Then You see Mr. Pieters I say that there is no dearth of models but what is lacking is the proper interpretation of them into prototypes. To even replicate a system that emulates the qualities of " Intelligence " one has to first define what it is that the AI community is trying to replicate / emulate in the first place.

The parameters of intelligent behaviours have to be decided upon ? For if we take the Turing Test as the basis for any Parameter or Bench Mark of Intelligent systems then it can be done by running linguistic algorithms which will not let the algorithm 'experience' a conversation as ours .

The Church Turing Thesis keeps cropping up in A.I. research. Not because it is very effective at giving us a model, but because it plays to the scientific prejudice that is based on the Laplacian model's assumption that "Rigid Discipline in thought" or "Reason" (Despite the fact that it often is NP incomplete [A Turing Concept]) is somehow better than a less rigid approach. Turings work purported to "Prove" that intelligence could be determined within an Axiomatic System. Unfortunately the main axiom, a definition of intelligence that all can agree on, is not available.

When you say that there is a Dearth of models, but that there is a lack of prototypes what you are also saying, is that prejudice against many models is either existent, or implied by the basis of the Computer which is the most prevalent Turing Based Architecture. The other models cannot be implemented as easily in a Turing Based Environment, even though it can expand to fit them in a sort of Virtual Machine manner.

Let us look at ANN's for instance. There are ANN models of parts of the brain but no ANN has ever been really well implemented as a computer algorithm, and the algorithms that are the most popular are often poor models of real neural function, if only because the ANN design crowd stabilized the design too early in the Neuroscience research. As Marcus has said, "Neuroscientists need to do a hostile takeover of Neural Networks to force new understandings of the cells into the models so that they can do more.

I have actually heard people espousing A.I. based on Perceptrons, which were debunked by Minsky back in the 1970's. They crop up again and again because they are easily understood, (linear functions) and can be mathematically rigorous in their modelling, while flexible enough to achieve convergence in a reasonably short term.

On the other hand, even though success has been found by Japanese researchers using Fuzzy Logic, the very name militates against it's use in Western Countries, and because it is not as simple as perceptrons, somehow mixing set theory, calculation, and logic in a manner that is profound but hardly simple, few claim to actually understand it.

Why is it, that Laplacian theories are more acceptable in Western Cultures, than things that work?

I have always wondered.

To get back to Lacan saying that there are three spheres, is a wonderful metaphor, but hardly helps untangle the Neuroscience behind it. Do you really expect to find a distinct three lobe brain with each lobe designated for its lacan derived sphere? So, since the brain is the only device we KNOW is intelligent, we must ask, how does this help? It just raises the question of what are the Neural Correlates of Symbolism.

Would you be terribly hurt, if I postulate that symbolism doesn't have correlates in the human brain?

Doesn't in fact need them, because they are inherently demanded by any form of coding and the human mind uses a degenerate (blame Dr. Edelman for that terminology) coding mechanism?

When Humans project symbolism onto things, like codes, they are simply saying that they are capable of seeing a similarity between the things they coded for, and something else. Does 0001 really equal one? or does 0000 0001 capture it better, I can continue of course, convention says today that today I should have at least 7 groups of zeros, and one more group to hold the 0001.

The convention changes with the data-path width of the most prevalent computers, as we move into 64 bit computers the length doubles, 128 bit computers would double again, and so on. Are they really the same number, or symbols of the same number in different (but compatible) systems of symbolism?

My theory of how implicit memory works, shows that in fact, each of these symbolic markers could very practically be triggered by the same stimulus, and could therefore cluster around the meaning of 1 as a natural result of chunking allowing the brain to determine which coding it wants for a specific computer, without actually needing a form of brain circuit sensitive to symbolism.

All you need is a lot of brain circuits sensitive to similarity, and an ability to partition all the data-elements that are similar to select one. I often am amused at my mind trying to assert that the 0 bit at the end of the binary phrase, is a symbol of the value, because of course it doesn't carry to the other bit locations and therefore is a case of overgeneralization.

When I first started in computing, Logic was thought to be the solution to everything, and I still get in trouble when I point out the functional limits to logic. I recognized then, that logic was just a means to manipulate symbols, and now I realize that symbols are just a means to manipulate similarities.

What I meant to convey Mr. Smith is that lacan speaks of orders not physical lobes of the human brain. He was a Doctor and his training would not have allowed him that but what he informs us with is ,this knowledge that by allowing an automotive system to analyze and aggregate data along the similar perceived-reality principles as ordered by lacan.

And by allowing an AI unit to aggregate and compare conclusions that separate or independent 'Data Processing Architectures' have drawn and leave the Unit the agency to formulate an amalgamated version of the three separate processings/ orderings into a whole and by trial and error allow it to apply that perceived-reality template to the physical world in active interactions.

Then may be we might come closer in understanding what actually is that 'Consciousness' that is fueling my defense of application of Lacan in AI and your very legitimate insight that architectures and models are often perceived similarities in systems that are not related by either context or structure.

Yes, well, what I was hoping to convey was that the Order of Symbolism, is probably no such thing, or is a process rather than an order, implied by the attractions of similar stimuli, in the implicit memory.

There have been a lot of A.I. type systems approaches, and I believe that Igor Aleksander of Empire University has made a case for what he calls an Imaginary representation of the external world. The trick, I think however is to link your Symbolic Order to Episodal Memory, since although symbolism is implied in skill, it is not required, and although symbolism is implied in discrete storage, it is not in implicit memory, and there is no need for symbolism in reactive memory. Can you think of how a Symbol could be mapped onto what/where and When? If so you have expanded episodal memory to conceptual memory without half trying.

Yes Mr. Pieters, Lacan indeed has the scope and dynamism to answer all of the three questions placed by you. Lacan's Ideas are anthropogenic but they are not anthropomorphic but being a subject modeled on the human psyche it is difficult to think of it as a functional algorithm on a computer but it is still possible.

Lacan's models can be duplicated in a system that is excluded from a larger system ' the other' but allows it to draw inputs and reflect and effect a larger outer envelop, an outer algorithmic shell to the lacanian shell if you will . Lacanian linguistic algorithm will then have to be a 'fluid algorithm' enveloped within a larger fluid algorithm that reflects patterns that has to be discovered by the inner 'fluid shell' ( excuse my neologism).This way the developmental fluidity and probability of the Lacanian system of modelling be truly realized.

Secondly Lacan is a more advanced choice for such a model but due to its organicity ( again excuse my neologism) and its space to allow the synaptic randomness in consciousness to be articulated it will impart to a strong AI unit the essence of the 'flesh of the mind'. This is more pertinent to a biologically view of consciousness than Dennet's more rationalistic linear causal view of it.

Thirdly I have but rudimentary understanding of computer languages but am trying to master some thing of C++ but with the correct Lacanian training a near accurate programming of a computer to run a viable and resilient Lacanian ' Simulcara Consciousness' is not an unreal possibility.

Kees: Thankyou for your comment. There is indeed a need to deal with ambiguity that has failed to happen partly because of Laplacian sentiment. I have been working with the idea, that even digital circuits could be degraded to become uncertain. The idea being that if we sacrifice certainty, we can with a little more work, recover the order, and reduce the chaos, in memory, without also sacrificing the ambiguity.

I believe that I personally tend to love to leave statements ambiguous in private conversation, as a speech device, especially when asked a double question. If the brain just filtered the ambiguity out of our statements then the speach trick could not work. I believe that tolerance of uncertainty, is one of the leading failures in A.I. theory today.

As part of my implicit memory model, I am working with a deliberate addition of ambiguity, in logic, as a way of overcoming logics and discrete memories limitations. By degrading the performance of the logic, I can reduce the speed of pattern recognition, so that it takes more layers of logic to deal with the ambiguities injected into the codes. With this, I hope to force discrete memory cells called CAM cells, to act like a less discrete simulation of neural processing. The trick to injecting uncertainty is a simple gate, I call a Satisfycing Gate, that is applied after pattern matching between layers of logic.

Well yes, from the introspective approach the subconscious is hidden from view. But as we understand the architecture of the neural processes, knowing the sort of hourglass like effect you are talking about, we can actually see the constraints being placed on the order of the senses, and then gradually being relieved by Self-Organizational effects at numerous levels. Are you familiar with Dr. Edelman's "Neural Darwinism"? I think a case can be now made for a Synaptic Darwinism, where the actual connections between synapses are often interchangeable, but self-organize in a manner that lets the cell conserve its materials, as it gets older and is not quite as active at gathering them. In other words, if we accept the idea of multiple level self-organization, it is reasonable to see virtually chaotic connections at birth (but a large number of them) and gradually more efficient connections over the life of the adult, (and much fewer or them). Until senescence eventually does away with the mass of neural tissues during dementia in the elderly.

A further level of self-organization probably happens within the homogeneous Tissues, where really small neurons feed relatively large ones that transfer information via the various different types of networks that make up the brain. By the time we have reached the "von Economo"/"Broadmann" area level of architecture we already have enough similarity in function to match not only between individuals, but between many species of mammals.

The interesting thing, is that if we can see ambiguous patterns, we already have, with the use of Self-Organizing map elements been able to show that the information content of the ambiguous patterns are high, and can be interpreted well enough topographically to define signal equivalency across multiple individual brains with out actual architectural equivalency. The primary need is to tolerate the chaotic differences instead of demanding order that just isn't there, in the signal architecture. All we really need is chaos tolerant mechanisms and we can remove the disorder and expose the order with simple topological means. It's sort of like tuning out the noise, to increase our signal to noise ratio in radio. Statistical techniques tend to deal with the pops and crackles, but to deal with the random noise, you really need something like a phase-locked loop that amplifies the base frequency, and reduces the random impulses in comparison, then you can use a threshold such as a squelch to cut off the random noise, without affecting anything carried by the carrier wave.

It's not quite the same because we are talking about structural noise rather than frequency, but the requirements are similar. With structural noise we need fractal filtering instead of frequency filters, but that as they say, just means we need a different mathematics.

Ok, what I hear you saying is that you have a problem with too fast a convergence. You are trying to make a "Choice". Once you have converged to a choice, of course multi-variable bias of the choice is simple. The trick, I think is not to converge so quickly to a single choice. In other words what we want is more subtle biases, and to use a lazy decision making system. In recent years there has been found evidence to suggest that in fact the brain does not make choices so much as reject ones that are not suitable, until it either converges, or allows conscious input at the last possible moment before the action to impose a "Won't Power" veto on the options. Of course sometimes none of the options will do, in which case conscious decision is to delay the decision a bit more while other options are checked, or pick a random step and "Jump" to it, because the time is up and there is no clear "best" decision.

@ kees I agree with what you and Greame are trying to do for the creation of a self referential entity is of paramount importance when dealing with AI. For intelligent behaviour is a function of the self interacting with the 'Other' world outside once an entity learns to differentiate between 'I' and the 'World' the patterns of choices in trying to develop a set of generic principles from sensory information is only some distance away.

Lacan could be much more straightforward:

"Je dis toujours la verité, pas toute, parce toute la dire on n'y arrive pas, les mots y manquent, c'est même par cet impossible que la verité tient au réèl."

I always tell the truth, not all, because telling it all, we cannot get there, words are missing there, as a matter of fact, it is by that impossibility that truth hangs to reality.

I actually have been absorbed in this question for more than 3 year now. And at the margins of academia. Some very interesting ideas are creeping up. One of them is about a negation engine, the concept of which I came up with some time ago and later discussed with a friend in the computer sciences.

An engine that reaches a decision by negating all other choices present to it, A No engine, if you will.