Arches

If you're stuck in a browser frame - click here to view this same page in Quantonics!

Darwin's Chip

Evolving A Conscious Machine

By Gary Taubes
for
Discover Magazine's June, 1998 issue, pp. 73-79

Reviewed by Doug Renselle, 15-18Sep2000

In some quotes, Doug removes Taubes' thelogos.

On 15May2001, we commence changing some Quantonics English Language Remediated Terms in this web page. Our initial effort here is to remediate some classical 'o' words with our remediated Quantonics '¤' words, especially in quantum comtexts.

When you see a classical word which has its 'o' remediated with an '¤,' look for that remediated term in our link above. We will link a few of these terms sparsely throughout areas which we revise below. Doug.

 Allow us to offer some introductory remarks which simplify what AI, genetic algorithms, and quantum neural nets require in order to be viable as quasi~real accoutrements in our every day lives.

Classical, SOM, CTM thing-king, when applied to AI, genetic algorithm and neural net technology simply disable "quasi~realness." How? They impose pure, ideal, classical mechanism on reality and any work we do with and in reality.

Pirsig says it like this, and we agree, paraphrased, "SOM removes and eliminates dynamic quality from reality."

Simply, a quantum~MoQ approach, using QTMs, puts DQ back!
(I.e., a quantum~MoQ approach assumes that DQ is real and
that our work products need DQ in order to be "quasi~real.")

There remain some very forward thinking academics who want to keep DQ with SQ. You may recall this from our 2002 TQS News,

"A huge indicator just appeared in The Chronicle's 1Nov2002 issue, an article titled 'THE FACULTY: Chancellor Says Transformation, Biologists say Mumbo-Jumbo.' Martha Winters Gilliland, chancellor of UMKC, Kansas City, Mo, has set a mission for UMKC to go 'quantum.' Can you imagine our glee when we saw those words? This is, for us, a first. A leader of an academic institution saying that under her leadership that institution must take a quantum approach. This is fabulous. It is as big an indicator as we have seen! But notice article's title: biology faculty members think a quantum approach is 'Mumbo-Jumbo.' Ha! Would that they understood how quantum biology is devastating classical approaches (Gilliland calls classical approaches 'Descartesian;' Descartes, René...was a SOMite par excellence, father of Bergson's despised 'spatial extensity')."

Institutions with leaders like Dr. Gilliland are ones who make quantum~avantis real! A recent example is use of a rat's brain cells to evolve a SON which can fly an F-22 simulator! See a key parenthetical from this local page, just below.

Over two years later, December, 2004, "In recent weeks, the College of Arts and Science, the School of Law, the Bloch School of Business, the School of Biological Sciences, the UMKC-Chapter of the American Association of University Professors, and an association of part-time instructors all voted, by substantial majorities, No Confidence in Chancellor Martha Gilliland. (Update: The School of Education subsequently also passed a vote of no confidence.)" "Mumbo Jumbo" Dumbos quote from bioscibuzz.

Roughly 1st week of December, 2004 The Kansas City Channel announced that Martha Gilliland is resigning as Chancellor! Gilliland was UMKC Chancellor from 2000-2004. The Mumbo Dumbo Jumbo UMKC SOMites have struck a tentative blow. They do not see it yet, but it is really a loss for that entire faculty, KC, and Missouri and Kansas states. A huge loss!

SOMitic CTMs are defending their legacy systems, and this is one way in which they do it. "Status quo is the way to go..."

Imagine three guys in an English pub. It's The Swan in Falmer, one hour south of London. The Swan has easy access to U. of Sussex where our two main protagonists play in its Center for Computational Neuroscience and Robotics.

First protagonist is Inman Harvey. Other is Adrian Thompson. Gary Taubes is interviewing these precursor artificial genetic designers for his article which we review here.

Both Inman and Adrian are computer scientists from Sussex. 50ish Inman used to study philosophy, but decided to get back into computers and now says he is an evolutionary roboticist. Inman wrote a chapter of a book on evolving robot consciousness.

Adrian is a much younger, 20ish, self-proclaimed "not an archetypal computer nerd." Adrian plays no computer games, and only does serious computer work and research (a different way to 'play').

Inman claims humans cannot assess machine consciousness due to anti-machine prejudices. He says we will avoid threats from large objects which are apparently uncontrolled by humans, where we will tolerate them more when we know a human is controlling them who probably has no intention of hurting us.

That innate prejudice is what invalidates our abilities to assess machine consciousness, Inman believes. He says our personal sense of our own self is font and grail of our acceptance and assessment of human consciousness.

Inman concurs with a common view that formal, mechanical, analytic, deterministic 'thinking' robots could never have humanesque consciousness.

However, he asks, "What if we could evolve a robot?" He thinks evolved robots will make human assessment of their consciousness more problematic. (Our review and discussion views this as a more general topic and asks, "What if we could evolve any contrived system?" Doug.)

Adrian disdains this approach, even though he has worked for years in a field called 'evolvable electronics,' or 'evolutionary hardware.' He uses FPGAs and genetic algorithms to dynamically evolve electronics via thousands of electronic generations of selection and mutation to learn to do simple tasks — for example, recognize a couple of monosyllable words.

FPGA is an acronym for Field Programmable Gate Arrays.

Begin aside:

You may remember first reading about PGAs in Tracy Kidder's fascinating book on DEC's and Data General's races to develop competitive super-minis through very late 1960s and very early 1980s. Kidder's book is titled, Soul of a New Machine, Avon, 1981. This is still a good read even after almost two decades.

Perhaps most fascinating about Kidder's book is that even in 1982, no one had any idea of an imminent massive Personal (Machine) Computer tsunami, and consequential demise of super-mini monolith machines Kidder was writing about. Nor did anyone (except inventors of internet precursors) imagine evolution away from supercomputer local servers, toward individual local non-networked machines, back to individual machines capable of multiversal volitional networking. As you read Doug's review, you may find it worthwhile to keep asking yourself, "What's next?"

:End aside.

Adrian tells us how evolved electronics and firmware (vis-à-vis evolved hardware) are much more efficient than anything humans can develop by design. By 'efficient' Adrian means:

Most books on genetic algorithms agree. They tell us that if we want to find an optimal solution to a problem, evolve its solution. Evolution is efficient in both selecting good solutions and evolving incremental solutions in nano-, pico-, or femto-second generations.

Adrian sees his work as experimentation, not something we can ponder philosophically. "I find myself incapable of taking part in a discussion of consciousness. I don't think the work I'm doing says anything about it. I'm just thinking about how I can use evolution to explore new ways of computing. When you start trying to extrapolate to matters relating to humans, you end up in philosophically dodgy areas."

Most well-propagandized classicists have this reaction. So we can quite easily assess Adrian as a classicist working in his own local, formal axiom set.

Those "dodgy areas" are where we make our greatest breakthroughs. Ignoring them, sweeping them under a scientific or philosophical carpet's edge is a major mistake. Like Robert M. Pirsig tells us, "Those mu answers are the ones we grow on."

However, Taubes uncovers some new memes which take us out of Adrian's classical mu avoidance realm and emerse incipience — new re-cognizing and emergent beginning notions of infant artificial consciousness' crux and source.

What fascinates us most about this article is Taubes' descriptions of evolving electronics which appear to tap into quantum nonactuality and generate phenomena which appear (to us) as nonlocal or superluminal tunneling affects.

Adrian's way of thinking about system evolution is classical, objective, and formal. But...he does not always act as he thinks...

Inman's work and his interpretations of it, appear to us more quantum.

Remember, to a classical mind, classical objects can only interact physically. Classical objects have, as mandated by Aristotelian and Newtonian syllogisms and axioms, no nonphysical, nonmaterial, nonobjective interactions. Quantum ¤bjects (i.e., quant¤ns) d¤ have n¤n-classical interrelati¤nships ¤f en¤rm¤us variety, including:

No classicist could rationally or reasonably accept any of these quantum memes. However, without them, your reviewer claims no contrived system will ever achieve consciousness.

Taubes tells us of some inexplicables which Adrian encounters. Why consider some phenomena inexplicables? We think it is because all three of these gentlemen are using classical thing-king methods (CTMs) to consider Inman's results. Perhaps it is only their uses of objectively biased English language which leads us to that impression. Simple use of 20th century English language almost implicitly makes a speaker objective.

Let's proceed with our review and see some of these inexplicables Taubes describes.

Taubes tells us we need FPGAs (e.g., Zilinx XC6216), a computer (e.g., AMD cpu and Linux OS based PC) to program them, and genetic algorithm methods (e.g., see Genetic Algorithms, by David E. Goldberg, Addison Wesley, 1989) to proceed with our discussion.

Enter Taubes' new protagonist: Hugo de Garis. He innovated an idea that FPGAs could be used to mimic biological evolution on a computer. Turns out his idea is a superb means to prototype new products. A great example is a John Deere assembly line Tom Petzinger wrote about in a Wall Street Journal edition several years ago. But almost any product may be prototyped using genetic algorithms.

Taubes develops a crude model of de Garis' evolutionary approach by comparing human DNA organization in sets of chromosomes to FPGA evolvable units which can be developed separately and then integrated into a larger whole which may then be further evolved, and so on…, to develop prototypes.

Much of language used in Taubes' article is highly objective. Taubes and his protagonists think of their work and their tools and methods almost wholly using CTMs. They see biological spawn and new genera as classical objects, or at least their classical language makes us perceive they are thinking objectively. For example use of 'offspring.' And "…nature throws in a few mutations…" And, "…DNA in compact units…" So when they encounter surprising results they call them "inexplicable."

Taubes does use "intermingling" to describe synthetic conjoining of "genetic information stored [propertyesque] in" haploid object sperm and "genetic information stored in" haploid object egg into a new human embryo object called an "offspring." Natural selection thus increases opportunity for survival of most fit species. (Our brackets.)

Taubes writes, "Because evolution has found solutions to extremely difficult problems in the natural world, computer scientists try to enlist evolution to solve difficult computational problems, using genetic algorithms." He goes on to describe a software engineering method which mimics nature to evolve better computational solutions.

He says, "The genetic algorithm generates numerous slight variations of genetic code and then 'individuals' are tested to see which perform best under some fitness scale…Mutations are added [objectively, in a many generation loop] for good luck in a next generation…New offspring are tested and mated…" (Our brackets. Our italicized thelogos.)

Taubes describes de Garis' innovative enlightenment, "I had this flash of an idea, that if you could send in software instruction to [an FPGA] to wire it up, maybe you could look on that software instruction as a genetic algorithm chromosome. You could mutate it randomly and maybe evolve the [FPGA]." (Our brackets. Our italicized thelogos.)

But Doug, what about those inexplicables you mentioned? OK, time for crux. Apparently genetic algorithm scientists do two kinds of evolution prototyping: intrinsic and extrinsic. Extrinsic prototyping is where all simulation is done on a PC. In intrinsic prototyping developers move code wholly to FPGAs which reside in a more realistic environment, and evolution is allowed to proceed there. (Now consider transistor tunneling affects here, in FPGAs, vis-à-vis in a PC where a PC's formal axiomatic instructions and innate architectural design intentionally eliminate most quantum affects (i.e., Murphies J).)

Let's stop just a moment and cover some ground Taubes failed to traverse.

Classicists assume an ideal, mechanistic, formal, deterministic reality. One illusory consequence of this is that man may manufacture ideal objects repeatedly. This concept is fundamental to modern mass production processes.

But reality is not classical. Reality is quantum. Quantum reality shows us that no two superatomic 'things' are ever identical to one another (e.g., people, cars, computers, FPGAs, apples, marbles, snowflakes, molecules, nucleons, etc.)! It is impossible to mass produce identical 'things.' This quantum reality rears its head in Taubes' story, but he nor none of his protagonists tumble to its presence and importance.

When we attempt to really (or realistically) evolve FPGA (or similar) solutions to difficult problems, quantum reality enters our problems' spaces, unless we design our hardware infrastructure to intentionally exclude quantum affects. (We can only partially achieve that classical-minded objective goal, however.)

Where classicists reside in their formal analytic enclosures, and do everything they can to guard from quantum reality's influences, those who dare enter a realm of quantum open system evolution suddenly have to admit to and embrace quantum reality's influences on their work products.

Our team of protagonists call these quantum reality influences "inexplicables." Why? Our protagonists are classicists, thing-king with their CTM hats and blinders firmly in place, wool pulled tightly over eyes and carpet's edge raised for sweeping away "inexplicables." You will often hear them referring quantum real influences as "Murphy." Get rid of Murphy or find ways to hide herm!

On with our review…

Moving our protagonists' evolutionary process from PC to FPGAs "…has the advantage of taking into account a chip's own reality, and whatever subtle phenomena might be happening therein that aren't included in a software simulation." (Our italicized thelogos.)

(Reader, you should be able to infer how incredibly important biological reality's vast quantum islands and quantum stages will become in future prototyping efforts. In fact, we are seeing enormous growth in quantum biological technology and emerscence sciences now. Why? Biologicals are natural genetic algorithms! Biologicals are physi! Biologicals are intrinsic quantum computers!)

Amazingly, Taubes tells us our anti-philosophy protagonist Adrian, "…is using those phenomena [inexplicables] to his advantage." Another way to say that, is "Adrian is using quantum reality to his advantage." Adrian decided not to make any assumptions about how evolution works.

Compare that last sentence with what classicists do. Everything a classicist does is based upon assumptions. All classical science is based upon a know ledge of assumptions called "laws, axioms, principles," all of which in sciences' closed classical box of reason are experimentally verifiable and valid by repeatable (i.e., classically manufacturable) scientific methods—CTM methods. What we see clearly now is that classical science and mathematics, using CTMs, manufacture their 'scientific' results.

So Adrian, like it or not, your own prescient intuitions against making assumptions already embark you on a potential quantum reality adventure.

Taubes describes a key decision: Adrian "resolved not even to tell his genetic algorithm that it was dealing with a digital device" whose innate essence is ideal classical ones and zeroes (which do not exist in quantum reality; see our August, 2000 News (~200k file); but reader, you should cognize that quantum transistors do, by intentional design, model those classical ones and zeroes).

Adrian's CTMs kept him in a classical realm though because he perceived his FPGAs playing in a classical analogue fuzzy world. Rather, he should have been think-king of his FPGAs as playing in a quantum paralogical multiverse.

Quantum reality is neither classically digital nor classically analogue fuzzy. Quantum reality is paralogical (i.e., sophist; see our Sophism Connection, Quantum Connection, and SOM Connection).

Adrian assumed that evolution's only intent is better behavior whose last event can always be improved to a new 'next.'

Adrian set up his experiment with these strange quantum non-assumptions, which no normal classicist could ever abide. Taubes gives good detail on how Adrian did his experiment, which we will not reproduce here. Taubes also offers superb metaphors of slowness of human or biological evolution (~weeks-months (drosophila)/~years (animals) per generation) vis-à-vis nanoseconds for FPGA generational evolution.

Adrian's experiment was to have his genetic algorithms evolve to recognize and distinguish a 1kHz signal from a 10kHz signal.

He invented a selection fitness score to compare generations (nature's score is survival). Better generations scored higher and were used to seed their successors. Seeding included latest best outcome plus minor pseudo random mutations to mimic nature's quantum Planck rate absolute change stimuli. Best chromosome was copied as-is, which Taubes refers as "elitism." Other 'better chromosomes' received mutations. Using genetic combinations of these a new generation is spawned and evaluated for fitness. Process repeats. After roughly 5000 generations Adrian has usable results.

Here is a list of Adrian's inexplicables:

  1. Adrian is unable to tell how his evolved process works.
  2. Adrian says his evolved process is 10 to 100 times more code efficient than a human designed equivalent.
  3. His chip appears to be using 'signals' not classically desirable nor assumed present in his system.

    When he attempted to 'explain' inexplicable three, his CTMs gave him classically naïve tentatives.

  4. His created systems are all unique.

    Remember what we said about classical reality assumptions vis-à-vis quantum reality assumptions? Quantum reality says no two things are alike, so inexplicable four is not a surprise, is it?

  5. His created systems stop working when their birthing environment changes.

    Examples are temperature changes, code changes (unrelated PC instruction changes which apparently should not affect his FPGAs while they are operating), etc.

Adrian concludes that if we are to be successful with genetic algorithms we must develop them (give birth to them) in their real intended home environment, and we must subject them to all possible environmental extremes. We think this is good advice.

For mass production, traditional classical manufacturing mindsets this is a disaster! No piece of code will be identical to any other piece of code. Maintenance will require evolution of product, perhaps in real time!

What are we hearing? We are hearing Adrian describe quantum reality! This is simply awesome! But there is more…

Now what Taubes wants to know is can machines like this 'be' conscious?

By our Quantonics' QTMs our answer has to be, "Yes!"

As long as scientists and engineers are working with quantons, their work products are conscious. All quantons are intrinsically self- and co-aware.

No classical object can be, in its ideal form, self- and co-aware. Classicists go to enormous efforts to eliminate quantum affects from their ideal, formal mechanisms.

QTMs demand that we embrace those quantum affects and use them to enormous advantage. Their monetary and concomitant cultural interrelationship beneficence values vastly exceed those of simply propertyesque classical systems.

Taubes exposes a still predominate classical view of consciousness in his, "Only the rare student of evolution believes that human consciousness was inevitable and not some perverse accident of nature." (Our italicized thelogos.)

Note that quote's arrogant non-quantum anthropocentricity. What about Alpha-Centaurian's consciousness? What about Betelgeusian's consciousness? What about monkey's consciousness? What about arthropodic consciousness? Classicism is anthropocentrically arrogant and uses that pulpit for its ex cathedra monistic, unilogical, unitemporal wails.

Quantum reality is both aware and self-aware across all scales of reality! You do not believe that? Ask yourself some questions: When two subatomic particles collide in a high energy particle accelerator like Long Island NY's new RHIC (Relativistic Heavy Ion Collider), do particle's make choices, during collisions, about what happens next? When a photon causes (actually affects; see affectation) an energy shell in an atom to change levels, AKA QED scintillation, are a photon and an atom making choices?

Our Quantonic heuristic is that reality is intrinsically aware, and that awareness scales to incredibly more complex forms one of which we call consciousness.

Quantum reality is aware and intrinsically capable of evolving consciousness among many other incredible quantum phenomena.

We thought this review might help those of you who have trouble finding Discover's original article.

Thanks for reading,

Doug. 16Sep2000


To contact Quantonics write to or call:

Doug Renselle
Quantonics, Inc.
Suite 18 #368 1950 East Greyhound Pass
Carmel, INdiana 46033-7730
USA
1-317-THOUGHT

©Quantonics, Inc., 1999-2014 Rev. 31Jan2010  PDR Created: 15Sep2000  PDR
(16-18Sep2000 rev - Minor editing and corrections.)
(2Oct2000 rev - Add red items and parentheticals to 'General quantum islandic memes.')
(12Feb2001 rev - Correct spelling errors.)
(15May2001 rev - Add comment paragraph at top of page on quantum language changes. Add link to English objective bias page.)
(14Nov2001 rev - Add 'What Happens Next' link just above.)
(8Dec2001 rev - Add top of page frame-breaker. Add/Extend some text in red, and add some new explanatory links. Some remediation too.)
(21Jul2002 rev - Change QELR links to A-Z pages.)
(8-18Dec2004 rev - Adjust colors. Reset legacy red text. Add page top red text box.)
(27-31Jan2005 rev - Correct spelling of Descartes. Add 'Missouri and
Kansas states' to our Gilliland precis.)
(28Apr2005 rev - Reset legacy red text.)
(14Mar2006 rev - Adjust colors, format, et al.)
(5May2008 rev - Reformat slightly.)
(31Jan2010 rev - Add links to 'affectation,' 'choice,' 'evolution' and 'scintillation.')


Arches