Is evolution intelligent?

[Mr Skeptic]

People sometimes talk about the design in organisms, but when pressed say it is metaphorical. But is it? We use evolutionary algorithms to create some of our designs. Now people have much less hesitation to call said products design, because intelligent humans designed the algorithm's parameters and ran it on computers that are also the product of our intelligence.

 

Evolution could be said to be a very narrow form of intelligence but with great capacity: it can only solve one problem. Some may instead call it "smart" rather than "intelligent".

 

Sayeth wiki:

Intelligence is an umbrella term used to describe a property of the mind that encompasses many related abilities, such as the capacities to reason, to plan, to solve problems, to think abstractly, to comprehend ideas, to use language, and to learn. There are several ways to define intelligence. In some cases, intelligence may include traits such as creativity, personality, character, knowledge, or wisdom. However there is no agreement on which traits define the phenomenon of intelligence agreed upon by a majority across the various concerned disciplines.

 

Problem Solving: The capability to solve problems is one of the central aspects of intelligence. The problem evolution solves is "design an entity that can produce surviving offspring". It should be noted that for all the intelligence of humanity, we do not even come close to being able to do that.

 

Learning: The evolutionary process can be said to learn; its memory is in the form of DNA that is dependent on the past. Stuff is also forgotten. Learning, or course, being considered another of the most important aspects of intelligence. However, the learning is again very narrow; evolution cannot learn a better algorithm.

 

Planning: Evolution does not plan, per se. However, some things from the past will be useful in the future despite not being useful in the present and are retained in the gene pool for a while at least.

 

Thinking, Reasoning, Comprehension: not so much.

 

Capacity for Language: We have DNA, RNA, and amino acids. These are sometimes compared to languages.

 

Creativity: The creativity of the evolutionary process is quite broad; however it is limited to being close to a local maxima.

 

Anyhow, what do you people think? Is evolution intelligent? Is evolution the Intelligent Designer?

[iNow]

No. Some changes are good. Some changes are not. No changes are intelligent.

 

Don't we have like three of these threads already active?

[Mokele]

Using such definitions to claim evolution is "intelligent" is much like saying that, since it can grow, metabolize, reproduce, etc., then fire is a living organism.

 

Intelligence itself is very hard to define. For instance, "language" is tricky, especially when including animals, since I've known completely asocial species which display problem-solving ability on a nearly primate level.

 

"Problem solving" is one I'd dispute - evolution solves problems by dumb luck (stumbling on the right mutation) and brute force (sheer number of individuals and duration of time).

 

"Learning" is another - even sponges can learn, and they don't even have nerves, let alone brains or intelligence.

 

Even if we grant "learning", that's not the same thing as "history dependence", which is what evolution shows. If you stretch a muscle, it contracts more forcefully that otherwise, so are muscle cells intelligent? No, they just show history dependence.

 

I'd actually dispute even evolution's creativity - developmental biology constrains many organisms to relatively limited morphological spaces. Plus, creativity implies the ability to deliberately come up with new solutions, rather than simply stumbling across them during brute-force high-throughput trying everything.

[Horza2002]

Well its not really design as every change that occurs to an organism goes through every minor intermedaite along the way. If it was designed, the new changes would simply appear. Evolution also uses the huge numbers of organisms to help its process

[Ophiolite]

"Problem solving" is one I'd dispute - evolution solves problems by dumb luck (stumbling on the right mutation) and brute force (sheer number of individuals and duration of time)..

Yep. that's how we problem solve at my company. And we're number one.

 

Well its not really design as every change that occurs to an organism goes through every minor intermedaite along the way. If it was designed, the new changes would simply appear.

You haven't worked in an engineering company, have you? Today many of those changes may tend to appear only in a software model, but still many make it to various forms of field testing and post launch modifications.

[Fuzzwood]

And we of course have to go to the site to view it as opposed to you giving a summarized version of it?

[Sayonara]

Evolutionary processes fall outside the scope of the domain which the term "intelligence" encompasses.

 

This is evident in the first line of the definition you posted: Intelligence is an umbrella term used to describe a property of the mind...

[Mr Skeptic]

Evolutionary processes fall outside the scope of the domain which the term "intelligence" encompasses.

 

This is evident in the first line of the definition you posted: Intelligence is an umbrella term used to describe a property of the mind...

 

But, if you consider evolution an algorithm and the universe the computer it runs on, wouldn't that potentially put it at the same level as A.I.? When run on a computer, for example, an evolutionary algorithm is considered A.I. But what about the original algorithm?

[BhavinB]

Although evolution has been said to be a mechanism for increasing 'information', saying it is intelligent might be a stretch.

[Sayonara]

But, if you consider evolution an algorithm and the universe the computer it runs on, wouldn't that potentially put it at the same level as A.I.? When run on a computer, for example, an evolutionary algorithm is considered A.I. But what about the original algorithm?

It depends on the domain.

 

Evolution as an algorithm is an intelligent way to approach a problem, but is the algorithm itself an example of intelligence?

[bascule]

No, evolution is not intelligent. Intelligence is inherently predictive. Natural selection has no predictive power; it merely creates random variations and the fittest variations survive and reproduce the best.

[marrykonta]

Hi Mr Skeptic,

 

Well,Thanks for providing good information here in these forum.I like the article and its way of presentation.Designs always contain patterns, but natural patterns never contain designs or symbols. There is a vast chasm between the most complex natural pattern and the simplest design. All designs start with symbolic representation of ideas through a code, which is always designed by a mind.

 

 

 

Thanks

[Mr Skeptic]

It depends on the domain.

 

Evolution as an algorithm is an intelligent way to approach a problem, but is the algorithm itself an example of intelligence?

 

Good point.

---

Merged post follows:

Consecutive posts merged

No, evolution is not intelligent. Intelligence is inherently predictive. Natural selection has no predictive power; it merely creates random variations and the fittest variations survive and reproduce the best.

 

So if I get what you're saying, using evolution in a model would be intelligent (predicts fit adaptations), whereas evolution itself would not. Much like using weather systems in a model would be intelligent (predicts future weather), whereas the weather itself wouldn't.

[bascule]

So if I get what you're saying, using evolution in a model would be intelligent (predicts fit adaptations), whereas evolution itself would not. Much like using weather systems in a model would be intelligent (predicts future weather), whereas the weather itself wouldn't.

 

Something which makes predictions based on a model isn't intelligent either. It's no more intelligent than knowing where pool balls will wind up when you hit them from a particular angle with a particular force.

 

To truly even begin to be considered "intelligent", a system must infer cause or otherwise make associations/classifications based on an arbitrary set of input data. Systems that manage to do this are certain Bayesian approaches, Numenta's NuPIC, and to a certain extent collaborative filtering based on a least squares fitting, and the Markov Chains used by Google as the basis of PageRank.

[D H]

But, if you consider evolution an algorithm and the universe the computer it runs on, wouldn't that potentially put it at the same level as A.I.? When run on a computer, for example, an evolutionary algorithm is considered A.I. But what about the original algorithm?

 

Whoa there!

 

First off, evolutionary algorithms are considered AI because the underlying concepts were developed by the AI community. Just because someone in the AI community developed some technique does not mean the technique itself is "intelligent".

 

 

There is a huge difference between evolutionary algorithms and biological evolution. You cannot just pick up an evolutionary algorithm off the shelf and tell it "go solve this problem". You have to numerically define the goal, you have to come up with a scoring algorithm that assesses the distance between a tentative solution and this goal, and you have to specify various parameters that control the evolution. Evolutionary algorithms in and of themselves are not intelligent. They are just math. There is always an intelligent designer, a person, behind EAs when EAs are used to solve a problem.

 

As has been discussed right here in this forum far too many times, biological evolution does not need an intelligent designer to function.

[forufes]

 

Planning: Evolution does not plan, per se. However, some things from the past will be useful in the future despite not being useful in the present and are retained in the gene pool for a while at least.

that is one of the most important points i hold against a "dumb" random evolution..

 

many mutations are not beneficial in them selves, but when combined with another mutation that comes afterwards, they become beneficial..so what made the trait survive the first "round" of natural selection to team up with the second trait and become useful?

 

 

No. Some changes are good. Some changes are not. No changes are intelligent.

keeping the good changes and rejecting the bad is intelligent.

[JillSwift]

many mutations are not beneficial in them selves, but when combined with another mutation that comes afterwards, they become beneficial..so what made the trait survive the first "round" of natural selection to team up with the second trait and become useful?

If the first mutation was not harmful, or carried a benefit of its own, it would not have been "selected out".

 

keeping the good changes and rejecting the bad is intelligent.

 

So, I want to take this bunch of sand I have and keep the smaller particles and reject the larger. So, I run it through a fine wire mesh.

 

The mesh selects for the smaller particles of sand, and rejects the larger.

 

The wire mesh is intelligent, by your argument.

[forufes]

If the first mutation was not harmful, or carried a benefit of its own, it would not have been "selected out".

theoretically, yes..but in reality, no.. any mutation comes at a price, and some mutation don't pay their price right away, but afterwards when they come in conjunction with another mutation (which also wouldn't be beneficial, and hence a waste, but combined with the previous one it pays off)..

 

 

 

So, I want to take this bunch of sand I have and keep the smaller particles and reject the larger. So, I run it through a fine wire mesh.

 

The mesh selects for the smaller particles of sand, and rejects the larger.

 

The wire mesh is intelligent, by your argument.

oh come on, you should know what logical fallacy that's supposed to called, you're good at naming them..

 

in a world where everyone wants to get smaller, and not bigger, yes, the wire mesh would be intelligent.

 

evolution does play the description of intelligence and the meaning it correlates to in our world very well..in our world, intelligence is how complex are the means something can devise to reach good or avoid bad, not big and small..

 

but what you said was something of a moment of enlightenment to how you guys think evolution is without a direction...and the experience is...twistedly wicked.

[Sisyphus]

What do you mean, "in reality, no?" That is how it works. Mutations persist unless actively "weeded out" by harming reproductive success rates, and even that is not a simple yes/no. They can be weeded out at different rates, or only under certain conditions (thus encouraging different behavior, which in turn changes the whole "environment," etc.)

[JillSwift]

theoretically, yes..but in reality, no.. any mutation comes at a price, and some mutation don't pay their price right away, but afterwards when they come in conjunction with another mutation (which also wouldn't be beneficial, and hence a waste, but combined with the previous one it pays off)..

Sisyphus already answered this one well.

 

 

 

oh come on, you should know what logical fallacy that's supposed to called, you're good at naming them..

 

in a world where everyone wants to get smaller, and not bigger, yes, the wire mesh would be intelligent.

 

evolution does play the description of intelligence and the meaning it correlates to in our world very well..in our world, intelligence is how complex are the means something can devise to reach good or avoid bad, not big and small..

 

but what you said was something of a moment of enlightenment to how you guys think evolution is without a direction...and the experience is...twistedly wicked.

There is no fallacy to name.

 

My analogy is simple (as analogies tend to be): The mesh is the environment, and the grain size is a phenotype. The mesh makes no decisions, doesn't think about what size of grain might fit through it, and otherwise does nothing that is associated with intelligence. It's simply a condition of environment.

 

You keep trying to include value judgments in the natural selection process - this is an entity that is not necessary to explain speciation, nor is there evidence to support such an entity. As such, it should be discarded.

[Mr Skeptic]

There is a huge difference between evolutionary algorithms and biological evolution. You cannot just pick up an evolutionary algorithm off the shelf and tell it "go solve this problem". You have to numerically define the goal,

 

Evolution already has the goal preset, to increase fitness in the organism's current environment. As I understand it, all evolutionary algorithms have the goal to maximize the fitness of the tentative solutions, based on the scoring algorithm you mention next.

 

you have to come up with a scoring algorithm that assesses the distance between a tentative solution and this goal,

 

Evolution has a preset algorithm to assess fitness (based on the laws of physics and the environment).

 

and you have to specify various parameters that control the evolution.

 

Evolution also has this preset, based on the laws of physics, the environment, the organism, and chance, which together determine new genetic variation.

 

Evolutionary algorithms in and of themselves are not intelligent. They are just math. There is always an intelligent designer, a person, behind EAs when EAs are used to solve a problem.

 

What if you input random laws of physics (like the many worlds hypothesis), and your algorithm can set its own fitness function and rules for new variability?

 

As has been discussed right here in this forum far too many times, biological evolution does not need an intelligent designer to function.

 

Well mostly my idea was that if evolution were intelligent, any intelligent design theory would have to include evolution among the possible intelligent designers. Ie, evolution would be the intelligent designer.

[D H]

Evolution already has the goal preset, to increase fitness in the organism's current environment.

That is not a preset goal. What constitutes "fitness" is very dynamic. There is no intelligence behind it. In fact, "fitness" can be downright unintelligent. Witness all the different sexual dimorphisms in animal world. Male deer waste a lot of energy in growing and shedding antlers. Antlers are downright tame compared to some of the oddities in the insect world. If there is an intelligent designer behind evolution it is a dang cruel one.

 

As I understand it, all evolutionary algorithms have the goal to maximize the fitness of the tentative solutions, based on the scoring algorithm you mention next.

Evolutionary algorithms, like evolution, do not have a goal. In the case of EAs the goal is in the hands of the intelligent designer who uses evolutionary algorithms. In the case of evolution there is no goal, period.

 

Well mostly my idea was that if evolution were intelligent, any intelligent design theory would have to include evolution among the possible intelligent designers. Ie, evolution would be the intelligent designer.

Why? An intelligent designer can do whatever it wants. An intelligent designer would just extract a rib and make it morph into some other form. Why use something slow and suboptimal like evolution? When it comes to designing things in the human world, using EAs (and machine learning in general) is often the last resort. Those techniques rarely escape the toy AI world because, like evolution, they are slow and often suboptimal.

[GDG]

that is one of the most important points i hold against a "dumb" random evolution..

 

many mutations are not beneficial in them selves, but when combined with another mutation that comes afterwards, they become beneficial..so what made the trait survive the first "round" of natural selection to team up with the second trait and become useful?

 

A mutation will survive quite a while, as long as it does not prevent reproduction. A mildly detrimental mutation will be passed on to one's descendents, and not die out for generations. (A mutation that drastically reduced reproduction would not be "mild" by definition.) Even pretty bad mutations like hemophilia can survive many generations.

 

How would you explain the fact that humans have many inherited diseases, most of which are not new this generation?

 

keeping the good changes and rejecting the bad is intelligent.

 

No, simply automatic. By definition, a "good" mutation is one that increases the organism's fitness in its environment, and results in its eventual predominance due to reproductive success. It is a mathematical fact, not an intelligent design.

[Mr Skeptic]

That is not a preset goal. What constitutes "fitness" is very dynamic. There is no intelligence behind it. In fact, "fitness" can be downright unintelligent. Witness all the different sexual dimorphisms in animal world. Male deer waste a lot of energy in growing and shedding antlers. Antlers are downright tame compared to some of the oddities in the insect world. If there is an intelligent designer behind evolution it is a dang cruel one.

 

The goal is always to increase fitness in the current environment; that does not change. The environment does, so fitness is a moving target. Regardless of the maladaptiveness of sexually selected traits, they are selected for because they increase fitness.

 

Evolutionary algorithms, like evolution, do not have a goal. In the case of EAs the goal is in the hands of the intelligent designer who uses evolutionary algorithms. In the case of evolution there is no goal, period.

 

The goal of all evolutionary algorithms is to increase fitness. For a designed evolutionary algorithm, the goal is set by setting the fitness function.

 

Why? An intelligent designer can do whatever it wants. An intelligent designer would just extract a rib and make it morph into some other form. Why use something slow and suboptimal like evolution? When it comes to designing things in the human world, using EAs (and machine learning in general) is often the last resort. Those techniques rarely escape the toy AI world because, like evolution, they are slow and often suboptimal.

 

Whoever said anything about an intelligent designer using evolution? What I was suggesting is that evolution itself could be considered an intelligent designer, though for that it would have to be considered intelligent.

[JillSwift]

Why "goal" and not "effect", Mr. Skeptic?