|
Author
|
Topic: Reverse Engineering the Human Mind
|
warren_bergerson
Member
Member # 262
|
posted 26. July 2002 14:22
AI from its inception has accepted the concept that a machine could ‘simulate the functioning on the human mind’ if the machine could ‘simulate the information processing of the human brain’. Simulating information processing in the brain involves 1)identifying how the brain processes information (called reverse engineering) and 2)building a machine capable of simulating the identified information processing. To better understand the issue of creating a ‘mechanical mind’, it is useful to consider two very serious unresolved technical problems associated with reverse engineering the human mind/brain.
To begin, the basic organization or structure of information processing in nervous systems has been well known for a long time. The basic processing unit in all nervous systems is the neuron. Although there are differences between neurons, all neurons are believed to have the same basic information processing mechanisms. The information processed in individual neurons is passed or communicated to other neurons via known processes and mechanisms.
THE NEURON
The first unresolved issue is understanding the functioning of individual neurons. It is relatively easy to precisely define and quantify both the input and output associated with neuronal information processing. It can be shown that
a) the relationship between input and output at a point in time t0 can be represented or modeled by a mathematical function ft0.
b) the function ft0 is adaptive or ‘increases the likelihood the organism will survive’ at time t0
c) the information processing in the same neuron at time tn can be modeled by a different function ftn which is adaptive at time tn
The functions ft0 and ftn have very high degrees of complexity (the ratio A/B is very large where A is the number of possible input/output functions and B is the number of adaptive input/output functions. ) The processes and mechanisms by which neurons create and maintain adaptive functions or programming are unknown. [This is logically very similar to the question of how do individual cells create and maintain adaptive genetic states at different points in time. ]
LOCATING HUMAN FUNCTIONALITY
The second unresolved issue is the ‘location’ in the brain of high level functionality. With respect to basic motor and sensory processing, it is possible to identify with a high degree of specificity which neurons are associated with which motor or sensory processing. With respect to ‘high level, voluntary, human behavior and decision making this specificity seems to disappear. [It even seems logically impossible to have specific neurons associated with ‘liking rock and roll’ or ‘buying peanut butter’.] Reverse engineering human functionality will be extremely difficult until there is some technique for identifying what is happening and where.
SUMMARY
Modeling and simulating the human mind is a very interesting and very challenging subject for scientific analysis. Understanding and modeling the human mind is, IMO, a very solvable problem. However, to suggest that the problem is easily solved by some simple, mechanical ‘reverse engineering’ process is very misleading. Given current knowledge, we couldn’t reverse engineer the brain of an ant, let alone the brain of a human. Given current levels of knowledge, we don’t even know where to look to begin reverse engineering high level human behavior.
[ 27 July 2002, 00:27: Message edited by: Moderator ]
IP: Logged
|
|
Leonid Andreev
Member
Member # 282
|
posted 02. August 2002 05:17
I completely agree with Warren’s summary: “Given current knowledge, we couldn’t reverse-engineer the brain of an ant, let alone the brain of a human”. Moreover, to those who deal with the problem directly, it must be painfully clear how little is known today about the basic mechanisms of information processing in a human brain and how hard it is to unequivocally define human’s mental abilities so that one could rely on such definitions when constructing an artificial prototype of human intelligence. Alan Turing’s test proposed in 1950 has still been the focus of discussions in the AI community – not due to its exceptional scientific merits but mostly because any other set of criteria would be too difficult to be squeezed through a simple cliché. (See, for example, Guilford, J. P. (1982) Cognitive psychology’s ambiguities: Some suggested remedies. Psychological Review, 89, 48-59. According to Guilford, there are theoretically at least 150 different components of intelligence.)
Therefore, the pursuit of the “copy-the-brain” task beholds more than simply an engineer’s dream or a scientist’s idée-fix. No one would refer to Elvis Presley impersonation as “reverse engineering”. It is not by accident that this term has been used in relation to a human brain.
It seems that AI studies repeat the history of Darwinism and evolution biology. Darwinism, neo-Darwinism, anti-Darwinism – it all transforms into an enormous amount of dissertations, books, ideas, and discussions, whose logic and argumentation provide moral or material satisfaction to authors and their successors, as well as a ground for rebuttals by their opponents. However, this ocean of relentless activity steadily evaporates, exposing the abyss that still hides the real history of the origin of life on Earth. Talks about micella as primary carriers of life in the pre-biotic broth remind me of discussions about synapses, neurons, dendrites in the context of the theory of consciousness. In both cases, there are a lot of eloquently worded scenarios. In neither of the cases, there is any proposal of a realistic mechanism that could corroborate with the scenarios and explain the phenomena of life or consciousness. Baron Munchausen is known to have hauled himself out of mud by lifting himself by his own hair. Does not that remind of the attempts of both AI and evolution biology to skip what cannot be skipped? Such an approach results in a naïve belief that quantity can make up for quality.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 02. August 2002 12:47
Leonid,
I would like to follow up on a couple of aspects of your comments.
First, I think it is useful to recognize the very close parallel that exists between ‘reverse engineering nervous system function’ and ‘reverse engineering genetic function’. AI and evolutionary science face very similar problems. And as you point out, despite much discussion, neither has had much success in solving the problem.
While it is important to recognize that the ‘reverse engineering nervous systems and genetic function’ problems have not been solved, it is also important to recognize that problems are potentially solvable. I personally am interested in using the design concept at the micro level to address the problem.
As I approach the issue, the ‘program controlling information processing in a neuron at a point in time’ and the ‘genetic state of cell at a point in time (defined in terms of the active/inactive states of individual genes)’ represent or define the design or biological design (or the biological programming) of the neuron and cell at a point in time. Developing models which model, explain and simulate the processes the produce changes in these elementary designs or programs is, IMO, the essential first step in understanding design, intelligent design, or biological design at the macro-level.
By starting with modeling, and simulating how these ‘design elements’ or ‘design programs’ change over short periods of time, it is eventually possible, my analysis suggests, to explain and simulate how these design elements change over long periods of time. Starting with modeling changes in small design units, my analysis suggests it is also possible to explain and simulate the larger macro forms of biological design(and macro changes in biological design).
Your comments very clearly articulated that 1)current knowledge of information processing in neurons and of genetic changes is far from complete and 2)extensive discussion of these topics often leaves the misleading impression that we have more knowledge than we actually do. I would add to your comments the observation that the misleading impression of the our current level of knowledge increases the difficulty of introducing and discussing technical issues that (may) have potential for addressing the unsolved issues.
IP: Logged
|
|
Leonid Andreev
Member
Member # 282
|
posted 04. August 2002 21:37
Warren,
I like your term “reverse engineering of genetic function” in a sense that without a well-designed conveyor-type reproduction system that takes into account the functionalities and function relationships in every small bit and at every fine level of an organism, life would be impossible. In this case, a ‘genetic function’ serves as a general characteristic of an assembly line for each of the species of organisms, and any malfunction or even slightest fault result in a “defective article” – a species that is now fossil.
However, not all pieces fall into their places for the term “reverse engineering of genetic function”. A question will arise: When and why did the necessity of reproduction emerge?
Let us consider various scenarios. Assume for a moment that life emerged in the form of a certain molecular cluster that had become the teleological matrix of the evolutionary development, which further led to cellular and higher levels of life. (In the past, I proposed a hypothesis that the primary molecular cluster, that had a lipoid nature and determined the whole path of the evolution of the living matter, had a remarkable ability for accumulation of energy from energetic noises. It was published in Russian, and I am planning on having posted its English translation on my website www.matrixreasoning.com.) Here comes up the question: At what stage of their development had protobionts got the necessity of self-reproduction? The answer based on common sense will be unequivocal: at no stage. Indeed, even if there would arise some sort of mysterious competition among protobionts, which is absolutely unlikely, given their concentration and functionality, then how exactly could they acquire the initial replication mechanism with so many perfectly relevant and precisely designed components that it all makes it impossible to have emerged “as is”, i.e. without progressive evolutionary perfection involved, which, in its turn, is impossible without a perfectly functioning reproduction mechanism? At this point, the whole matter boils down to the classical question of which came first, the chicken or the egg, and therefore the scenario does not make any sense.
The predicament only worsens when we try to involve creationism in this issue. One might think that the answer to the question of when and why protobionts started reproducing is simple: at the Creator’s discretion. However, even for the Creator it would be impossible to first create a sound and only then craft a pipe to make a sound – that is to say, first came the organism, and then the reproduction mechanism. Given that evolutionary transformations are surely being observed in the nowadays living organisms, i.e. after the Life has been created, normal logic cannot accept the creationism theory: what, sloppy job? That would disavow the very idea of Creation.
Of course, the issue is too complex in order to be resolved in the course of superficial discussions. But there are two points about it that, in my opinion, go without proof: a) the emergence itself of a genetic function is one of the biggest problems of evolution biology and of origin and evolution of life; and b) however amazingly complex the teleology and autonomy of a genetic function may seem to us, it is nothing more but a production line, although capable for self-adjustment and self-correction. No doubt that the research community is more interested in the conveyor belt itself rather than in what is being assembled on it, which is certainly an absurdity and will look even more absurd as time goes, because it clearly shows the tendency of avoiding unsolved and unsolvable problems by jumping into new, more complex problems.
The AI field seems to suffer from the same type of aberration. It is clear that nothing is clear about the mechanism of information transmission and processing in the brain. Without having that problem resolved, any most bold AI project will be a colossus on clay legs (including “reverse engineering of a human brain”). Creating the artificial intelligence and an apparatus congruent to human mentality is an ultimately exciting and honorary endeavor whose practical and theoretic success can be safely predicted. However, once created, ‘artificial intelligence’ will be as euphemistic a notion as, for example, ‘artificial neural networks’ that are not even close to brain neuron networks. Like DNA in live organisms, intelligence is a certain superstructure on top of the system for information storage and processing in the brain. Modern science has been successfully avoiding facing some of the unresolved basic issues of biology, which is, however, becoming more and more difficult.
Both in AI and evolutionary science, simple and clear-cut solutions and ideas, which, in fact, may explain the phenomena of life and consciousness, are looked upon as too speculative and unworthy of consideration in view of the modern investigation technologies. This is too obvious to be unnoticed.
[ 05 August 2002, 00:14: Message edited by: Leonid Andreev ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 05. August 2002 12:41
Leonid,
I certainly agree with your assessment of both the need to understand basic genetic and nervous system functions and with the current lack of understanding of these basic functions.
It is, IMO, useful to recognize the rather interesting structural similarities between the two problems. Both share what I refer to as a four phase or four dimensional structure. The three phases are probably most apparent in the operation of a neuron where we have:
Phase 1: The adaptive or teleological impact of nervous system processing
Phase 1: input/output or cause/effect information processing performed by neurons
Phase 2: the neuronal program which defines the input output process and
Phase 3: The processes or operations which change the programs controlling the input/output program
Genetic functionality has the same structure with:
Phase 1: The adaptive or teleological impact of genetic processing
Phase 2: The input/output or cause/effect processing(assembly of organic compounds) produced by genetic mechanisms
Phase 3: The genetic programs which defines the input/output processing and
Phase 4: The processes or operations which change the programs controlling the input/output programs.
Not only to genetic functionality and neurons share this common structure, but traditional analysis makes the same basic mistake in interpreting this structure. Analysis of both neuronal and genetic programs make the faulty assumptions that the programs involved change very slowly. In fact, observation shows that both genetic and neuronal programs change from second to second, not from life time to life time. This means that the programs responsible for reprogramming (the power or intelligence in the system) is far, far greater than suggested by traditional analysis. [I am currently developing a formal definition of biological design which incorporates the four dimensional concept.]
I would agree with you that explaining the evolution of genetic functionality is one of the key issues that needs to be addressed by any theory of evolution or genetic change. It is somewhat interesting that the ‘evolution of neurons and nervous systems’ is not similarly viewed as a critical issue (IMO, the evolution of humanness or the evolution of human social behavior should similarly be viewed as one of those major evolutionary leaps. )
However, IMO, once you are able to define genetic functionality and to identify the processes responsible for maintaining the adaptive state of genetic functions or programs, then explaining the origin of genetic functionality will be a far more manageable task. Take for example the issue you raise on origins of reproduction. One of the essential features of life is the existence of multiple distinct individuals. It seems likely that precursors of life forms must have similarly exhibited this ‘sets of distinct units’ property(I like to think of this as sets of some type of bubbles.)
Some individual pre-life bubbles would have been destroyed by accidents. In the absence of some primitive reproductive process these bubbles would have disappeared. These pre-life bubbles, IMO, would have had to contain what I call ‘adaptive design element’- in effect chemical reactions that increased in some manner the probability that a bubble survived. Any such chemical reactions would quickly disappear unless it was compatible with primitive reproduction.
Either some simple ‘genetic chemicals’ or possibly some precursor or genetic chemicals, IMO, occurred naturally in the pre-life form as an adaptive design element. As adaptive design element with appropriate properties, again IMO, would develop or evolve into a large set of what I call ‘biological design elements’. Genetic material is an example of a large set of what I call biological design elements.
IP: Logged
|
|
nobody
Member
Member # 145
|
posted 30. August 2002 14:40
Hello again Warren,
My compliments on a very fascinating idea for a thread. I hope we can get more participation on your thoughts about the brain.
Leonid says:
"1. However, even for the Creator it would be impossible to first create a sound and only then craft a pipe to make a sound – that is to say, first came the organism, and then the reproduction mechanism.
2. Given that evolutionary transformations are surely being observed in the nowadays living organisms, i.e. after the Life has been created, normal logic cannot accept the creationism theory: what,
3. sloppy job? That would disavow the very idea of Creation."
* * * * *
1. I disagree. Who are we to set "impossible" limits on our Creator?
2. Again I disagree. I see no "evolutionary transformations" that are "being observed in the nowadays living organisms". I see life of a wide variety that functions and reproduces within the parameters of its DNA programming. We see many sizes, shapes and colors of dogs but all of them are still dogs.
3. Since creating life is far beyond human capability, I don't think it's wise to call any of the designs "sloppy".
.02
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 30. August 2002 23:02
Dear nobody,
Let me correct some of your claims:
first of all if you claim that nothing is impossible for the Intelligent Designer then his/her Intelligent Design would seem to have placed itself outside the realm of scientific inquiry since anything goes.
Creating life is at present beyond human capability but that need not mean that it will always remain so.
I agree that we may have certain faith beliefs but they should not be confused with scientific inquiry I would argue.
IP: Logged
|
|
Mark Szlazak
Member
Member # 391
|
posted 31. August 2002 00:18
I seriously doubt that reverse engineering the brain would succeed in also reverse engineering the mind or even "information processing." The problem with all materialistic approaches to mind is that mind gets squeezed out of the measurements or analysis. Unless of course your dealing with "objective consciousness" what ever the hell that is. The problem with current scientific approaches to mind is they don't deal with it in its own "terms." Basically science is on all fours when dealing with mind and needs another enlightenment as much as the rest of society. This means a new ontology and epistemology. The problem is physics! Materialism and physicalism just won't cut it beyond mere correlations of mental states to physical states, and that doesn't mean just brain states. There is no evidence that says brain is the only thing that correlates with mental states, since mental states have been associates with other parts of the human body and beyond.
Panpsychists would say that some degree of subjectivity/interiority exists in any entity including things like quarks, etc. Consciousness goes all the way down as well as up.
[ 31 August 2002, 20:02: Message edited by: Mark Szlazak ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 31. August 2002 15:24
Nobody,
Thanks for the kind words. I would like to make two quick comments on the following statement.
Quote: Again I disagree. I see no "evolutionary transformations" that are "being observed in the nowadays living organisms".
First, it is useful to note, that if we broaden our perspective to include ‘adaptive transformations in biological design’ rather than limiting our analysis is ‘long term genetic transformations, then there is are very large numbers of ‘transformations in biological design’ which can be observed and analyzed.
Second, there is a form of ‘evolutionary transformation’ that is "being observed in the nowadays living organism" but which is generally not recognized as an evolutionary transformation. Organisms or more specifically groups of organisms have the ability to ‘maintain a species identity’. This maintenance of identity requires ongoing active ‘evolutionary transformations’. In the absence of such processes, you would get the type of divergence measured by so called genetic clocks.
As an interesting aside, it appears to be ‘mathematically/logically impossible’ to model, simulate or explain this ‘maintenance of identity’ process using realistic mutation and selection rates, and mechanisms included in either Darwinian or neo-Darwinian theories. I will be addressing this in the next few days.
It is also worth ‘observing’ that the failure to reverse engineer neurons, and the failure to reverse engineer genetics doesn’t appear to be lack of interest in the subject, but the lack of a practical methodology for performing the reverse engineering. Existing techniques could be used to ‘validate’ how the processes worked if we had a good starting point for understanding what form the processes take. We aren’t likely to get a good idea of how the processes work unless we are ‘lucky enough’ to find 1)a relatively easy to study example of the processes, and 2)if the process studied was logically similar to other more difficult to study processes.
IP: Logged
|
|
nobody
Member
Member # 145
|
posted 02. September 2002 15:36
Hello Frances,
You say:
"1. Let me correct some of your claims:
2. first of all if you claim that nothing is impossible for the Intelligent Designer then his/her Intelligent Design would seem to have placed itself outside the realm of scientific inquiry since anything goes.
3. Creating life is at present beyond human capability but that need not mean that it will always remain so."
1. Actually you didn't.
2. Here's what I posted: "Who are we to set "impossible" limits on our Creator?" Your sentence does not seem to relate in any way to my comment. We humans can study what God has created. However that does not mean we will ever be able to understand how He created life.
3. We are not currently intelligent enough to create life. The more scientists study the incredible design, assembly and programming of life, the more difficult it becomes for them to fool anyone with the abiogenesis hypothesis.
IP: Logged
|
|
nobody
Member
Member # 145
|
posted 02. September 2002 15:40
Hello Warren,
You say:
"it is useful to note, that if we broaden our perspective to include ‘adaptive transformations in biological design’ rather than limiting our analysis is ‘long term genetic transformations, then there is are very large numbers of ‘transformations in biological design’ which can be observed and analyzed."
Okay, but I don't consider that to be "evolution". I see that as falling within the range of programming that is already within DNA.
I hope I am understanding your point correctly.
I look forward to your further information.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 02. September 2002 17:05
Nobody,
Just a brief Labor Day observation on the difference between ‘evolution’ and ‘other adaptive ‘design changes’ in biological systems’. If we ignore changes in ‘neuronal programming’ and changes in ‘changes in human behavior’, we still have a wide range of ‘adaptive change processes associated with genetic material’. Based on what is currently known about regulatory genes, we know there are lots of different types of ‘adaptive genetic changes’. These include both ‘changes between generations’ and ‘changes within an organisms lifetime’.
Traditionally, a major distinction has been drawn between ‘inter-generational change’ or ‘evolution’ and ‘within life time changes’ or ‘genetic programming’. The ‘assumption’ that there is a clean demarcation between ‘inter-generational change’ and ‘within lifetime change’ was introduced by man. Nature has a nasty way of crossing the arbitrary lines drawn by man. If you look at different types of adaptive change, and the different rates of adaptive change, you will, I believe, find a continuum, not a discontinuity.
However, as I mentioned before, even if you limit ‘evolutionary change’ to changes which are clearly intergenerational, there are readily observable and measurable evolutionary transformations occurring all the time. These are ‘transformations to reverse divergence’. The stability in species we observe all around us is happening in rapidly flowing river. If ‘species’ weren’t swimming or evolving rapidly upstream, they would be changing (diverging) rapidly.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 03. September 2002 11:53
Nobody,
In order to discuss reverse engineering the mind or reverse engineering evolutionary change, it is useful to first back-up and define some basic concepts. To begin, it will be recognized that evolutionary change and adaptive change represent a ‘change in form between some point in time t0 and t1 of some continuous phenomena lx’. In design science terminology, adaptive or evolutionary change is defined as ‘a change in the ‘point in time design’ of some phenomena lx between time t0 and t1".
We can define Q as a ‘scientific modeling rule’ which allows us to assign a unique numerical value to the ‘point in time design of lx at time ty’ . Using this terminology and notation, an adaptive/evolutionary change process is expressed or modeled by an expression of the form:
F(Q(lx,t0)) = Q(lx,t1).
Developing a scientific model or theory of adaptive/evolutionary change involves addressing a number of ‘technical’ problems including
1) defining the range of lx and ty values the model is to apply to
2) defining the quantification rule Q and
3) defining the process rule or algorithm F.
To go back in time a bit, Darwin ‘fact of evolution’ was a proposal hypothesizing the existence of some algorithm F, which for any lx a life form could predict/explain any change from Q(lx, ty) to Q(lx, tz) where tz occurs after ty. Darwin also proposed that 1) there was some modeling rule Qd which would produce a unique value for each living organism, and a process involving diversity and natural selection which could explain all changes in the values of Qd. One of the basic flaws with Dawin’s TOE as a scientific theory was that it couldn’t offer a precise, workable definition of Qd.
Discoveries in genetics appeared initially to offer a clear solution to the problem of a modeling rule Qd. DNA offered a clear precise rule for assigning a unique value to each individual organism based on the sequences of pairs in the individual organisms DNA. It also appeared initially that it might be possible to explain or model all ‘evolutionary’ changes in DNA in terms of the interaction of ‘random mutations’ and ‘natural selection’. RM&NS looked initially like a ‘hard science’ model or theory of evolutionary change.
Given what appears or appeared to be a hard science theory of evolutionary change, the questions arose:
1) How do you test the theory? and
2) If the initial form of the theory is falsified, how do you formulate an alternative to replace it?
A great deal of time could be spent discussing how evolutionary biology and genetics addressed these issues. For the discussion here, it is sufficient to note that the design science approach are significantly different. As a starting point it is useful to note four major differences:
1) Unlike evolutionary biology, design science offers objective, verifiable techniques for testing/validating/falsifying evolutionary/adaptive change models and theories.
2) Design science offers a variety of alternative techniques for quantifying ‘point in time design’ including ‘point in time design’ associated with what is called evolutionary change.
3) Design science offers techniques for explicitly formulating alternatives to RM&NS.
4) Design science offers of a wide range of possible alternatives to RM&NS model or theory of evolutionary/adaptive change.
While some of the other topics may be more interesting, I think it might be useful to start the discussion by considering some of the different options for quantifying point in time design. I will try to address that topic tomorrow.
IP: Logged
|
|
|
Printer-friendly view of this topic