The AI Question and ChatGPT: "Truth-Mills"?

 

Everyone seems to be complaining about the Artificial Intelligence application "ChatGPT" : From passionate leftwing atheist PZ Myers through moderate evangelical atheist Larry Moran to cultish fundamentalist Ken Ham it's nothing but complaints!

PZ Myers refers to ChatGPT as a Bullsh*t fountain. Also, in a post titled "How AI will destroy us" he blames capitalism and publishes a YouTuber who calls AI "B.S.".  Biochemist Larry Moran gives chatGPT a fail mark on the basis that it is "lying" about Junk DNA (that's also a complaint of Myers, although "lying" is rather too anthropomorphic in my opinion). The Christian fundamentalists go spare and lose it completely: Kentucky theme park supremo Ken Ham, in a post titled "AI - It's pushing an Anti-God Agenda" (March 1st) complains that ChatGPT isn't neutral but is clearly anti-God - what he means by that is that its output contradicts Ken's views! We find even greater extremism in a post by PZ Myers where he reports Catholic Michael Knowles claiming that AI may be demonic!  Ken Ham is actually much nearer the mark than Knowles when Ken tells us that AI isn't neutral: The irony is that although we are inclined to think of AI as alien, inhuman, impartial and perhaps of superhuman power, it is in fact inextricably bound up with epistemic programming that has its roots in human nature and the nature of reality itself. It is therefore limited by the same fundamental epistemic compromises that necessarily plague human thinking**.  Therefore like ourselves AI will, of necessity, hold opinions rather than detached cold certainties. Let me expand this theme a bit further. 

***

From 1987 onwards I tried to develop a software simulation (eventually written in C++) of some of the general features of intelligence. I based this endeavor on Edward De Bono's book "The Mechanism of Mind". I tell this story in my "Thinknet" project and although it was clear that it was the kind of project whose potential for further development was endless, I felt that I had taken its development far enough to understand the basics of how intelligence might form an internal model of its surroundings. The basic idea was simple: it was based on the generalised Venn diagram. Viz:

The whole project was predicated on the assumption that this kind of picture can be used as a general description of the real world. In this picture a complex profusion of categories is formed by properties distributed over a set of items*. If these properties are distributed randomly then there are no relations between them and it is impossible to use any one property as the predictor of other properties. But our world isn't random; rather it is highly organized, and this organization means that there are relationships between properties which can be used predictively. As I show in my Thinknet project the upshot is that the mechanism of mind becomes a network of connections representing these nonrandom relations.  The Thinknet project provides the details of how a model of thinking can be based on a generalised Venn diagram.

One thing is fairly obvious; if we have many items and many properties a very complex Venn picture may emerge and the epistemic challenge then arises from the attempt to render this picture as a mental network of connections.  Epistemically speaking both humans and AI systems suffer from the same limitations: In trying to form a network of connections they can only do so from a limited number of experiential samples.  This would be OK if the world was of a relatively simple organization, but the trouble is that yes, it is highly organised but it is not simple; it is in fact both organized and yet very, very complex. Complexity is a halfway house between the simplicity of high order and the hyper-complexity of randomness. To casual observers, whether human or AI, this complexity can at first sight look like randomness and therefore present great epistemic challenges in trying to render this world as an accurate connectionist model given the limits on sampling capacity. On top of that let's bear in mind that many of the connections we make don't come from direct contact with reality itself but are mediated by social texts. In fact in the case of my Thinknet model all its information came from compiled text files where the links were already suggested in a myriad Bayesian statements. This "social" approach to epistemology is necessary because solitary learning from "coalface" experience takes far too long; that would be like starting from the Paleolithic. 

Like Thinknet we learn far more from social texts than we do from hands-on experience.  Those social "text files" are extremely voluminous and take a long time to traverse. There is no quick fix that allows this textual experience to be by-passed. This immediately takes us into the realm of culture, group dynamics and even politics where biased sampling is the natural state of human (& AI) affairs. The complex mental networks derived from culture means that intelligence, both human and AI, is only as good as the cultural data and samples they receive. So, in short, AI, like ourselves, is going to be highly opiniated, unless AI has got some kind of epistemic humility built into its programming. AI isn't going to usher in a new age of unopinionated and error-unadulterated knowledge objectively derived from mechanical "Truth-Mills". The age old fundamental epistemic problems will afflict AI just as it afflicts human beings: PZ Myers might call ChatGPT a Bullsh*t fountain, but then that's more or less also his opinion of the Ken Hams, the Michael Knowles and Donald Trumps of this world. On that matter he is undoubtedly right! As with humanity (e.g. Ken Ham) then so with ChatGPT. The bad news for PZ Myers is that Bullsh*t production has now been automated!

Truth Mills: Is AI going to automate the production of theoretical fabric?

ChatGPT for dogs

Footnotes:

* Venn diagrams don't have the facility to form a set of sets. However, this can be achieved using another "higher level" Venn diagram; we thus have Venn diagrams that are about other Venn diagrams. See here:

Quantum Non-Linearity: The Thinknet Project. Footnote: On Self Description (quantumnonlinearity.blogspot.com)

** Epistemic difficulties surrounding accessibility and signal opacity loom large in historical research. "Epistemic distance" is a big issue in human studies. 

Breaking Through the Information Barrier in Natural History Part 5

In this series of posts I have been critiquing de facto IDists Nametti and Holloway's (N&H) concept of "Algorithmic Specified Information" (ASC) a quantity which they claim is conserved. The other parts of this series can be found here, here, here and here. 

As we have seen ASC does not provide robust evidence of the presence of intelligent activity; its conservation is also questionable. The underlying motive for N&H's work probably flows out of their notion that "Intelligence" is fundamentally mysterious and cannot be rendered in algorithmic terms no matter how complex those terms are. Also the de facto ID community posit a sharp dichotomy between intelligent activity and what they dismiss as "natural forces". They believe that the conservation of what they identity as "Information" can only be violated by the action of intelligent agents, agents capable of creating otherwise conserved "information" from nothing. 

.

For me, as a Christian, de facto ID's outlook is contrary to many of my own views that readily flow out of my understanding that those so-called "natural forces" are the result of the creative action of an immanent Deity and will therefore reveal something of God's nature in their power to generate difference and pattern; after all human beings are themselves natural objects with the ability to create pattern and configuration via artistic and mathematical endeavour. 

Although I would likely see eye-to-eye with de facto IDists that there is no such thing as something coming from nothing (a notion that is the proposal of some atheists), for me the very glory of creation is that it generates otherwise unknown & unforeseen patterns and configurations; as Sir John Polkinghorne puts it our cosmos is a "fruitful creation": Whilst for the Divine mind there may be nothing new under the sun, for those under the sun the unfolding of creation is a revelation: So whilst it's true that the realisation that something has to come from something prompts us to probe for a formal expression of the conservation of something, on the other hand that creation creates pattern and that humans learn from the revelation of this creation suggests we also probe for a formal expression of our intuition that information is created. De facto IDists behave like crypto-gnostics unable to acknowledge the sacredness of creation (albeit corrupted by Satan and Sin).

All in all I find the IDists obsession with that of trying to prove an all embracing theorem of information conservation as misguided and futile as the atheist project to show how something can come from nothing. 

***

1. Evolution and teleology

In this part 5 I want to look at atheist Joe Feslenstein's reaction to N&H's efforts. In fact in this post Joe Felsenstein criticizes the concept of ASC on the basis that it simply doesn't connect with the essential idea of evolution: that is, the selection of organic configurations based on fitness:

FELSENSTEIN: In natural selection, a population of individuals of different genotypes survives and reproduces, and those individuals with higher fitness are proportionately more likely to survive and reproduce. It is not a matter of applying some arbitrary function to a single genotype, but of using the fitnesses of more than one genotype to choose among the results of changes of genotype. Thus modeling biological evolution by functions applied to individual genotypes is a totally inadequate way of describing evolution. And it is fitness, not simplicity of description or complexity of description, that is critical. Natural selection cannot work in a population that always contains only one individual. To model the effect of natural selection, one must have genetic variation in a population of more than one individual.

Yes, the resultant configurations of evolution are about population fitness (or at least its softer variant of viability). The sum of a configuration's Shannon improbability minus the relative algorithmic complexity is an insufficient condition for this all important product of evolution. 

Fitness (and its softer variant of viability) is a concept which humanly speaking is readily conceived and articulated  in  teleological terms as the "goal", or at least as the end product of evolution; that is, we often hear about evolution "seeking" efficient survival solutions. But of course atheists, for whom teleological explanations are assumed to be alien to the natural world, will likely claim that this teleological sounding talk is really only a conceptual convenience and not a natural teleology. For them this teleology is no more significant than cause-and-effect-Newtonianism being thought of in terms of those mathematically equivalent "final cause" action principles. As is known among theoretical physicists there is no logical need to think of Newtonian mechanics in terms of final causation: Algorithmically  speaking "final cause" action principles are just a nice way of thinking about what are in fact algorithmically procedural Newtonian processes, processes driven from behind. Pre-causation rather than post-causation rules here. 

However, although Felsenstein the atheist will likely acknowledge there is no actual teleology in evolution he nonetheless accepts that thinking about evolution in terms of its end result of fitness makes the whole process (pseudo) meaningful:  Felsenstein points out that the earlier ID concept of Complex Specified information (CSI) is intrinsically more meaningful than ASC and notes that CSI was explicitly stated by William Dembski in terms of end results: Viz:

In the case where we do not use ASC, but use Complex Specified Information, the Specified Information (SI) quantity is intrinsically meaningful. Whether or not it is conserved, at least the relevance of the quantity is easy to establish. In William Dembski's original argument that the presence of CSI indicates Design (2002) the specification is defined on a scale that is basically fitness. Dembski (2002, p. 148) notes that:

"The specification of organisms can be cashed out in any number of ways. Arno Wouters cashes it out globally in terms of the viability of whole organisms. Michael Behe cashes it out in terms of the minimal function of biochemical systems. Darwinist Richard Dawkins cashes out biological specification in terms of the reproduction of genes. Thus in The Blind Watchmaker Dawkins writes "Complicated things have some quality, specifiable in advance, that is highly unlikely to have been acquired by random chance alone. In the case of living things, the quality that is specified in advance is ... the ability to propagate genes in reproduction."

The scale on which SI is defined is basically either a scale of fitnesses of genotypes, or a closely-related one that is a component of fitness such as viability. It is a quantity that may or may not be difficult to increase, but there is little doubt that increasing it is desirable, and that genotypes that have higher values on those specification scales will make a larger contribution to the gene pool of the next generation.

Herein we find the irony: Both Felsenstein and Dembski note that Specified Complex Information is only meaningful in terms of a conceived end result: e..g fitness, viability, minimal function, gene propagation or whatever. But of course what appears to be teleology here would simply be regarded by a true-blue atheist as an elegant intellectual trick of no more teleological significance than the action principles of physics. 

2. Evolution and the spongeam

It is relatively easy to determine whether a given organism has viability; that is, whether it is capable of self-maintenance and self replication; just watch it work! But the reverse is much more difficult: From the general requirements of self maintenance and self-replication it is far from easy to arrive at detailed structures that fulfill these requirements: This is where evolution is supposed to the "solve" the computational problem: It is a process of "seek and find"; a "find" registers as successful if a generated configuration is capable of self-maintenance and self replication given its environment. In evolution the self-maintaining and self-replicating configurations are, of course, self-selecting. That's the theory anyway. 

Strictly speaking, of course,  talk about evolution "solving" a computational problem is not going to go down well with the anti-teleologists because the activity of "solving" connotes an anthropomorphic activity where a problem has been framed in advance and a "solution" sought for; computation in most cases is a goal motivated process where the goals of problem solving are its raison d'etre. But to a fully fledged atheist evolution has no goals - evolution just happens because of the cosmos's inbuilt imperative logic, a logic implicit from the start, a logic where evolutionary outcomes are just incidental to that logic. If we believe evolution to be driven by causation from behind, then evolutionary outcomes will be implicit in the initial conditions (at least probabilistically). It has to be assumed that these outcomes will at least have a realistic probability of coming about given the size & age of the universe and the causation laws that constrain the universe. To this end I have in various papers and posts caricatured evolutionary processes as the exponential diffusion of a population of structures across configuration space. Each reproductive step is constrained by the conditions of self-maintenance and self-replication: But this process will only work if a structure I call the spongeam pervades configuration space. I'm not here going to air my doubts about the presence of this structure but simply note that conventional evolution requires the spongeam to be implicit in the laws of physics. The spongeam is in effect the depository of the up-front-information which is a prerequisite of evolution and also OOL. More about  the spongeam can be found in these references......

https://quantumnonlinearity.blogspot.com/2016/01/necessary-conditions-for-evolution.html

https://quantumnonlinearity.blogspot.com/2016/02/on-structuralism-and-spongeam.html

https://quantumnonlinearity.blogspot.com/2016/03/the-mathematics-of-spongeam.html

https://quantumnonlinearity.blogspot.com/2019/09/evolution-naked-chance.html

Although I think reasonable atheists would accept that evolution requires a burden of up-front-information there may still be resistance to this idea because it then raises the question of "Where from?". Like IDists who are determined to peddle the notion of information conservation some atheists are constantly drawn toward the "something from nothing" paradigm.  See for example the discussion I published in the comments section of the post here where an atheist just couldn't accept that the natural physical regime must be so constrained that evolution effectively has direction. He may have been of the "naked chance" persuasion.   

3. Generating complexity

As we have seen in this series, random configurations are identified by the length of the algorithm needed to define them: Viz: If the defining algorithm needs to be of a length equal to the length of the configuration then that configuration is identified as random. However, that a random configuration can only be defined by an algorithm of the same length doesn't mean to say that the random configuration cannot be generated by algorithms shorter than the length of the random configuration: After all, a simple algorithm that systematically generates configurations, like say counting algorithms, will, if given enough time, generate any finite random sequence. But as I show in my book on Disorder and Randomness small space parallel processing algorithms will only generate randomness by consuming huge amounts of time. So basically generating randomness from simple beginnings takes a very long time (actually, by definition!). In this sense randomness has a high computational complexity. 

Although complex organised entities like biological structures obviously do not classify as random configurations they do have properties in common with randomness in that they show a great variety of  sub-configurational patterns and are potentially a huge class of possible configuration (but obviously a lot smaller than the class of random configurations). Therefore it is very likely that such complex configurations as living structures, being somewhere between high order and high disorder in complexity, are themselves going to take a long time to generate from algorithmic simplicity, much longer time, I'll wager, than the age of the universe even given the constraining effect of the laws of physics. This would mean that to generate life in a cause and effect cosmos and in a timely way sufficient up front information (such as the spongeam) must be built into the cause and effect algorithms from the outset. The cause an effect paradigm, even if used probabilistically, requires that the outcomes of a process are implicit (if only probabilistically) in the current state of the procedure. But if there is insufficient up-front-information built into a cause and effect system to drive the generation of life in a short enough time how could it be done otherwise?  If we imagine that there was no spongeam could life still be arrived at?  I believe there are other possibilities and ideas to explore here.

4. Declarative languages and computation

In conventional evolution (and OOL) the potential for life is thought to be built into a physical regime, a regime driven from behind in a cause and effect way. But if those cause and effect laws are simple parallel imperative algorithms and provide insufficient constraint (i.e. insufficient up front information) life can then only be developed by consuming considerable time and space resources, more time and space, in fact, than the known cosmos provides. So, as I have proposed in my Thinknet and Meloncolia I projects one way of solving the generation time problem is to use expanding parallelism. But for this technique to work there is, however, another important ingredient needed here. This is the declarative ingredient which means that what is generated is subject to selection and therefore in a declarative context the algorithms are explicitly teleological and driven by a goal seeking intentionality.

Most procedural programs are actually implicitly teleological in that they are written in an imperative language with the aim of producing some useful output; that is, an end product. But in a true declarative program the procedures aren't written down but rather the declarative language itself is used to state the desired end product in a logical and mathematical way and the compiler translates this formal statement into procedures. A practical example of a simple declarative language would be as follows: 

5. Example of a declarative language

Our problem might be this: Is there a number that has the following set of properties: Its digits add up to a specified number N1, it is exactly divisible by another specified number N2 and it is also a palindrome. This is a declaration of intention to find if such a number or numbers exist. One way to conceive the solution of this problem is to imagine all the natural numbers actually exist, at least platonically as locations in some huge imaginary "memory" that can be flagged by a signalling processes; in this sense the availability of memory space is assumed not to be an issue. In fact we imagine we have three processes at work; Viz: Process A which systematically flags numbers whose digits add up to N1, process B which systematically flags numbers which are multiples of N2 and finally process C which systematically flags palindromes. If all these three processes flag the same number then we will have a solution to the stated problem (there may be more than one solution of course). Hence a solution, or a range of solutions, can then be selected as the goal of the cluster of processes. 

A method of this sort could be employed for other mathematical properties of numbers. Hence we could extend the tool box of flagging processes indefinitely, A, B, C, D, E....etc. Each of these labels represents a process which generates a particular kind of number. So we could then make declarations of intent Thinknet style such as:

[A B C  D] 

1.0

This expression represents a simple declarative program for finding numbers with all the properties flagged by A, B , C and D. This search is a two stage operation: Firstly the configuration A B C D represents the operation of forming halos of numbers flagged with their respective properties.  The square brackets [ ] represents the operation of making a selection of those numbers which simultaneously have all the properties that the processes A, B, C and D flag.

In analogy to Thinknet we could further sophisticate this simple language by allowing nesting; that is:

[[A B] C D]

2.0

...where the nest [A B] is solved first by selecting a set of solutions which have properties flagged by both A and B. Processes C and D then run, flagging their respective properties. The outer brackets are eventually applied to make a selection of numbers which simultaneously have all the properties flagged by A, B, C and D.  It is likely that different bracketing scenarios will come up with different results.  Hence in general:

[[A B] C D] != [A B [C D]]

3.0

No doubt there is a lot of scope for sophisticating this simple declarative language further - for example we could have processes which seek numbers which do not have certain declared properties; these are what you might call negation processes: Hence we might have:

[A !B]

4.0

...where !B means numbers that don't have the property flagged by the process B.  Other logical operators could no doubt be added to this language. 

Clearly the details of the processes A B C etc. are procedural; that is, they must be programmed in advance using a procedural cause & effect  language.  Nevertheless,  the foregoing provides a simple example of a declarative language similar to Thinknet where once the procedural work has been done setting up A, B, C... etc, the language can then be used to construct declarative programs. However, simple though this language is, practically implementing it is beyond current computing technology: Firstly it is assumed that working memory space isn't limited. Secondly, the processes A, B, C,.... and [ ] would be required to operate with huge halos of possible candidate numbers; this latter requirement really demands that the seek, flag and selection processes have available to them an expanding parallelism rather than the limited parallel computation of our current computing technology.

In and of themselves the flagging procedures designated A.B, C etc. do not appear to obviously manifest any goal driven behaviour: it is only when A, B, C etc are brought together with [ ] that the teleology of the system emerges. But having said that we must realise that if it were possible to implement the above declarative language on a computer it would in fact be only a simulation of declarative mechanics. For in a practical computer not only would the procedural algorithms defining the A, B and C reside as information in the computer memory from the start, but so also would the code needed to implement the selection process [ ]. Thus a human implementation of declarative computing has to be simulated with cause and effect software. The usual computational complexity theorems would therefore still apply to these strings of code. But although in human designed computers the seek and select program strings are found in computer memory from the outset, this doesn't apply in nature. After all, we don't actually observe the laws of physics themselves; all we observe is their apparent controlling and organizing affects on our observations. Thus physical laws are more in the character of a transcendent reality controlling the flow of events. Likewise if there is such as thing as teleological  selection criteria in our cosmos then they too would, I guess, be a transcendent reality and only observed in their effects on the material world. Nature as far as we can tell doesn't have a place where we can go to see its stored generating programs doing the seeking and selecting. But when we do conventional computing we can only simulate transcendent generation and selection laws with explicit program strings residing observably in memory.

6. Creating/generating Information?

I've schematically represented a declarative computation  as:

[A B C] => Output

5.0

...where A, B, C, are cause and effect processes which (perhaps using expanding parallelism) flag configurational objects with particular properties. The square brackets represent the operation of making a selection of configurations which simultaneously possess the sought for properties flagged by A, B and C. This selection is designated by "Output". 

The computational complexity of the computation represented by 5.0 is measured by two resources:

a) If the above operation were to be rendered in conventional computation there would be programs strings for A, B, C and [ ] which when executed would simulate the generation and selection of configurations. The length of that string would be one aspect of the complexity of the computation.

b) The second measure is the count of operations required to reach the output. If we are simulating the declarative computation using parallel processing then linear time would be one way to measure the complexity, but if we are using expanding parallelism it is better to measure the complexity in terms of the total count of the number of parallel operations. 

In light of this paradigm is it right to make claims about either information being conserved and/or information being created?  As we will see below the answer to that question is 'yes' and 'no' depending on what perspective one takes. 

Firstly let us note that 'information' is, connotatively speaking, a bad term because it suggests something static like a filling cabinet full of documents. In contrast we can see that expression 5.0 is in actual fact a representation of both static information and computational activity. If we are going to perceive 'information' in purely static configurational terms then 5.0 clearly creates information in the sense that it creates configurations and then flags and selects them; the teleological rationale being that of finding, flagging and selecting otherwise unknown configurations which are solutions to the stated problem. So, secondly we note that the creation of  the configurations which are solutions to the stated problem cannot take place without computational activity. 

Spurious ideas about information somehow being conserved may well originate in the practical problem of the storage of configurational information where storage space is limited: Algorithmic information theory is often concerned with how to map a large string to a shorter string, that is, how to compress the string for the purpose of convenient storage without loosing any information. Here algorithmic information theory reveals some kind of conservation law in that clearly the information in a string has a limit beyond which it can not be compressed further without loosing information. In fact as is well known a truly random configuration can not be compressed at all without losing information. In this "finite filing cabinet" view of information, a view which deals with configuration, (as opposed to activity) we find that the minimum length a string can be compressed without loss of information is fixed; in that sense we have a conservation law. 

But when we are talking about the generation and selection of configurations, configurational information isn't conserved; in fact the intention is to create and flag otherwise unknown configurations. Moreover, we are not talking here about an informational mapping relationship because this activity need not be one that halts but just continues generating more and more complex configurations that fulfill the criteria A, B, C, ...etc. Nevertheless, it may still be possible that some kind of conservation law could be formally constructed if we mathematically bundle together the program strings needed for A, B, C and [ ] along with the quantity of activity needed to arrive at a selection. Hence we might be able to formalise a conservation of information along the lines of something like: 

Program strings + activity = output.

6.0

...that is the computation of the required output is the "sum" of the initial static information and the amount of activity needed to arrive at a result; here initial information and computational activity have a complementary relation. Of course, at this stage expression 6.0 is not a rigorously formal relationship but really represents the kind of relationship we might look for if the matter was pursued. 

As I noted toward the end of this paper in my Thinknet project the declarative paradigm symbolised by 5.0 provides a very natural way of thinking about "specified complexity".  As we have seen the word 'specified' connotes some kind of demand or requirement conceived in advance; hence the word 'specified' has a teleological content, a goal aimed for. 'Complexity' on the other hand will be a measure of both the initial static information and computational activity required to meet the stated goal. These two aspects are more formally expressed in relationship 5.0 where the specifications are represented by properties flagged by processes A, B and C and the complexity of the computation is measured by 6.0.

I would like to moot the idea that expression 5.0, where its form is understood to apply in an abstract and general way, is an important aspect of intelligent activity. This is of course by no means a sufficient condition of the dynamics of intelligence and in fact only represents one aspect, albeit a necessary condition, of intelligence; namely that of goal seeking. 

7. The poverty of de facto ID. 

The de facto ID movement comes with strong vested interests. Firstly there are political interests: North American de facto ID finds itself leaning into the arms the right-wing, although to be fair this may in part be a reaction against some of the left slanting atheism of the academic establishment. Secondly there are intellectual interests. As we have seen de facto ID is all but irreversibly committed to a paradigm of intelligence that is beyond human investigation in so far as they have mooted the concept that intelligence is some kind of oracular magic that cannot be simulated in algorithmic terms. This has naturally fitted in with their dualistic outlook which sets intelligent agency over and against the "natural forces" of nature, forces which are thought of by them to be too profane, inferior and "material" to be at the root of a putatively "immaterial" mind with power to create information in a mysterious magical way. This almost gnostic outlook neglects all the potentiality implied by that fact that nature (which includes ourselves) is an immanent God's creation. 

What may not have helped the de facto IDists is that current scientific attitudes are slanted almost exclusively in favour of a cause and effect view of the physical regime, a regime where there is no room for "final causes" i.e teleology. In the cause and effect paradigm the future is seen to be exclusively written into past events (at  least probabilistically) and little or no credence is given to the possibility that there may be transcendent selection effects waiting invisibly in the wings. 

As we have seen, without explicitly referring to the dynamics of end results and/or intentionality it is very difficult to define "specified complexity". This has in fact hamstrung the de facto IDists attempts to define it themselves. This has lead N&H to define "specified complexity" in terms of static configurations alone thus neglecting the important dynamic aspect of information generation. According to N&H a configuration is judged to have Algorithmic Specified Complexity (ASC) if it has a high improbability but with a low relative algorithmic complexity. Once again they've non-noncommittally hedged on the concept of intelligence by placing it beyond analysis into the hidden libraries of relative algorithmic complexity. The result is a definition of specified complexity that is full of loopholes as we have seen: At best it can identify complex organisation. But as we have also seen this doesn't always work and moreover ASC isn't as strongly conserved as they claim it is; it is poor definition that is far from robust and gets nowhere near identifying the presence of intentionality. Their problem traces back to the fact that they are trying to identify the operation of intelligence without cognisance of the dynamics of intentionality and goal seeking.  

The intrinsic configurational properties of an object such as an unexpected degree of ordered complexity are not reliable predictors as to the operation of intentionality; they may be just incidental to the cause and effect processes.  When we look at objects of archaeological interest, whether they be complex or simple, we look for evidence of intentionality. But we can only do that because we are human and have some inkling of the kind of things humans do and the goals that motivate them. In archaeological work, as in police work, attempting to identify the presence of purpose, (that is identifying an underlying teleology) is a feature of that work. It is ironic that the atheist Joe Felsenstein should spot the inadequacy of N&H's definition to cope with even the pseudo-declarative nature of standard evolution. 

The fact is N&H haven't really grasped the concept of specified in information in terms of a dynamic declarative paradigm and therefore have failed to come up with a useful understanding of the intelligent activity. Given that their vision goes no further than procedural algorithmics and configurational compressibility (connections where the information is in one sense present from the start) it is no surprise that they think that information, which they perceive as a very static object,  is conserved. In contrast the whole rationale of intelligent action is that it is a non halting process forever seeking new configurations & forms and thereby creating them. Intelligent output is nothing if creative. This is what I call intelligent creation. 

***

Summing up

As we have seen even an atheist like Joe Felsenstein is tempted to accept that "specified information" makes little sense outside of a teleological context and that is why evolution is conveniently conceived in pseudo-teleological terms - that is, in terms of its end result - when in fact with evolution, as currently understood, all the cause and effect information must be built in from the outset. Of course, for a true-blue atheist any teleological rendition of evolution is at best a mental convenience and at worst a crass anthropomorphism.  For myself, however, I have doubts that even given the (procedural) laws of physics there is sufficient built-in information to get to where we are now biologically with a realistic probability after a mere few billion years of the visible universe. Hence, I'm drawn toward the heretical idea that both expanding parallelism and that transcendent seek and selection criteria are somehow built into nature.

Do these these notions of the teleology of declarative problem solving help to fill out the details of the mechanisms behind natural history? If we understand "evolution" in the weaker sense of simply being a history of macro-changes in phenotypes and genotypes, then what goals are being pursued and what selection criteria are being used apart from viability of form? How are the selection criteria being applied? What role, if any, does quantum mechanics have given that it looks suspiciously like a declarative process which uses expanding parallelism and selection?  l'll just have to wait on further insight; if it comes. 

Intelligence, Oracles, Magic and Politics

The de facto ID concept of intelligence.

As I have remarked many times on this blog the de facto Intelligent Design movement affects to leave the internal details of the "intelligence" they believe to have stepped in and directly created life as a mystery. There is some justification in this policy: When handling great mysteries (e.g. Divinity) caution is sometimes the better part of valour and so it may be best to proceed apophatically; that is, to define the mystery in terms of what it isn't. An apophatic approach to intelligence seems to be stock in trade of the de facto ID community in North America. In fact as far as I can tell the mainstream IDists believe that the intelligent agent which created life is neither explicable in terms of so-called "natural forces" or even for that matter any process which has the potential to be expressed algorithmically no matter how complex that algorithm may be. I find their views a little ironic: As many of them make claim to a Christian faith one might think that those so-called "natural forces" which we as Christians believe to be God's sublime Creation may hold one or two surprises for us as to what these "forces" (under Sovereign management) can do; after all, Quantum Mechanics alone has left enlightenment humankind thoroughly perplexed as to what it all means (For a start it is no longer meaningful to talk of matter as having identity of substance; identity comes via configuration). But no, in the de-facto IDist world  the "profane natural forces" vs "sacred intelligent agency" dichotomy is their habitual thesis and anti-thesis. In their view "matter" is too debased and inferior to be a secondary source of the dignified sublimity of  mind.

So, in the light of all this I was not in the least bit surprised to find a post on the de facto Intelligent Design website Uncommon Descent with links to ID material  giving the clearest evidence I've yet seen that de facto ID prefers to think about true "Intelligence" as a property tantamount to a magical power, setting it apart from anything else we encounter in this world^*1. The UD post in question alerts us to one of de facto ID's gurus who is attempting to identify human intelligence as having the ability to act as a "partial halting oracle". That is, it is assumed that human intelligence is an oracle which can in some (but not all) cases solve the halting problem. According to Wiki. the concept of an "Oracle" as used in computational theory is defined as follows:

An oracle machine can be conceived as a Turing machine connected to an oracle. The oracle, in this context, is an entity capable of solving some problem, which for example may be a decision problem or a function problem. The problem does not have to be computable; the oracle is not assumed to be a Turing machine or computer program. The oracle is simply a "black box"  that is able to produce a solution for any instance of a given computational problem:

A "black box" capable of doing the right thing sums up those inscrutable oracular powers. This manoeuvre by an IDist guru well and truly places the essence of intelligence all but beyond analytical probing ^*2. As I have said many times before the de facto IDist's preference for an esoteric notion of intelligence traces back to their use of their "explanatory filter" which once it has been used to settle on intelligent agency as the cause of a pattern doesn't really allow one to proceed much further. This of course contrasts with my own approach to intelligence which doesn't resort to super-analytical processes; well nearly: In my Thinknet project I see intelligence as a teleologically driven search process by a "Thinknet" like system. Thinknet systems are potentially chaotic which means that they can amplify those quantum ambiguities up to the macroscopic level, ambiguities which if they remained un-collapsed would give us people who could be in two places at once. Well, we can't have that at the macroscopic level so if the mind is constantly collapsing those wave-function, then, I tender, it is this process of constant collapse which generates consciousness.  But if the mind amplifies those apparent random collapses  up to macroscopic level there is therefore the potential for it to manifest that great incomputable - absolute randomness; so in that sense mind has an incomputable aspect to it. Nevertheless, what I'm proposing is no blackbox concept of intelligence: I'm working on a notion of intelligence that is much more resolvable than ID's magical oracular black box and this is why I have to sophisticate the explanatory filter.

Turning to my subjective perspective on my own thought life I must say that it certainly doesn't feel like some magical oracle able to coolly solve a problem just like that! In contrast problem solving requires the hard graft of mental searching as one attempts to make connections which lead to solutions. To me my thought life feels much more like the seek, reject and select trial & error grind of a Thinknet search than it does ID's magical oracle where genius solutions just pop into the head. I see the hard work associated with thinking as a consequence of the overheard incurred by using a very general-purpose thinking system with a general purpose connectionist language to solve the generic problem; as this system is a jack-of-all trades-problem-solver it can be slow at solving specialised problems as it has to first translate the problem into its connecionist terms.

I don't have a strong claim to having clinched the essence of intelligence anymore than do the defacto IDists. But like myself they have just as much right to investigate an avenue of possibility in their search for what intelligence is about. In fact I believe their presence is a good thing; the more people investigating different avenues the better. For all I know the IDists might be right! Also, like the IDists I believe that intelligence of some sort underpins the nature of the cosmos.  So under any other circumstances I would applaud the IDists efforts at tentatively trying to move forward with something new; after all that's science for you.  But I'm afraid in this case I can't applaud. Why is that?

***

Well, the answer to that is politics especially the politics in North America. It's the catalyst that has precipitated and hardened a "natural forces" vs "intelligent agency" polarisation. The IDists are persona non grata among the academic establishment and so it is no surprise that these IDists have been tempted to put all their eggs into the "intelligence-is-magic" basket in order to batter academia's evolutionary and algorithmic rendering of the processes of life, processes the academics believe to have generated human intelligence. Some times I wonder if the de facto ID people aren't really being serious with their proposals and simply come up with their stuff just to rile the academic establishment!

But the politics doesn't stop there. IDism is all part of a greater right vs left wing tribal conflict which means that the right wing sharply disagree with the government tenured academics over one or more of a set of well contended issues (as mentioned in my last blog post): e.g. vaccinations, climate change, gay rights, deep government conspiracy theories, the regulatory role of government, the covid-19 lock down, hyper-market libertarianism, gun rights etc. The common underlying theme running through all this is the diffidence right-wingers have toward central government interventions; no!, make that the status quo interventions:  When it comes down to it the right-wing is just as capable of supplying individuals of totalitarian inclination as any other human sub-culture, if not more so. Do you think those characters one finds in America's quasi-militias would have the slightest respect for the argumentative cut and thrust of an authentic parliament? Unlikely: More to their taste would be for one of their plutocrats to do a Cromwell and clear parliament using AR-15 armed thugs.

Crackpot daftness can be found on the extremes of both left and right, but my argument here is with the right-wingery of the de facto ID community. Right wing sentiments ultimately drive their all but exclusive commitment to an Oracular paradigm of intelligence. They've backed  themselves into the cramped corner of this paradigm because they are suspicious of those government tenured academics who for the most part will get rubbed up the wrong way by de facto ID's support of oracular intelligence.

The republican language coming out of England's 1642 civil war fed into the American war of independence (from tax) and now the North American right-wing endlessly recapitulate the sentiments of this language Viz: interference coming from a tax funded government is at best regarded with suspicion and at worse as evidence of a deep government conspiracy.  For example, on Uncommon Descent one can find references to "climate change alarmism" and also "covid-19 lock down alarmism". The emotive term "alarmism" is the keyword expressing right-wing apprehensions about projects largely emanating from government sponsored tax funded bodies. In my view coordinating the social responses to the black swans of climate change and covid-19 requires centralised information and control; such a response is well beyond the powers of the sluggish market with its distributed blind-watch-maker decisions. But such government involvement is the right-winger's worst nightmare come true; especially if government should muff it (which they often do!)

The pretext supporting the "libertarian" polemic about covid-19 and climate change is, however, entirely plausible if not sound: The world's wealth generating markets could be so affected by central government policies that it causes huge economic hardship or perhaps even an apocalyptic economic collapse. But this line of argument cuts both ways. Covid-19 and climate change, if left to run their courses, could conceivably also cause economic collapse. Moreover, the right wing's emotive language can be used against them: One might accuse them of promulgating "economic hardship alarmism", or "totalitarian new world order alarmism". Both sides are faced by the same dilemma: The  fix may be worse than the problem!

Whilst I strongly reject the border-line Marxism and anti-theism found among some academics, neither can I support the right-wing affectation for so-called libertarianism. Libertarianism is to the free market as fundamentalism is to Christianity; they are the kiss of death for the things they purport to uphold. Sociopathic libertarianism is a source of social disaffection thus helping to serve up a discontented society on a platter to either Marxist or right-wing dictators. For example, allowing covid-19 to take its course is likely to strike harder among the poor than the rich and therefore this solution to our problems is readily perceived as the solution in favour of the rich. Moreover, self-branded "libertarianism" with its connotation of "liberty" comes under the heading of "self-praise is no recommendation": Looking at the mix of potential plutocrats, domineering characters and the well armed quasi-militias (in America) who make claim to the name "libertarian" it is easy to imagine a would be dictator arising from their ranks. And it wouldn't be the first time that "liberty" and "hegemony" have walked hand in hand; let's recall the outcomes of the English civil war of 1642, the French revolution of 1789, the October revolution and Mao's China. Idealism and hegemony are closely linked.

The many wildcards of socio-economics don't stop some people thinking they are clever predictors and planners. The open-endedness of socio-economic systems is a bottomless pit of new data that can be cherry-picked and tailored to support the favoured planning polemic. In a chaotic world human beings are necessarily complex adaptive systems and therefore by definition much better opportunists than they are planners. They make their decisions and take their opportunities on the hoof. Like other biological organisms society is a mix of central as well as distributed control and this mix no doubt better suits a chaotic world where black swans create new problems and at the same time deliver otherwise unforeseen opportunities. But the time honoured overriding concern of human beings is that of hanging on to the immediacies of survival at all costs and that's probably why many people favour social distancing rather than the long shot of saving an abstract economic system that more likely favours lining the rich man's pockets in his ivory tower before it gets to line your pockets (if you've survived covid-19!). While there's life there is hope, hope that the new opportunities open up into vistas of  fruitful originality and prosperity.  We can only plant and water; it is God that gives growth.

POSTSCRIPT 

27/4/20

In a post on Uncommon Descent that I wouldn't necessarily want to take issue with we find an interesting comment from a character called "Polistra". Viz:

Polistra April 26, 2020 at 2:48 pm

This is silly and illogical. It wasn’t the virus that stopped the world.

The virus just wandered around and found tissues to infect, and the humans who own the tissues killed the virus using standard weapons and tactics. A very few humans were unable to maintain the war, and they lost.

The world was stopped by GOVERNMENTS. The virus was just the latest fake “reason” for stopping the world.

This commentator doesn't like the fact that the UD post suggests 900 bytes of covid-19 DNA is the reason why the world has shut down. Polistra clearly wants a much clearer statement that the culpability lies with GOVERNMENTS.  Polistra doesn't tell us why governments want to shut the world down with what he calls a "fake" reason any more than flat earthers will tell us why the UN wants us to believe in a spherical earth instead of their flat earth. Although I don't think most UDers would go along with this kind of conspiracy theorism it's probably significant that they don't challenge him: He's one of them, he's part of their anti-government tribe!  The irony is, as I have already said, that it's so easy to see dictators readily emerging from the ranks of the domineering fanatical right wingers if they should ever get power.


Footnotes

*1 I'm not quite sure how this works out with human beings, objects which from a third person perspective are observed to be entirely a product of  complex organisations of  God's atoms.

*2 Turing's halting theorem and Godel's incompleteness theorem are closely related in that both use  the "runaway self-referencing" reasoning found in the diagonalisation procedure. Roger Penrose proposed that the human ability to understand Godel's argument proved that human thinking was an incomputable process. Hence Penrises ideas are also favoured by IDists. Whilst it is wrong to dismiss Penrose outright I have submitted my reasons why I don't follow him down this particular avenue..

Breaking Through the Information Barrier In Natural History. Part I

(This post is still undergoing correction and enhancement)

Propeller technology was always going to have a problem

breaking the sound barrier whereas jet technology didn't.

I was fascinated to read this post on Panda's Thumb by mathematical evolutionist Joe Felsenstein. The post is about the application of Algorithmic Information theory by Intelligent Design theorists to the evolution question. Felsenstein is rather concerned that these IDists are attempting to use Algorithmic Information to prove (yet again) that evolution is impossible. But once again their attempts go awry: They are over interpreting a genuine complexity/improbability barrier as an impossibility barrier. But it turns out to be no more an impossibility barrier than the sound barrier was to flight; with the right technology the barrier can be broken. And once again they interpret the barrier as the sign of some kind of information conservation law which prevents so-called "natural forces" bringing about the emergence of life.

But having said that let me say that my own position occupies a space somewhere between, on the one hand, the IDists who don't abide by evolution because it uses what they believe to be creatively inferior so-called "natural forces" and on the other the atheists who are determined to show that evolution is a cinch and more-or-less in the bag. The fact is some evolutionists (although this may not apply to Felsenstein) do not fully appreciate the information barrier that evolution actually presents (See Larry Moran for example). In fact the implicit parallel computational paradigm found in the standard concept of evolution is not going to be up to the task unless it starts with a huge burden of up-front information. This is where I believe IDists like William Dembski's work is relevant and valid as I have said before, although Dembski and his ID interpreters have inferred that Dembski's work also implies some kind  of "Conservation of Information", whatever that means.

The consequence of IDists over interpreting evolution's information barrier in terms of an absolute barrier is that they then conclude they have in their hands proof that "natural forces" cannot generate life and that some extra "magic" is needed to inject information into natural history in order to arrive at bio-configurations. They identify that extra magic not as the "supernatural" (that would look too "unscientific"!) but instead as "Intelligent agency"

To a Christian like myself, however, this IDist philosophy raises questions: For although there is a clear creation vs God dualism in Christian theology I find the implicit dualism within creation implied by de facto IDism problematic: If an omniscient and omnipotent God has created so-called "natural forces" then it would seem to be quite within his capabilities to provision creation in such a way that natural history could conceivably include the "natural" emergence of living configurations. The "magic" may already be there for all we know!

Moreover, it is clear that human intelligence, which is one of the processes of the created order, can "create information" and I don't think the IDists would deny that. And yet as far as we know human intelligence appears not to transcend God's created providence. IDists, however,  are likely to attempt to get round this observation by trying to maintain that human intelligence has a mysterious and extraordinary ingredient which allows it to create information - for example, I have seen IDists use Roger Penrose's idea that human intelligence involves incomputable processes as the mysterious super-duper magic needed to create information. Penrose's ideas, if correct, imply that human intelligence (and presumably the intelligence that IDists claim on occasions injects information into  natural history) cannot be described algorithmically. If this line of argument can be maintained then it would justify the IDists dualism. But this IDist paradigm can be challenged. For a start I believe Penrose was wrong in his argument about the incomputablility of human thinking; see here and here. Yes, human thinking may have some extraordinary ingredient of which we are unware, but it may have been part of the covenant of creation all along: I don't, however,  believe it to be an incomputable process.

This post (and the next post) is my take on the "information barrier" debate between evolutionists and IDists. I will not be going into the minutiae of how the IDists or their antagonists arrive at their conclusions but I will be looking at the conclusions themselves and comparing them with my own conclusions based on three projects of mine which throw light on the subject. Viz:

1. Disorder & Randomness. This project defines randomness algorithmically. 

2. The Melancholia I project: This project discusses the possibility of information creation and/or destruction.

3. The Thinknet project: This project defines the term "specified information" teleologically and notes the parallels with quantum mechanics.

Summarising my position: I can go some of the way with Dembski and the IDists in that there is an issue with standard evolution in so far as it demands some mysterious up-front information in order to work, but there is no such thing as "the conservation of information"; information can be destroyed and created; but in any case the arguments used by IDists are unsafe because there is more than one understanding of what "information" actually is. It is likely that the IDist's "conservation of information" may result because we are most familiar with linear and parallel computing resources, a computing paradigm that has difficulty breaking through the information barrier. This contrasts, as we shall see, with the exponential resources of expanding parallelism and the presence of these resources exorcises the dualists ghost in the machine which haunts IDist philosophy. On the other hand some atheists are unaware that there is an information barrier (probably not true of Joe Felsenstein - see here and here) and are unlikely to see Dembski's work as laying down a serious challenge.

As usual I don't dogmatically push my own ideas from a polemical partisan soap box seeking conversions to my case. For me this is a personal quest, an adventure and journey through the spaces of the mind which quite likely may not lead anywhere. The journey, as I often say, is better than the destination.

***

In this post I want to introduce the information barrier via William Dembski's work. In the video of a lecture I embedded in my blog post here Dembski introduces his concept of the "conservation of information" via the following simple relationship:

Probability of life being generated by a given physical regime =  r <  p/q

1.0

....where p is the unconditional probability of life and where q is the conditional probability of life given a physical regime whose probability is r.

I give an elementary proof of this theorem in the said blog post. Relationship 1.0 will tell us what we are looking for if we rearrange it a bit. Viz:

q < p/r

2.0

Now, we expect p, which is the unconditional probability of life, to be very small; that is, if we were using a computational method which involved choosing molecular configurations at random, then it is fairly obvious that the complex organisation of configurations capable of self replication and self perpetuation will, by virtue of their rarity in configuration space, have an absolutely tiny probability. If we want to fix this problem of improbability and get a realistic chance of life emerging then conceivably we could contrive some physical regime whereby the chance of life coming about with a better than a random selection computation is q, where q >> p and where q is the conditional probability of life; that is the probability of life given the context of the physical regime.  But if q is to be realistic then from 2.0 it follows that we must have r ~ p; that is, the only way of increasing the conditional probability of life is to first select a highly improbable physical regime. As Dembski points out the improbability has now been shifted onto the probability of the physical regime. Dembski's point becomes clearer if we convert our probabilities to formal "information" values as follows.

Now, the so-called information function I(p), as used by Dembski, is defined for a probability of p using:

I(p) = - ln (p)

3.0

...where ln is the natural logarithm function. From this definition it is clear that for very small values of p function 3.0 is going to return a large value of I; that is a large "information" value.

The so-called "conservation of information" becomes clearer if we first take the natural log of expression 2.0, followed by applying definition 3.0 and then with a bit of rearranging we arrive at:

  I(q) + I(r) = I(p) 

4.0

Those looking for "natural explanations" don't expect the emergence of life from a "natural" physical regime to be a surprise but rather a very "natural" outcome given that regime. This is tantamount to requiring that I(q), the "surprisal" value of the conditional emergence of life, to be relatively low. Trouble is, because I(p) is so high, it follows from 4.0 that if I(q) is to be low then I(r), the information value of the physical regime, is necessarily very high. Relationship 4.0 being effectively a "zero sum game" expression  means that something has to soak up the information; either I(q) or I(r) or both. We are therefore always destined to be surprised by the extreme contingency nature flings at us from somewhere within equation 4.0. So, at first sight we seem to have an information conservation law expressed by 4.0.

Relationship 4.0 is in fact borne out by a closer look at conventional evolution, a process which somehow generates structures that in an absolute sense are highly improbable. Joe Felsenstein himself implicitly acknowledges equation 4.0 in his suggestion that the information for life is embodied in the physical regime we call the "laws of physics". (see here and here). If so then from 4.0 we infer that these laws must be a very improbable state of affairs and therefore of very high information. Evolution as it is currently conceived requires that this information expresses itself in what I call the "spongeam" about which I say more in this blog post. (Actually, my opinion is that the spongeam doesn't exist and that some other provision applies - more about that another time)

Equation 4.0 is beguiling: It seems to come out of some simple and rigorous mathematics. But it embeds an assumption. That assumption is that I(p) has a very high value because we assume from the outset a computation method which involves a serial "throwing of the die" as it were, a method which is going to require many, many conventional computational serial steps and therefore has a prohibitively high time-complexity as far as practice is concerned*². But then if we have dice rather than just a die we can then have more than one trial at a time and the chances of creating life by chance alone increase, although it is clear that there would have to be an enormous number of parallel trials to return a significant probability of generating living configurations in this way. This multi-trials technique is effectively the brute force resort of the multiverse extremists. It is a fairly trivial conclusion that increasing the number of parallel trials has the effect of "destroying" information in that increasing trial numbers increases the probability of an outcome and so its information value goes down. Clearly in the face of huge numbers of parallel trials an outcome, no matter how oddly contingent it might be, is no longer a "surprise" in such a "multiverse". Not surprisingly this concept appeals to anti-theists who feel more at home in a Godless multiverse.

In the paper linked to in this post of the Melancholia I project I looked into the effect of increasing parallel trial numbers and in particular I considered the subject of expanding parallelism in the generation of outcomes. It's a fairly obvious conclusion that increasing parallel trials increases the probability of a result! But it also goes to show, perhaps a little less obviously, that information isn't conserved in such a context; in fact in this context information is effectively destroyed by the increasing trial numbers and in particular by expanding parallelism. This, sort of thing is likely to go down well with anti-theists because the "surprisal" value ( i.e. -ln(p) ) associated with outcomes is eroded, although of course anti-theists may still be surprised that such a multiplying system exists in the first place!

As I contend in this post multiverse ideas which posit a sufficient number of trials needed to destroy our surprise at the universe's amazing organised contingencies leaves us looking out on a cosmos whose aspect is empty, purposeless and anthropologically meaningless. And yes, I say it again; this kind of universe suits the anti-theists down to the ground; it seems to be the sort of universe they eminently prefer. But in spite of that there is something to take away from these multiverse ideas, in particular the idea of expanding parallelism, hinted at by quantum mechanics, and which  is evidence of the potential availability of huge computational resources. Given the concept of omniscience & omnipotence implicit in the Christian notion of God, positing the existence of these huge computational resources doesn't seem so outrageous. But in a Christian context the computational potential of expanding parallelism has, I suggest, purpose and teleology and is in fact evidence of a declarative seek, reject and select computational paradigm.

***

Although the -ln(p) concept of information used by Dembski succeeds in quantifying some of our intuitions about information it does have some notable inadequacies and it is these inadequacies which take us on to the subject of Felsenstein's Panda's Thumb post, namely Algorithmic Information theory. Let me explain...

Ironically I called the paper that explores the subject of expanding parallelism "Creating Information" rather Destroying Information. This is because my Melancholia I project is really about a concept of information very different to -ln(p). The need for this different concept of information becomes apparent from the following considerations. Although the function -ln(p) adequately quantifies our level of surprisal at outcomes this definition of information is not good at conveying the idea of configurational information. Take this example: The chances of finding a hydrogen atom at a designated point in the high vacuum of space is very small and therefore we have  a high information event here should it happen. But it is a very elementary event, an event which only conveys one bit of information: 'yes' or 'no' depending on whether a hydrogen atom appears or not. The trouble is that a one bit configuration is hardly what one would like to call a lot of information! Therefore we need something that is better at conveying quantity of information. From the function -ln(p) it follows that a one bit configuration can "contain" the same amount of information as a large n-bit configuration. This doesn't feel very intuitive, particularly if we are dealing with potentially large and complex configurations; it seems intuitively disagreeable to classify a complex configuration as possibly having the same level of information as a one bit configuration.*¹

Algorithmic information theory attempts to measure the information content of a configuration via its computational complexity and this returns a measure of information which agrees with our intuitive ideas about the quantity of information found in a configuration, something that -In(p) doesn't necessarily convey.  However, using this concept of information we find that once again the IDists think they have stumbled on another information barrier that scuppers any creation of life by those inferior but dreaded "natural forces"! In the next post this contention will take me into the subject of my book on Disorder and randomness which also deals with algorithmic information theory. Once again we will find that expanding parallelism bursts through the information barrier. Although Joe Felsenstein and his buddies certainly won't need any help from me to engage the IDists I will in fact be using my own concept of Algorithmic Information to look into the IDist claims because it provides me with something immediately to hand.

There is one more ingredient that needs to be added to the mix to complete the picture of information creation and this is an ingredient which will certainly not be to the taste of anti-theists whose world view is one of a purposeless cosmos without teleology. I'm talking of my speculative proposal that the cosmos isn't working to some meaningless and mindless procedural process that just goes on and on and leads to nowhere but rather is operating some kind of purposeful declarative computation that uses expanding parallelism to seek, reject and select outcomes. It is in this context that the notion of specified information suddenly jumps into sharp focus; in fact I touch on this subject toward the end of the paper I've linked to in part 4 of my Thinknet project. (See section 11).

I have to confess that if the cosmos is using a purposeful declarative computational paradigm that makes use of expanding parallelism I'm far from having all the details: All I have currently is an understanding of the effect that expanding parallelism has on computational complexity and the metaphor of my Thinknet project which seems to have parallels with quantum mechanics; quantum mechanics looks suspiciously like a seek, reject and select declarative computation which taps into the resources of expanding parallelism. Contrasting alternatives to my conjectures are that we either have  the anti-theist's meaningless procedural multiverse or the primitive notion that God did indeed simply utter authoritarian magic words and via brute omnipotence was able to speak stuff into existence! The latter seems very unlikely theologically speaking: David Bump, who is a (nice) young earthist Christian I am currently corresponding with, has kindly and respectfully supplied me with a long document of his thoughts on what it means to be a Christian who sees God as creating things via spoken words  "as is" about 6000 years ago. Frankly I doubt it! As I have analysed David's arguments I have found that for me all this leads to huge theological problems and unless I turn to fideism these problems don't look as though they are going to go away! I will in due course be publishing my response to David.

Footnotes:

*1 A bit stream can carry a lot of information in the sense of definition 3.0 because its information value is a product of many probabilities and this may equate to a very small probability and therefore a correspondingly high information content. But the trouble with -ln(p) is that a one bit configuration could be equally as information laden. Another problem with -ln(p) is that once a configuration becomes a "happened event" and is recorded, all its information is lost. This is because probability is a measure of subjective knowledge and therefore once a large configuration becomes known ground, no matter how complex, it loses all its information.... a sure sign that "information" in this subjective sense is easily destroyed and therefore not conserved.

*2 There is also another assumption here (or perhaps it's a confusion rather than an assumption) that probability and randomness are identical concepts - they are not: See my paper on Disorder and Randomness. Dembski uses the principle of indifference in assigning equal probabilities to outcomes for which no prior knowledge exists as to why one outcome should be preferred over the other and hence the outcomes from this subjective point of view have equal probability. This procedure is correct in my opinion; but two outcomes which subjectively speaking have an equal probability are not necessarily equally random; randomness is an objective quality deriving from confrontational disorder. 

Thinknet, Alexa and The Shopping List. Part I

In a "Computerphile" video my son Stuart Reeves explains in high level functional terms the stages involved in "Alexa" parsing, processing and responding to verbal requests. In the video he starts by asking Alexa what may seem a fairly simple question:

Alexa, how do I add something to my shopping list? 

Alexa responds (helpfully, may I say) by regurgitating the data on "Wikihow". But Stuart complains "This is not what I meant! It's a very useful thing if you didn't know how to make a shopping list, but it's not to do with my shopping list!". It seems that poor Alexa didn't twig the subtle differences between these two readings: a) Alexa, how do I add something to my shopping list?  and b) Alexa, how do I add something to a shopping list? Naturally enough Alexa opts for the second generic answer as 'she' has no information on the particularities of the construction of Stuart's shopping list. 

Important differences in meaning may be connoted by a single word, in this case the word "my". Moreover, whether or not this word actually impacts the meaning is subject to probabilities; somebody who wants a generic answer on how to construct a shopping list may have an idiosyncratic way of talking which causes them to slip in the word "my" rather than an "a".  If the question had been put to me I might at first responded as Alexa did and miss the subtlety connoted by "my". However, depending on context "my" could get more stress: For example if I was dealing with a learning difficulties person who was known to need a lot of help this context might set me up to understand that the question is about a proprietary situation and that the generic answer is inadequate.

Stuart's off-screen assistant is invited to put a question to Alexa and he asks this: "Alexa, what is 'Computerphile'?". Alexa responds with an explanation of  "computer file"!  It is clear from this that poor old Alexa often isn't party to essential contextual information which can throw her off course completely. In fact before I saw this video I had never myself heard of "Computerphile" and outside the context of the "Computerphile" videos I would have heard the question, as did Alexa, as a question about "computer file" and responded accordingly. But when one becomes aware that one is looking at a video in the "Computerphile" series this alerts one to the contextualised meaning of "Computerphile" and this shifts the semantic goal posts completely. 

On balance I have to say that I came away from this video by these two computer buffs, who seem to get great pleasure in belittling Alexa, feeling sorry for her! This only goes to show that the human/computer interface has advanced to the extent that it can so pull the wool over one's eyes that one is urged to anthropomorphise a computer, attributing to it consciousness and gender!

Having run down Alexa Stuart then goes on to break down Alexa's functionality into broad-brush schematic stages using the shopping list request as an example.

It is clear from Stuart's block diagram explanation of Alexa's operation that we are dealing with a very complex algorithm with very large data resources available to it. Although some of the general ideas are clear it is apparent that in the programming of Alexa the devil has been very much in the detail. But as we are in broad brush mode we can wave our hands in the direction of functional blocks declaring that "This does this!" and leave the details to be worked out later, preferably by someone else!

***

The precise example Stuart unpacks is this:

                                             Alexa could you tell me what's on my shopping list?

Although it is clear the "Alexa" application makes use of a huge suite of software and huge data resources, I would like to show that there is in fact a very general theme running through the whole package and that this general theme is based on the "pattern recognition" I sketch out in my "Thinknet" project. 

***

An elementary Thinknet "recognition" event occurs when two input tokens results in an intersection. The concept of an intersection is closely related to the concept of the intersection between two overlapping sets. For example, items which have the property of being "liquid" and items which have the property of being "red", overlap  under the heading of "paint".  In  its simplest form a Thinknet intersection, D, (such as 'paint')  results from two input stimuli A & B (such as 'liquid' and 'red') and this is what is meant by a Thinknet "recognition" event. We can symbolise this simple recognition process as follows:

                                                                                           [A B]  → D

1.0

Here inputs A and B  result in the intersecting pattern D. (In principle it is of course possible for an intersection to exist for any number of input stimuli, but for simplicity only two are used here) If we now  input a third stimulating pattern C and combine it with the intersection D we represent this as:

[AB] C → DC

2.0

Since A and B have resulted in the selection of the intersecting pattern D we now have D and C as the stimuli which are candidates for the next level intersection between D and C; if indeed there exists an intersection for D and C. If this intersection exists (let's call it E) then the full sequence of intersections can be represented thus:

[AB]C → [DC] → E

3.0

As an example the stimuli D might be "tin" and together with "red" and "liquid" this might result in the final intersecting pattern E being"tin of red paint".

Expression 3.0 could be equivalently written as:

[[AB]C]  → E

4.0

The square brackets are used to represent a successful intersection operation and in 4.0 the bracketing is applied twice: First to AB and then to the intersection of A and B and the stimuli C. The simple construction in 4.0 is now enough to give us indefinitely nested structures. For example, if we have the general pattern:

ABCDEFG

5.0

Let us assume 5.0 has strong intersections which result in a full recognition event. To exemplify this full recognition event using square bracket notation we may have for example:

[[[ABC]D]E[FG]]

6.0

This nested structure can alternatively be represented as a sequence of intersections: The first layer intersections are:

[ABC] → K and  [FG] → M

7.0

The second layer intersection (only one in fact) is:

[K D] → L

7.1

The third layer intersection combines the output of 7.0 and 7.1 with the residue E in 6.0 as follows:

[LEM] → J

8.0

...this would mean that pattern 5.0, if the intersections are strong enough, implies the likely presence of the pattern J.

The operation of forming intersections is not unlike that of searching for pages on the web where tokens are input and these tokens then define a list of pages which contain these tokens: If the combination of input token has sufficient specification value it will narrow down the search to just a few pages. However, there is a difference between web searches and Thinknet in that in Thinknet the intersection itself is then submitted along with other patterns to the next layer of intersection problem solving resulting in the nested structures we have seen above.

For advanced intersections to occur it is clear that Thinknet must have available a large data resource which effectively embodies the category information required to reach intersections. This large data resource actually takes the form of a weighted association network and this network is way of storing category information. How this network is formed in the first place is another story.

The forgoing examples give us a framework for discussing pattern recognition, Thinknet style. But as we are in broad-brush mode we can ignore the tricky details needed to get something to work really well and instead only talk about the general idea. 

***

If we take a context where we have the input of an unordered collection of stimuli like A, B, C, and D, we may find that at first attempt these fail to form an all embracing intersection and therefore Thinknet fails to recognise the whole context. For example:

[A, D]  [C B]

9.0

Or  expressed in more explicit terms:

[A D] → E  and  [C B] → F 

10.0

Here  [A D] and [C B] have generated intersections E and F. But the intersection formation on E and F has failed to go any further. There are however at least two ways in which this "intellectual blockage" may be circumvented. The high level goal seeking programmed into Thinknet means that it lives to understand and for Thinknet "to understand" means forming intersections. It is good, therefore, if there is more than one way of reaching an understanding.

***

One way to relieve the deadlock expressed by 9.0 may be to "contextualize" the problem by adding another stimulus; let us call this stimulus G. Hence, the input becomes A, B, C, D, G. Adding G may result in a complete solution as represented by the bracketing below:

[[[A, D] G]  [C B]]

11.0

Expressed explicitly in terms of intersection layers:

First layer of intersections:

[A D] → E  and [C B] → F   .....as before - see 10.0 

12.0

Second layer intersection:

[E G] → H 

13.0

...here. the introduction of G means that E and G combine to generate the new intersection H.

The third layer of intersection generation results in full resolution:

[H F] → J

14.0

Thus the result J represents a complete recognition of the inputs A, B, C, D and the contextualising input G.

***

A second way which may relieve the deadlock requires a bit more sophistication. In Edward De Bono's book The Mechanism of Mind, a book on which much of my Thinknet thinking was based, we find a way of getting round this "mental blockage"; I call it a "mental blockage" because on re-submission of problem 9.0 Thinknet as it stands would simply generate the same result. But it wouldn't necessarily generate the same result if the "state" of Thinknet changed slightly every time a problem was submitted. This is achieved by ensuring that when a pattern is activated as a result of an intersection that pattern subsequently needs a greater signal threshold to activate it next time. This means that it may fail to become an intersection on the second attempt. and this may open the way for other patterns with a lower activation threshold to become intersections instead.  

For example, let us suppose that E and F as in 12.0 fail to become intersections on second attempt (or it may take third or even forth attempts) as a result of their thresholds being raised and instead we find a complete intersection solution forms as follows:

[[A D C]  B]

15.0

Or in terms of intersection layers:

First layer:

[A D C] → G 

16.0

Second layer:

 [G B] → H

17.0

The dead lock expressed by 9.0  has been broken by the  threshold increases on E and F, preventing them becoming intersections on later tries; it's a bit like raising land to prevent water pooling and stagnating at the same place by forcing it to pool elsewhere. The important point is that because the path of least resistance has become blocked by increasing thresholds Thinknet has found a new route through to a complete solution. Another way of thinking of the raising of thresholds with use is as a kind of "boredom" factor which encourages Thinknet to move on and try elsewhere. 

***

When I worked on the Thinknet software I got as far as programming nested intersections, but what I didn't do was add threshold changes as a function of use; this would effectively have given Thinknet a feedback loop in so far as outcomes would effect thresholds and the thresholds would effect future outcomes. Adding such a functionality would open up a vista of possible devilish detail: In particular, making the feedback non-linear would introduce the potential for complex chaotic behaviour. If we can think of a Thinknet "thinking session" as a session involving repeated re-submission of an intersection problem, a nonlinear Thinknet would never quite return to its starting state. This would turn Thinknet into a system which searched for intersections by changing its state chaotically: In so far as chaotic evolution is a way of ringing-the-changes (chaotically) Thinknet becomes an intersection search engine. Thus, the more time Thinknet spends "thinking" the more chance that a complete intersection solution pops out of the chaotic regime that thinking entails. But I must add a caution here. A chaotic Thinknet is far from an exhaustive and systematic search engine; its driving energy is chaos, a process which approximates to a random search - it is therefore not an efficient search. But one thing it is: A chaotic Thinknet is "creative" in the sense that it has the potential to come up with unique solutions where the search space is too large, open ended or ill defined to embark on a search via a systematic and exhaustive ringing-of-the-changes. 

I will be using my sketch of Thinknet principles (and how it could work if the details were sufficiently worked out) to discuss if and how Alexa's task, at all stages, can be formulated in terms of Thinknet style pattern recognition. The general nature of this style of pattern recognition encapsulates a simple idea: Thinknet uses a few input clues which, like a Web search engine, narrows down the field by "set fractionating"; that is, where multiple "Venn diagram" sets overlap,the resulting subset may be very narrow. However, where Thinknet differs from this kind of simple search is in problem re-submission and its non-linear "thinking" process. But I must concede that this underlying chaotic searching may not be suitable for an Alexa type application because there is probably a demand for predictable and controllable output results when it comes to slave systems like Alexa. In contrast a fully fledged Thinknet system has strong proprietary goal seeking behaviour and is orientated towards seeking idiosyncratic, creative and unpredictable intersections rather than finding servile answers through an exhaustive systematic search.  In short, it's too human to be of use!