't such things be coded by the brain from ground up. Why can't we all be Cartesian code houses? Solving the problems as they arise? And why can't we just use the same very powerful idea to solve all of these problems?

Further, when you realize the power of the idea, you also realize all the little things that cognitive science typically leaves out of the theories. Even the simple act of learning to focus your eyes involves a lot of prediction. You need to know which part of which eye fits with which part of the other eye and when those two points are actually further apart than expected it gives one the prediction that the perceived object is further away. While also accounting for the jiggling of the eye, and staring at different things, while not changing anything about the ones internal idea of what the world should look like. Your brain does a powerful amount of work that at some level everybody gives credit to evolution but somehow escapes people's that if its really evolution all the way down, all the problems go away.

William Paley is credited with one of the better versions of the Argument from Design, which in turn is probably the best argument in favor of a God (even despite it's shortcomings). The argument is admittedly an argument from ignorance. It says, basically when you look at a tree you know that it needs an explanation far different than what you could give for a stone because there's some kind of intellectual real work there. That there is an amazing amount of real design work that has gone into that tree and when we consider things like watches we equally well know that a lot of real design work went into them and they need a real explanation. He then "reluctantly" admits this is must be God. Then after Darwin, the evolutionists chime back that the equivalency is false. If I can explain exactly how the tree came to be using only natural well understood laws, it can't be intelligence. Therefore Paley is wrong and there's no equivalency between the intelligent works of man and the evolved works of nature.

This is wrong. It's true that the 'God conclusion' that Paley attempts is unwarranted but there is a conclusion you can draw and it is that the intellectual works of man and intellectual works of nature are cut from the same cloth. Establishing that evolution is the foundation of the works of nature doesn't set asunder the ingenuity of man might not also be evolution incarnate. After all, what kind of refutation is that: "This can't be intelligence because it's evolution." Why can't it be both! On one end we have a well understood, natural process, and on the other end we have a glorified and fundamentally unexplained skyhook. The only problem that seems to arise is that accepting that it is true demands that we finally have an explanation for intelligence.

The problem with the "intelligent design" movement isn't the claim that there is "design" but rather that "intelligence" is a solution. Intelligence isn't a solution unless it's on fundamentally firm ground. You can't claim that intelligence is the answer unless you can explain intelligence. After all, Alan Turing provided the only answer for the question "What is intelligence?" that was ever met with any good reception. It was pretty well Justice Potter Stewart's explanation of what obcenity is: "I don't know but I'll know it when I see it." It seems a bit silly but prudent to note that I know something that passes the Turing Test: an oak tree. People have for millennia been identifying the products of nature as being the product of an intelligent mind, maybe, just maybe, they aren't all completely stupid? After all it is admittedly rather naive to suggest that humans don't have a spidey-sense of design. That we can look upon the natural world and see real intelligent design work having gone into every twig and every critter while at the same time knowing that that it's evolution at work. Why must we tell ourselves that these are mutually exclusive? Why can't they be a very fundamental equivalency?

Nature is the product of non-random selection of randomly varying replicators. To say that because we understand this, it certainly can't be intelligent, is rather naive. If nature is so stupid, why do we steal many of our best ideas from nature? If nature is dumb, why are natures designs so much better than our designs?

“I said it was impossible to look at these without seeing that they were the effect and the expression of mind. I shall never forget Mr. Darwin’s answer. He looked at me very hard and said, “Well, that often comes over me with overwhelming force; but at other times,’ and he shook his head vaguely, adding, ‘it seems to go away'” - Duke of Argyll (F. Darwin 1888, I:316)

With much being played at via this observation of intelligence in nature from Paley's arguments to William Dembski arguing that there's some amount of "Specified complexity" where evolution no longer suffices to explain design and only intelligence on the other end of this same fuzzy line can manage to explain anything (while explaining nothing, as it taken as a hollow skyhook). There's no limit to the amount of obvious interplay between the jaw-dropping power of evolution proper and the cunning wit of sharp intelligence. And while we noddingly agree with things like Orgel's second rule "Evolution is cleverer than you are." we don't take it seriously, and we should.

The truth will set us free. Nature's design, nature's intelligence, and nature's brilliance are not merely illusions they are reality and the production of evolution and evolution is intelligence. Evolution is the real intellectual work of nature. If we are to manage to have any hope at all of programing intelligent computers to do real intellectual work, we need to know that intelligence is to do more than code a tiny subset of poorly executed solutions. We need to make the solution to solutions, and with evolution this is exactly what we have. With evolution working on prediction space, we'd can answer everything. Until we see life itself as the most basic form of intelligence, we stand little chance of programming anything of note in Artificial Intelligence (a term which itself makes about as much sense as artificial evolution.)

It has been said that geology is just physics in slow motion with some trees. Well, biological evolution is just intelligence in slow motion carried out in nature, rather than in a skullful of neurons. They are the same exact thing. And the sooner we pay attention to that fact, the sooner we can use the universal acid of evolution to melt away our problems with intelligence.

"In the distant future I see open fields for far more important researches. Psychology will be based on a new foundation, that of the necessary acquirement of each mental power and capacity by gradation." - Charles Darwin, Origin of Species




Googling: "intelligence is evolution".

  • "The cause of human intelligence is evolution"
  • "But the essential ingredient in using these structures in computational intelligence is evolution: population-based search with random variation and selection."
  • "If intelligence is evolution squared then what comes next will be expolution"
  • "One major advantage that animal intelligence has over robotic intelligence is evolution."
  • "Intelligence is evolution's unique gift to humans."
  • "The cause of human intelligence is evolution."
  • "The seeking for Intelligence is Evolution and bangers big or small"
  • "less sophisticated forms to complex life (or even intelligence) is evolution."
  • "Intelligence is evolution on steriods."

Googling: "evolution is intelligence"
  • "It was these extensive research areas that later became the watershed for his book, “Nuclear Evolution is Intelligence, and Intelligence is Life”"
  • "As the major survival element in human evolution is intelligence, the conclusion that the higher evolved race, the Caucasians, possessed a superior"
  • "The biggest issue I have with evolution is intelligence. If we arrived over millions of years of chaos, chance, uncertainty and random processes"
  • "The key to this evolution is intelligence and information management. In order to hit a target,"
  • "The positive tool of evolution is, intelligence used with honesty, love, understanding, compassion, sensitivity and kindness."
  • "Evolution is intelligence/information accumulation. Development is intelligence/information preservation. Evolution causes ongoing unpredictability"
  • "I'm inclined to agree with Idiocracy: the determining factor of procreation (and therefore evolution) is intelligence, though I would add"
  • "A more likely sign of Human evolution is intelligence level."
  • "Would they argue about evolution? Is intelligence a blessing or a curse?"
  • "The most important trait in human evolution is intelligence."
  • "I agree that the pinnacle of evolution is intelligence"
  • "So, I guess that what I am saying, is that evolution is intelligence at work. Not there yet. What we have is mechanical principles at work"
  • "Evolution is intelligence/information accumulation. Development is intelligence/information preservation. Evolution causes ongoing unpredictability"
  • ""Evolution is intelligence denying intelligence." Sorry to ruin your party, guys."
  • "If evolution is intelligence then why can't we identify it? -- Some have claimed evolution to be the cause behind design and yet who has isolated the scource of this ability for evolution to think or built?"
  • "'Evolution Vs. Intelligence?' - Evolution is intelligence. -- We think we're intelligent, so we look at nature and some think that it shows intelligence, while others think it is down to evolution. The basic principles of both intelligence and evolution are the same"
  • "Details of Evolution Is there intelligence behind it ?"
  • "Further, you make an implicit assumption that the 'aim' of evolution is intelligence – that is completely untrue."

This is every example I've can find of those words together. And the solutions themselves seem so very satisfying that I cannot fathom that they are out there, in earnest, and I've somehow never heard of them. Please consider the idea, and please tell me what you think.

3 comments:

Tracy said...

Have you thought about the specifics yet on your idea in terms of either programming and/or actual brain physiology? How do you implement this without it becoming a “Bayesian process”? Every time I think it through, it seems to me that during the process of comparing the prediction to the sensory input etc. that in the end it will turn out to be just another “Bayesian Process”. I must be missing something in my understanding of how you are envisioning the evolutional process in such situations as this. Would you say that a better understanding of the immune system process of which you spoke provide me with the answer to this?

Tatarize said...

Yeah, you don't run it like that. Bayesian processes tend to be the best way to sort of predict things but really it needs to run via evolutionary algorithms. Bayes is great, but it's a fundamentally stupid process. Evolution is the key. Prediction is important but it's not the difference between functioning and non-functioning.

Also on top of my blog there's a link to a thread with enough interaction to make my thoughts on the matter far more clear.

Tracy said...

I figured that I will have to do quite a lot of catching up to do before I will really grasp a clear understanding of what you are proposing. The good thing is that I am fascinated by what you are proposing, especially at the level of implementation in programming and/or hardware. I will start systematically going over the resources you have provided. It will take me a while to work through it all because I am not as quick at processing information as I used to be. I may be calling on you for help on clarification of things from time to time.

At the risk of being presumptuous, I would like to make a suggestion. If I were in your position, I would seriously think about going back to university to get a degree in cognitive science focused on the physiology of the brain. This would put you in the unique position of having a solid background in these two disciplines, computer science and cognitive science, which, IMO, would let you better implement your ideas. Even if you disagree with some of the conclusions that have come out of cognitive science so far, it would give you a complete understanding of how and why they have come to these conclusions and possibly put you in a position to make corrections. More importantly, having a background in both fields would put you in a position to bring the two fields into a better cooperative union which might eliminate some of the misdirection that exists in both fields. You, with your mind’s thinking and perceptual skills coupled with your fresh ideas, there is no reason why you could not easily direct these fields toward new and better directions. I, myself, had been seeking a degree in computer science with a minor in psychology focused on cognitive science with the hope of putting myself in a good position to better bring the two fields together, if not outwardly then at least inwardly within my own mind. Because of my situation, I doubt that I will ever become productive enough to seek work in these areas. That is if I even finish my degrees which is not looking too promising at this point. lol