Can designs be evolved without intellectual contamination?

[QED]

In the topic titled Should Intelligent Design be allowed in classrooms? Rob took great exception to something I wrote:

QED asks the following question about a computer program that “outputs random numbers,” created in the mind of a conscious human being, and compiled on a deterministic device (i.e., computer),

First you have to understand that there are two separate algorithms involved. To make this clear let's say we want to write a computer program that we will call "A" that writes a program of its own called "B".

You want to be sure that ALL the intellect that goes into A is contained and does not enter the product B. Imagine a very simple model: we write a program that outputs random numbers....

Did any of the intellect we used to construct the program find its way into the pattern? Clearly it didn't, and this demonstrates precisely how there can be containment between the different "intellectual" entities involved in the process.

Of course, anyone who has studied computer science knows that contrary to this simplistic and erroneous claim:

Unfortunately, generating random numbers is a task that looks a lot easier than it really is, primarily on any deterministic device. Von Neumann [Neu63] said it best: “Anyone who considers arithmetical methods of producing random digits is, of course, in a state of sin.” All we can hope for are pseudorandom numbers, a stream of numbers that appear as if they were generated randomly.

Any computer scientist is well aware of this fact, and would never make the fallacious claim that QED does above. Any computer scientist worth his paycheck is well aware of the serious consequences of using a poorly engineered random number generator when it comes to security concerns.

Clearly, this example is logically flawed both in fact and truth; it is a fallacious argument because QED is attempting to base his conclusion upon the false premise that a human can write a computer program on a deterministic device (i.e., computer) and generate a truly random input rather than a pseudorandom input. This premise is false as it is impossible; therefore, the conclusion does not follow. Both program A and program B are generated within the creative mind of a human, and subject to the limitations of both the human mind and the deterministic device (i.e., machine) the program runs upon, which is also a creation of human mind. A pseudorandom-number generator is not a true random number, and cannot be made one by the simple magical fact of QED’s defining it as such. It is nothing more than a mind created deterministic algorithm that returns numbers that “look” statistically random.

There is never any so-called “containment between” program A and program B, as both are part of a whole system created by human mind.

And all of this tells us only that humans are clever enough to fool themselves that pseudorandom-number generators generate truly random numbers, and intelligent enough to create deterministic devices to run their clever programs upon.

It tells us nothing about how they got so clever (or foolish) to both design a deterministic machine (computer), write a deterministic algorithm to generate pseudorandom-numbers as input for another computer program, yet still fool themsemselves despite all this cleverness into believing these are truly random numbers, and claim the this tells them something fundamentally about the universe.

I think this whole exercise tells us more about the little "i"ntelligent designer than any big "I"ntelligent Designer.

Rob is certainly correct in pointing out that the typical "Random" function provided by various programming languages returns a pseudo random number. These functions are typically deterministic algorithms and the sequences of numbers they generate will repeat after a finite number of iterations. When true, non-deterministic, sequences are required electronics can be used to convert thermal (Johnson) noise into sequences of numbers that are never expected to repeat.

Interestingly though, I still disagree with Rob on this question of the containment of intellect in the case of the pseudo random sequence. I have often created my own pseudo random sequences in hardware, from long shift-registers and "Exclusive OR" gates as well as doing pretty much the same in software. Now let's say I know nothing about the design of radio antennas. But I need a good design for one...

Well, given that I know I want it to radiate the maximum amount of power for a given signal going in to it, I can always Evolve one using Darwin's insights about nature. All I need to do is write the appropriate computer program "A" that causes a program "B" to evolve -- involving random mutations to its expression genes using one of my pseudo random generators. Either by simulation or by real-world measurement (robot arms twisting lengths of wire or whatever) the evolving antenna design of program "B" eventually reaches an optimal performance level and hey presto! I've got a great little aerial!

Just a minute though -- when I chose the XOR taps along my shift register, was I unwittingly transferring knowledge about what constitutes a good antenna? Maybe I didn't know that a quarter wave is a significant length, but the evolved product emerging as a result of my pseudo random generator design certainly finds this out.

I'm pretty sure then that there is total containment of intellect here -- even with psuedo random numbers. But if Rob wants to keep on nit-picking this to death, let's just say we use thermal noise instead. Now that the source of random mutation is truly random is there any other path through which the intellect of the engineer could get into the evolved product?

[QED]

In the topic tited: Is there philosophical evidence for design?

... We tend to think that we can use our intellect to fulfill our wishes and therefore design things which don't already exist in the world, for our own purposes. In doing so we use our knowledge and understanding of the world to make novel arrangements of matter -- a useful definition of intellect for the purposes of this debate.

Now, as it turns out, there is this algorithm that people such as Darwin have identified which has the ability to yield novel design by a totally different process. This algorithm can be replicated by engineers (a procedure that involves the use of "intellect" but when the work is done, the resulting "intellectually devised" design generator, when switched on, produces its own novel designs using its own version of intellect. This version is somewhat different to one defined above as it does not operate on knowledge and understanding of the world as preconceived notions in the mind of an engineer, however, it can be argued that the nature of the feedback resulting from trial and error ultimately reduces to the gaining of "knowledge" about the way the world works.

Compare an artifical neural network. If we design the neural net (its structure basically) and let it "learn" some patterns, does it now have "intelligence"? We've represented some of our intelligence in the neural net so it has a derived kind of intelligence.

Another example is a robot. If we design a robot that moves around, and it does something we didn't anticipate (because we didn't forsee all possibilities), is the robot now "intelligent"? Yes, it has a derived intelligence build in.

Monte Carlo or genetic algorithms that employ pseudo-random number generators have the same kind of derived intelligence.

We want to know if the "derived intelligence" in the above examples contains any intellectual property of the original human inventor. I'm sure that lawyers could equivocate on this matter for as long as they were being paid to do so, but there seems to be only one answer to me:

I've already "roughed out" the principles of evolving an antenna design in my previous post here. Back in 1965 Richard Feynman was co-awarded the Nobel Prize for Physics along with Julian S. Schwinger and Tomonaga Shin'ichiro for their work on Quantum Electro-Dynamics: the Theory of the interaction between light and matter. This Theory ties together all the phenomena of electromagnetism and codifies with stunning experimental precision the way that nature operates when, for example, electric currents are generated at radio frequencies.

So where would an "antenna evolver" be getting knowledge of all this this from? Is it working its way through the minds and books of physicists and engineers and being unwittingly transferred through to the design of the evolver? I would argue that it is easy to imagine a scenario like the one I describe above where this couldn't be the case -- because the person inventing the "antenna evolver" may never have picked-up any knowledge of the relevant science. Unless you can argue that it is impossible in principal for someone to be in such a position, it proves that designs can be evolved without intellectual contamination.

The neural net you mentioned strikes me as another good example. The network may contain a degree of knowledge passed on by its designer, but any further knowledge it gains through subsequent interactions is novel and may remain utterly unknown by its designer. It really does look like a no-brainer issue (sorry for the bad pun) to me. And as for any pseudo-random generators used in GA being a route for the transfer of the designers knowledge into the evolving solution, this sounds like a bizarre idea to me. I wish either you or Rob could explain how this would work as it would have extraordinary consequences for information technology. Setting aside that tantalizing prospect, we can always revert to using random number generators based on Johnson noise for this application. That would totally close this off as a potential route for intellectual contamination.

[rigadoon]

...So where would an "antenna evolver" be getting knowledge of all this this from? Is it working its way through the minds and books of physicists and engineers and being unwittingly transferred through to the design of the evolver? I would argue that it is easy to imagine a scenario like the one I describe above where this couldn't be the case -- because the person inventing the "antenna evolver" may never have picked-up any knowledge of the relevant science. Unless you can argue that it is impossible in principal for someone to be in such a position, it proves that designs can be evolved without intellectual contamination.

The neural net you mentioned strikes me as another good example. The network may contain a degree of knowledge passed on by its designer, but any further knowledge it gains through subsequent interactions is novel and may remain utterly unknown by its designer. It really does look like a no-brainer issue (sorry for the bad pun) to me. And as for any pseudo-random generators used in GA being a route for the transfer of the designers knowledge into the evolving solution, this sounds like a bizarre idea to me. I wish either you or Rob could explain how this would work as it would have extraordinary consequences for information technology. Setting aside that tantalizing prospect, we can always revert to using random number generators based on Johnson noise for this application. That would totally close this off as a potential route for intellectual contamination.

If I make a telescope and see things I didn't image before, does the telescope have an intelligence that I didn't give it? No. These algorithms are just tools, instruments, inventions -- they may allow us to do things we didn't think of before but that doesn't give them an independent intelligence.

[QED]

If I make a telescope and see things I didn't image before, does the telescope have an intelligence that I didn't give it? No.

The telescope has no intelligence it simply makes something that already exists easier to see.

These algorithms are just tools, instruments, inventions -- they may allow us to do things we didn't think of before but that doesn't give them an independent intelligence.

When you say they do not posses an independent intelligence do you mean that they posses our intelligence or no intelligence at all? (like the telescope). I've explained how these tools can be devised such that they do not possess our intelligence, and I'm not arguing that they posses an independent intelligence of their own. I would rather say that they display apparent intelligence and produce apparent design because we tend to attribute these properties to living things not tools.

As with the telescope the "design" is already "out there" in the state-space of all potential designs. The trick of the "tool" is to search this space and discover a solution. So, here's a functional system for delivering apparent design into the world -- one that has no intelligence in it -- operating instead on a natural logical principle.

[rigadoon]

These algorithms are just tools, instruments, inventions -- they may allow us to do things we didn't think of before but that doesn't give them an independent intelligence.

When you say they do not posses an independent intelligence do you mean that they posses our intelligence or no intelligence at all? (like the telescope). I've explained how these tools can be devised such that they do not possess our intelligence, and I'm not arguing that they posses an independent intelligence of their own. I would rather say that they display apparent intelligence and produce apparent design because we tend to attribute these properties to living things not tools.

Intelligence is an ability. If we make devices with such an ability, they have the ability we design them to have. Take away the designer and there's no device. They have nothing independent of their designer. Devices are dependent on their designer.

[QED]

Intelligence is an ability. If we make devices with such an ability, they have the ability we design them to have.

This is not always the case. Yes, it is true for the vast majority of the devices we design which is why I think you and others like you have this reaction. But there is a particular class of design which we are considering here and it is one which has an independent ability. We can go on in pantomime fashion saying "Oh yes it is", "Oh no it isn't" but in order to break out of this pointless loop, I started this topic to show that it is possible for intellectual contamination to be avoided. So if you want to refute the claim that we have created design generators that do not rely on our intellect for the novel design solutions they generate you need to show where the information bleeds through. Rob tried his hand at the random number generator and failed. Perhaps you can think of something else.

Take away the designer and there's no device. They have nothing independent of their designer. Devices are dependent on their designer.

You are right in your first sentence. This however equates to another paradox that has yet to be resolved -- how the universe with its materials and physical law came about in the first place. But you fall down with your next sentence. This is what I have been showing here; that systems of material and logic can have independence from their "designer". GP/GA are concrete examples of this fact.

[rigadoon]

Take away the designer and there's no device. They have nothing independent of their designer. Devices are dependent on their designer.

You are right in your first sentence. This however equates to another paradox that has yet to be resolved -- how the universe with its materials and physical law came about in the first place. But you fall down with your next sentence. This is what I have been showing here; that systems of material and logic can have independence from their "designer". GP/GA are concrete examples of this fact.

Two variables are independent if a change in one is unrelated to a change in the other. All devices are dependent on the person devising them. If the person does things differently, the result is a different device. Will devices sometimes do things the person devising them didn't think of? Of course. In software they are called bugs. Now you want to call them intelligence? Do you think you are Dr. Frankenstein?

[QED]

Two variables are independent if a change in one is unrelated to a change in the other. All devices are dependent on the person devising them.

The device you talk of here is the "design generator". Certainly whatever form it takes will be dependent on the person devising it. But you're far from successfully arguing that the designs it generates are necessarily dependent on the person who devised the generator. You go on to say as much in your next sentence anyway.

When you say "All devices are dependent on the person devising them." can you not see that this is not necessarily true? This is readily refuted by the extremely crude (but ultimately valid) example of a random sequence of numbers being "introduced to the world in some way" by means of a device devised by me or you. Using a non-deterministic noise source, any pattern that emerges (irrespective of whether it makes sense or not) is independent of the "pattern" represented by the design of the generator.

If the person does things differently, the result is a different device. Will devices sometimes do things the person devising them didn't think of? Of course. In software they are called bugs. Now you want to call them intelligence? Do you think you are Dr. Frankenstein?

This is exactly the sort of "intelligence" that can emerge. While it may be amusing to equate it with the fictional monster, the equation is irrelevant. The fact remains that randomness combined with selection can give rise to the appearance (to us) of intelligent design. Again GA/GP demonstrates this and because intellectual independence can be achieved between the developer of the generator and the generated product as described in this debate, it shows that there is a natural source of apparent design within nature.

I think you need to review the rough outline of the antenna evolver mentioned above and ask yourself where the "knowledge" that leads to the optimization of the design comes from. It looks to me as though you think this is a trivial question, but I assure you it is not. Just because you can picture a bunch of geeks tinkering with hardware and software does not mean this is a no-brainer. The only real unanswered question is what the natural equivalent of the "geeks with their hardware and software" might be

[Scrotum]

changed my mind, disregard this post

[rigadoon]

When you say "All devices are dependent on the person devising them." can you not see that this is not necessarily true? This is readily refuted by the extremely crude (but ultimately valid) example of a random sequence of numbers being "introduced to the world in some way" by means of a device devised by me or you. Using a non-deterministic noise source, any pattern that emerges (irrespective of whether it makes sense or not) is independent of the "pattern" represented by the design of the generator.

So you want to "introduce" a device and then claim no responsibility for it? What if it harms someone? Your argument won't get you out of a lawsuit.

If the person does things differently, the result is a different device. Will devices sometimes do things the person devising them didn't think of? Of course. In software they are called bugs. Now you want to call them intelligence? Do you think you are Dr. Frankenstein?

This is exactly the sort of "intelligence" that can emerge. While it may be amusing to equate it with the fictional monster, the equation is irrelevant. The fact remains that randomness combined with selection can give rise to the appearance (to us) of intelligent design. Again GA/GP demonstrates this and because intellectual independence can be achieved between the developer of the generator and the generated product as described in this debate, it shows that there is a natural source of apparent design within nature.

Now you're talking about apparent intelligence. So this "independent" intelligence is really "apparent" after all?