Evolutionary Algorithms now surpass human designers.

Lengthy article. I'm gonna have to swish this one around in my mouth a bit.

Evolutionary Algorithm, was Richard Dawkins' way, to demonstrate how from random assortment of objects, a higher order (or a lower entropy) can be formed, merely by shuffling and re-shuffling the elements for periods long enough for stable formations to occur. "Stable" were formations, or a combination of elements, resistant to change through re-shuffling.

He used it to prove and demonstrate that in spite of the religious dogma, a very complex hierarchy of living forms can be created from utter meaningless chaos, without the initial design to start with, of a divine intelligence, we may or may not call God.

Ref: "The Blind Watchmaker" 1986. A computer program was attached for the book readers to try out.

Later, all sorts of "Life" "Breed" and "Colonise" video games popped out, I forgot their brand names, but whoever remembers the funky eighties, may recall.

Evolutionary Algorithm, was Richard Dawkins' way, to demonstrate how from random assortment of objects, a higher order (or a lower entropy) can be formed, merely by shuffling and re-shuffling the elements for periods long enough for stable formations to occur.

At the risk of sounding religiously dogmatic I think this is a false analogy with living systems. A living thing is a highly complex system inherently capable of reproducing. It is 'sorted' by a very simple process of living in the natural world and competing with its neighbours.

The evolutionary algorythm method uses a very sophisticated sorting mechanism (a computer, a program, inputs from user, electricity) to sort relatively simple mathematical elements.

I understand in principle what Richard Dawkins was trying to say, but I think he can be a bit dogmatic sometimes.

Well, 'natural world' is a pretty sophisticated system. Perhaps more sophisticated than any computer system available by the simple fact that it includes and has produced all available computing systems.

Dawkins tried to show that the 'typing monkeys producing Shakespeare, given enough time' is not a good analogy for biological evolution because all these monkeys cannot short circuit the 'long enough time' by continuously noticing and selecting partially meaningful combinations among the gibberish and continuously favouring them in the ongoing process. Nature can and does.

For an evolutionary algorithm (EA) to be a legitimate analogue of biological evolution (BE) the rigour of selection must be appropriate. A creationist could argue that they had designed an EA which showed that any mutations became extinct and that the species showed no change over time. An evolutionist could set the selection criteria so that mutated individuals survived. Who is correct? EA can't answer this.

I agree completely. No conceivable EA can mimic evolution. But, being very anti-creationist, I would emphasize the difference in size and complexity, perhaps because I cannot and would not find qualitaitive differnces between EA and biological evolution.

Existing practical EA programs, not graphical games, are quite successful in what they aim to do, like predicting traffic.

Dawkins knew how to present a case. He once proposed a program which draws random shaped flowers on a computer touch screen. It then counts contacts made by fooled bees and selects for shapes with relatively many bee contacts. this is evolution, isn't it?

Hi Dovy, I appreciate your polite manner compared to some anti-creationists I have encountered. RE: "This is evolution, isn't it?" This is an interesting question. If the question was "Is this an example of a superficial change with no increase in complexity that has benefited the organism?" then I would say 'yes'. If the question was " Does this prove that 'evolution' can account for the origin, complexity, diversity and disparity between all life on Earth?" then my answer would be 'no'. Davo

1. I was quoting the mentioned book to give a practical example where Evolutionary Algorithm was used prior to the emergence of all these "Colonise" games.

2. "Shuffling" was my analogy not Dawkins'. By Shuffling I meant to point to the genetic exchange, and to the random appearance of mutations, as a mechanism to re-combine and give room for diversity in the emergence of new formations, some of which are randomly more resistant to change, thus 'immortalising' themselves within the genome.

3. "Sorting" in such analogy might be used to point to a natural-selection process, in it's common agreeable reference.

It was a demonstrational aid, which in the eighties, I was surprised to see, how effective it really was:

He took a random assortment of unrelated shapes, to randomly re-combine, versus an algorithmic 'filter' of the alphabet, to match for a few millions of iterations, and voila, slowly but surely, one by one, in a random order, all the alphabet letters were combined.

As the saying goes "Give a billion chimps typewriters, and let them play with it for a billion years, and sooner or later, all of Shakespeare's writings will emerge"

This of course, is a paraphrase of the "Big Numbers Law", but the point Dawkins was making, was indeed clear enough.

is there any algorithm that's truly random? if it is a math generated function, it has to receive inputs from somewhere whether it be human or from a computer. i used this example the other day, i own a texas instruments calculator (TI-89 Titanium) and it has a "random number generator" function, and i wanted to test how random it was. two buddies of mine had two different types of texas instruments calculators (one was a TI-92 and the other was a TI-86) and none of us had ever used this function before. we all started and all three calculators gave the same values for every iteration we went through. i understand that the calculators were from the same company with very similar software, but if it is truly a random number generator why did it give us the same values every time?

Shame on TI!

Random number generators produce pseudo-random number sequences. Some are better than others but none is truly random. But! there cannot be any absolute method for evaluating randomness although there are various very partial tests for it.

Decent generators use a different starting seed for each sequence, such as the number of seconds since last midnight multiplied by the date.

We oldsters remember a pre-pc publication, made by Rand Corporation, which was called 'A Million Random Digits'

In an internet site called fourmilab you can get good random numbers. Good because they are produced by a decaying radioactive isotope.

Then again you may ask "Is there a true randomal occurrence in nature in general" taking into account that any occurrence is dependant on the strict laws of nature, vectorial pre-determination, and other restricting factors.

It's like the use of the terms "Real Time" or "Thermodynamically Closed System". these too are not absolute, definitive terms. These are subject to your expected range, or definition.

You determine that something is "Adequately Randomal" for your expected range or system definition.

Too many variables to account for in mere mortal math. Example: we can't predict the weather; nor the stock market; etc. It's not just genetics. It's also natural selection and that itself is a huge unknown.

EAs are just another example of human vainglory.

Bob

I don't think engineering will ever become the exclusive provence of artificial intelligence. Invention and intuitive design (art, for instance) are certainly able to use artificial intelligence to aid in refining parameters, however the genesis of those domains remains in the creative epiphanies of the human mind.

Hi taejonkwando,

You are absolutely correct. The one thing an algorithm (or any computer program) will never come up with is a serendipitous moment, a sudden flash of insight! We humans reign supreme in that arena.

Regards

Romans 1:20 (New International Version)

For since the creation of the world God's invisible qualities—his eternal power and divine nature—have been clearly seen, being understood from what has been made, so that men are without excuse.

I do believe there is a valid place for Eas in many, if not most, engineering applications. Consider this from the article: "Pierre Legrand and colleagues at the University of Bordeaux 2, France, developed an evolutionary system to configure the electrodes for cochlear implants. Up to 22 electrodes on the auditory nerve let cochlear implants restore lost hearing, but the voltages and timings of the signals applied to them are highly individual, requiring much adjustment for speech to be audible. Legrand's team took just one-and-a-half days to configure an optimal pattern for one patient whose doctors had not succeeded in 10 years." Read: doctors had not succeeded in 10 years!

That's one of the greatest advantages of Eas; that they can perform thousands of evolutionary iterations that human engineers just do not have the time for.

I think the biggest widespread blockade to this technology is simple human pride! [There is an inherent] "tension between evolutionary biology and ID (intelligent design). That's the challenge we need to rise to in winning people over."

Regards,

-John

My problem is with the article not the subject matter. I am frustrated when some 'journalist' makes such grandiose propositions as to the ultimate ability of some new technology or tech improvement.

Why not suggest that a new and optimum development for space travel or grain yields are likely as a result of EA? Instead, again, it is posing science and spirituality (more intently organized "R") against one another, or so it seems to me.

I am not sure that I am speaking clear but it is quite early yet, and a Monday to boot. So I will reserve the right to revisit this topic. Post caffe.

Hi C. Rummel,

I think the lack of clarity may be in what is being compared. As you said "it is posing science and spirituality [...] against one another".

The only way that EAs have been used (to my knowledge) is to produce evolutionary iterations of products, or ideas, that have already been conceived, or produced, by the mind of man. As the opening paragraph of this thread indicates, science is not being pitted against spirituality, rather, evolutionary algorithms are simply surpassing (due to their speed) what a human mind can achieve.

Thus an EA is simply another engineering tool that you, as a scientist, can use to obtain a superior design of whatever widget you happen to be dealing with.

Perhaps therein lies the rub. Can we humans accept the fact that a machine can create (meaning evolutionize) something better than we can, since we (meaning nature) have been doing it, exclusively, since the first uprising from primordial slime.

Sorta like: "But we've always done it that way".

-John

As I re-read I took your point in agreement.

"...some 'journalist' makes such grandiose propositions as to the ultimate ability of some new technology or tech improvement..."

True, true. It's probably in attempt to appear visionary