Platt Perspective on Business and Technology

Reconsidering Information Systems Infrastructure 16

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on July 23, 2020

This is the 16th posting to a series that I am developing, with a goal of analyzing and discussing how artificial intelligence and the emergence of artificial intelligent agents will transform the electronic and online-enabled information management systems that we have and use. See Ubiquitous Computing and Communications – everywhere all the time 2 and its Page 3 continuation, postings 374 and loosely following for Parts 1-15. And also see two benchmark postings that I initially wrote just over six years apart but that together provided much of the specific impetus for my writing this series: Assumption 6 – The fallacy of the Singularity and the Fallacy of Simple Linear Progression – finding a middle ground and a late 2017 follow-up to that posting.

I have been discussing the first of a set of four topics points since Part 13 that I have been offering as a tool kit, or rather a set of indicators as to where a set of tools might be found. And these tools would be used for carrying out at least part of a development process for enabling what would ideally become true artificial general intelligence.

Those topics points are:

• The promotion of ontological development that is both driven by and shaped by self-learning behavior (as cited in Part 13 in terms of antagonistically positioned subsystems, and similar/alternative paradigmatic approaches),
• Scope and range of new data input that might come from the environment in general but that might also come from other intelligent agents (which might mean simple tool agents that carry out single fully specified tasks, gray area agents that carry out partly specified tasks, or actual general intelligence agents: artificial or human, or some combination of all of these source options.)
• How people or other input providing agents who would work with and make use of these systems, simplifying or adding complexity to the contexts that those acting agents would have to perform in, shift tasks and goals actually required of them either more towards the simple and fully specified or more towards the complex and open-ended.
• And I add here, the issues of how an open ended task to be worked upon and goal to be achieved for it, would be initially outlined and presented. Think in terms of the rules of the road antagonist in my two subsystem self-driving example of Part 12 here, where a great deal of any success that might be achieved in addressing any overtly open-ended systems goal will almost certainly depend on where a self-learning agent would begin addressing it from.

Note: my focus here, insofar as I will bring that to bear on general intelligence embryogenesis, is on startings for that per se. The above repeated fourth tool set point begins, for what it addresses, by assuming that this is possible and perhaps even inevitable and addresses the issues of optimization so as to enable greater long-term developmental potential and its realization, from such a starting point.

I focused on the first of those points in Part 15, and primarily on the bootstrap problem of initially developing what could be considered an at least embryonic artificial general intelligence there, and with a supporting discussion of what ontological development might encompass, at least in general terms, in this overall context. And that led me to a set of Point 1-related issues that I raised there but that I held off on delving into:

• The issues of understanding, and deeply enough, the precise nature and requirements of open-ended tasks (n.b. tasks that would call for general intelligence to resolve) per se,
• The roles of “random walk” and of “random mutational”, and of “directed change” and how they might be applied and used in combinations, and regardless of what starting points are actually started from, ontologically, in developing such an intelligence capacity.

I went on to say at the end of Part 15 that I would in fact address these issues here, and I added, in anticipation of that discussion to come, that doing so will lead me directly into a more detailed consideration of the second tool packet as repeated above. And I will in fact at least begin to cover that ground here. But before doing so and in preparation for those lines of discussion to come, I want to step back and reconsider, and in somewhat further detail than I have offered here before, exactly what artificial or any other form of general intelligence might entail: what general intelligence as a general concept might mean and regardless of what form contains it.

I am going to address this from an artificial construct perspective here, as any attempt to analyze and discuss general intelligence from an anthropocentric perspective, and from the perspective of “natural” intelligence as humans display it, would be essentially guaranteed to bring with it a vast clutter of extraneous assumptions, presumptions and beliefs that revolve around our all too human expectation of our somehow being a central pivot point of universal creation. Plato’s and Aristotle’s scala naturae lives on, and with humanity presumed to be at the very top of it and with that held as an unassailable truth and for many.

• As a here-pertinent aside, the only way that we may ever come to understand what general intelligence really is: what it really means, is if we can somehow find opportunity to see it in a non-human form where we can in fact view and come to terms with it without that baggage.
• And this ultimately, may prove to be the only way that we can ever really understand ourselves as humans too.

But setting aside that more philosophical speculation to address the issues at hand here, I turn to consider general intelligence from the perspective of an artificial general intelligence to be, and from the perspective of a perhaps embryonic beginning to one as cited in passing in Part 15. And my goal here is one of at least hopefully shedding light on what might be the source of that first bootstrap lifting spark into intelligent being.

I am going to pursue that goal from two perspectives: the first of which is a broadly stated and considered general framework of understanding, that I will offer as a set of interconnected speculations. Then after offering that, I will delve at least selectively and in broad brushstroke detail into some of the possible specifics there. And I begin the first of those two lines of discussion by making note of an at least ostensibly very different type of series that I have been offering here, that has nevertheless prompted my thought on this area of discourse too: Some Thoughts Concerning a General Theory of Business as can be found at the directory Reexamining the Fundamentals for its Section VI and at its Page 2 continuation for its Section IX. And I specifically cite the opening progression of postings as offered in that Section IX here for their discussion of closed axiomatic systems that are entirely self-contained, and open axiomatic systems that are developed and elaborated upon, on the basis of outside-sourced information too.

I would argue, with the lines of reasoning offered there as impetus for this, that:

• The simple essentially tool-only specialized artificial intelligence agents that I write of here are limited to that level and type of development because they are limited to simple deductive reasoning that is tightly constrained by the rigid axiomatic limitations of the pre-specified algorithms that they execute upon.
• At least potential artificial general intelligences will only be able to advance to reach that goal to the extent that they can break away from such limitations. This of necessity means they’re at least attempting to successfully execute upon open ended tasks that by their very nature would fit an open axiomatic model, as discussed in my business theory series, where new and even disruptively new and different might arise and have to be accommodated too. And this means allowing for and even requiring inductive reasoning as well as its more constrained deductive counterpart.
• Gray area tasks, or rather the gray area agents that would carry them out might simply remain over time in this in-between state, or they might come with time to align more with a simple specialized artificial intelligence agent paradigm, or a more and more genuinely artificial general intelligence one. Ultimately, that will likely depend on what they do and on whether they canalize into a strictly deductive reasoning form, or expand out into becoming a widely inductive reasoning-capable one.

Even if this is necessary as a developmental and enablement requirement for the formation of a true artificial general intelligence I expect that it can on no way be considered to be sufficient. So I continue from that presumption to add a second one, and it involves how deeply interconnected the information processing is and can be, in an entity.

• Tightly compartmentalized, context and reach limited information processing that separates how tasks and subtasks are performed and with little if any cross-talk between these process flows beyond the sharing of output and the providing of input can make for easier, more efficient coding, as object-oriented programming proves.
• But when this means limited at most, and even stringently prevented cross-process and cross-systems learning with rigid gatekeeper barriers limiting the range of access to information held and with that established on an a priori basis, that would almost by definition limit or prevent anything like deep learning, or widely inclusive and involving ontological self-development as might be possible from accessible use of the full range of experience that is held within such a system.
• This means a trade-off between simpler and perhaps faster and more efficient code as a short-term and here-and-now imperative, and flexible capacity to develop and improve longer-term.
• And when I cite an adaptive peak model representation of developmental potential in this type of context, as I have recurringly done so in a concurrently running series: Moore’s Law, Software Design Lock-In, and the Constraints Faced When Evolving Artificial Intelligence (as can be found at Reexamining the Fundamentals 2 as its Section VIII) I would argue that constraints of this type may very well be among the most significant in limiting the maximum potential that an ontological development can reach in achieving effective general intelligence, or just effective functionality per se for that matter.

I am going to continue this narrative in a next series installment where I will add a third such generally stated puzzle piece to this set. And then, as promised above, I will proceed to address the second basic perspective that I made note of above here, and at least a few more-specific points of consideration. Beyond that I will more explicitly address two points that I said that I would delve into here in this posting, in Part 15 but that I have only approached dealing with up to this point in this overall narrative:

• The issues of understanding, and deeply enough, the precise nature and requirements of open-ended tasks (n.b. tasks that would call for general intelligence to resolve) per se,
• The roles of “random walk” and of “random mutational”, and of “directed change” and how they might be applied and used in combinations, and regardless of what starting points are actually started from, ontologically, in developing such an intelligence capacity.

And I will finish my discussion of the second of four main topic points that I repeated at the top of this posting, with that. And then I will move on to address the above-repeated Points 3 and 4 from the tools list that I have been discussing here:

• How people or other input providing agents who would work with and make use of these systems, simplifying or adding complexity to the contexts that those acting agents would have to perform in, shift tasks and goals actually required of them either more towards the simple and fully specified or more towards the complex and open-ended.
• And the issues of how an open ended task to be worked upon and goal to be achieved for it, would be initially outlined and presented, and how the capability of an agent to develop will depend on where it begins that from.

Meanwhile, you can find this and related postings and series at Ubiquitous Computing and Communications – everywhere all the time and its Page 2 and Page 3 continuations. And you can also find a link to this posting, appended to the end of Section I of Reexamining the Fundamentals as a supplemental entry there.

Moore’s law, software design lock-in, and the constraints faced when evolving artificial intelligence 13

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on July 17, 2020

This is my 13th posting to a short series on the growth potential and constraints inherent in innovation, as realized as a practical matter (see Reexamining the Fundamentals 2, Section VIII for Parts 1-11.) And this is also my tenth posting to this series, to explicitly discuss emerging and still forming artificial intelligence technologies as they are and will be impacted upon by software lock-in and its imperatives, and by shared but more arbitrarily determined constraints such as Moore’s law (see Parts 4-12.)

I have for the most part, focused here in this series on electronic computers and on the underlying technologies and their implementations that make them possible. And that has meant my at least briefly addressing the historical sweep of what has been possible and of what has been accomplished in that, and from the earliest vacuum tube computers on to today’s more advanced next generation integrated circuit-driven devices. And I have addressed issues of design and development lock-in there, and in both a software and a hardware context, and Moore’s law as an integrated circuit oriented, seemingly self-fulfilling prognosticator of advancement. Then I shifted directions in Part 12 and began at least laying a foundation for discussing quantum computing as a source of next-step advancement in this still actively advancing history.

I ended Part 12 by offering this to-address list of issues and topics that will arise in this still embryonically forming new computer systems context, adding that I will address them from both a technological and a business perspective:

• A reconsideration of risk and of benefits, and both as they would be viewed from a short-term and a longer-term perspective,
• A reconsideration of how those issues would be addressed in a design and manufacturing context,
• The question of novelty and the challenges it creates when seeking to discern best possible technological development paths forward, where they can be clear when simple cosmetic changes are under consideration, but opaque and uncertain when more profound changes are involved,
• And lock-in and certainly as that becomes both more likely to arise, and more likely to become more limiting as novelty and the unexpected arise.

I begin at least preparing for a discussion of those issues here by offering a more fundamental science and technology perspective on what is involved in this new and emerging next technology advance context. And to put that in a need and demand based perspective, I begin addressing those issues by sharing links to two relevant news-oriented pieces that relate to the current state of development of quantum computing, and a link to a more detail clarifying background article that relates to them:

Google Claims a Quantum Breakthrough That Could Change Computing,
Quantum Supremacy Using a Programmable Superconducting Processor and this piece on
Quantum Supremacy per se, where that represents an at least test-case proof of the ability of a quantum computing device to solve a problem that classical computers could not solve as a practical matter, for how long its resolution would take with them (e.g. seconds to minutes with even just our current early stage quantum computer capabilities, as opposed to thousands to tens of thousands of years with the fastest current supercomputers that we have now.)

Think of the above news story (the first reference there), the research piece that directly relates to that same event (the second reference there), and the detail clarifying online encyclopedia article (the third reference offered there) as representing proof that quantum computing has become a viable path forward for the development of dramatically more powerful new computer systems than have ever been possible.

That noted, quantum computing really is still just in its early embryonic stage of development so we have only seen a faint preliminary glimmer of what will become possible from this advance. And given the disruptive novelty of both this technology and its underlying science, and certainly as that would be applied in anything like this type of context, we cannot even begin to imagine yet, how these devices will develop.

• Who, looking at an early generation vacuum tube computer could have imagined anything like a modern cutting edge, as of this writing, solid state physics based supercomputer, of the type so outpaced now in the above cited benchmark test?

And with that noted for orientation purposes if for no other reason, I turn to consider quantum computing per se, starting with some here-relevant quantum physics. And that means starting with at least a few of the issues that arise in the context of quantum indeterminacy as that plays a central role in all quantum computer operations.

Let’s consider the well known but perhaps less understood metaphorical example of Schrödinger’s 50% of the time more than just maligned cat.

• A cat is put in a box with a radioisotope sample that has a 50% chance of producing one radioactive decay event in a given period of time. And if that happens, that will trigger a mechanism that will kill the cat. Then after exactly that 50% chance, interval of time has passed, the box is opened and the cat is removed from it, and with a 50% chance of it still being alive and a 50% chance of it now being dead.
• According to the principles of quantum indeterminacy, the condition of that cat is directly, empirically known as a fixed matter going into this “experiment.” The probability of that condition pertaining then and there as a valid empirical truth is 100%. The cat starts out alive. And when the box is opened at the end of this wait, its condition is once again directly, empirically known too, whatever it is. And once again, the probability of that condition, whether alive or dead can be empirically set at 100%, as a known, valid and validated truth. But the condition of that cat is both unknowable and undetermined while it is locked inside that box. And that last detail: the indeterminacy of that cat’s condition while in the box, is the crucially important detail here, and both for how this narrative applies to quantum physics per se and for how it applies in this specific context. And that most-crucial detail is also where this is least understandable.

Is the cat close to 100% alive and just a fraction over 0% dead immediately after the box is closed, with those condition specifying percentages “equilibrating” to 50% and 50% just as the box is opened? Mathematically at least, that is what the basic rules of quantum indeterminacy would specify. But that description just serves as an attempt to force fit classical physics and related scientific expectations and terminology into a context in which they do not actually apply. Those terms: dead and alive apply in a pre- and a post-experiment context where the condition probabilities are empirically resolved at 100%. They do not in fact hold meaning while the cat is in a here-assumed quantum indeterminate state.

Now let’s take this same pattern of reasoning and apply it to a circuit element, or rather its equivalent, that would manipulate a qubit, or quantum bit of data in carrying out a calculation in a quantum computer. Going into that calculation, that piece of data might have a known or at least knowable precise 0 or 1 value. And the same applies when this calculation step is completed. Think of those states as representing the classically observable states that that cat is in, immediately before and after its “experiment.” But the value of that qubit can best be thought of as a probabilistically shaped and determined smear of possible values that range from 0 to 1 while that calculation step is actually taking place.

What is the information carrying capacity of a qubit? The basic answer is one bit, though it is possible to pack two bits into a single qubit using a process called superdense coding. But that, crucially importantly only represents the information capacity inherent to a qubit outside of that in-between indeterminacy period when that calculation step is actually being carried out. Then, during that period of time, the actual functional information carrying capacity of a qubit becomes vastly larger from the vastly larger range of possible values that it in effect simultaneously holds. And that is where and when all quantum computer calculations take place and that is where those devices gain their expansively increased computational power – where that is a function of the speed at which a calculation takes place and a function of the actual volume of information processed during those involved time intervals.

I have to admit to having been in a mental state that is somewhat equivalent to a quantum indeterminacy here, as I have been thinking through how to proceed with this posting Part of me has wanted to in effect dive down an equivalent of Lewis Carroll’s rabbit hole and into some of the details buried under this largely – if still briefly stated metaphorical explanation. Part of me has wanted to keep this simple and direct, and even if that means leaving out areas of discourse that I at least find endlessly fascinating, and with discussions of Dirac notation and linear algebra and of Hilbert space and vectors in it – sorry, but I mention this for a reason here and even when I have chosen to follow the second of those two paths forward from here.

• Quantum computing and the basic theory behind it, and the emerging practice of it too, involve very long and very steep learning curves, as they contain within them a veritable flood of the unfamiliar to most, and of the disruptively New to all.
• And that unavoidable, 100% validatable truth will of necessity shape how the four to-address bullet points that I started this posting with and that I will discuss in detail, will be generally understood, let alone acted upon.

I repeat those topics points here, noting that my goal for this posting was to at least begin to address the technological side to quantum computing, in order to at least start to set a framework for thinking about them as they would arise in real world contexts:

• Where I will offer a reconsideration of risk and of benefits, and both as they would be viewed from a short-term and a longer-term perspective in this new and emerging context,
• A reconsideration of how those considerations would be addressed in a design and manufacturing context,
• The question of novelty and the challenges it creates when seeking to discern best possible technological development paths forward, where they can be clear when simple cosmetic changes are under consideration, but opaque and uncertain when more profound changes are involved,
• And lock-in and certainly as that becomes both more likely to arise, and more likely to become more limiting as novelty and the unexpected arise.

I will at least begin to more explicitly address these points and their issues starting in a next series installment. Meanwhile, here are two excellent reference works for anyone who would like to delve into the details, or at least a lot more of them than I have offered here (where I only skirted the edge of that rabbit hole in this posting):

• Bernhardt, C. (2019) Quantum Computing for Everyone. MIT Press.
• Hidary, J.D. (2019) Quantum Computing: an applied approach. Springer.

The author of the first of those book references claims that they should be understandable to anyone who is willing to put in some work on this, who is comfortable with high school mathematics. It makes use of linear algebra and a few other topic areas that are not usually included there, but it does not assume any prior knowledge of them. The second of those books uses more advanced mathematics and it presumes prior experience in them of its readers. But it goes into corresponding greater depth of coverage of this complex and fascinating topic too.

• Both books delve into issues such as quantum entanglement that are crucially important to quantum computing and to making it possible, so I do offer these references for a reason. And crucially importantly to this discussion, and as a perhaps teaser to prompt further reading, it is quantum entanglement that in effect connects a quantum computer together, so that the calculations carried out qubit- by-qubit in it can be connected together in carrying out larger and more complex calculations than any single information processing element (as touched upon above) could manager or encompass on its own, and no matter how it is recurringly used.
• And both books at least begin to address the issues of quantum computational algorithms and that is where the there-undefined mathematical terms that I cited in passing above come into this.
• All quantum computational algorithms function by shaping the probability distributions of possible values that can arise between that 0 and that 1 in an indeterminate state. And that is described in terms of mathematical machinery such as linear algebra and Hilbert space theory. And as all of the data that would be processed in or created by the execution of those algorithms, tends to be expressed in terms of Dirac notation, or bra–ket notation as it is also called, that actually becomes quite important here too.

Setting aside this set of references for the moment, and my at least attempted argument in favor of looking deeper in the technology side of quantum computing through them, or through similar resources, I add that you can find this posting and this series and related material at Ubiquitous Computing and Communications – everywhere all the time 3. And also see Page 1 and Page 2 of that directory. And I also include this in my Reexamining the Fundamentals 2 directory as topics Section VIII. And you can also find further related material in its Page 1 directory listing too.

Some thoughts concerning a general theory of business 35: a second round discussion of general theories as such, 10

Posted in blogs and marketing, reexamining the fundamentals by Timothy Platt on July 14, 2020

This is my 35th installment to a series on general theories of business, and on what general theory means as a matter of underlying principle and in this specific context (see Reexamining the Fundamentals directory, Section VI for Parts 1-25 and its Page 2 continuation, Section IX for Parts 26-34.)

I have been successively discussing three closely interrelated topics points and their ramifications in this series since Part 31, which I repeat here for smoother continuity of narrative:

1. Provide a brief discussion of the generality of axioms, and of how a theory elaboration can lead to reconsideration of the axioms underlying it, where that can lead to accepted axioms already in place, now being reconsidered as special cases of more general principles,
2. Adding, subtracting and modifying such axioms in general, and
3. The whole issue of consistency, and even just within constrained sets of axioms. And I will frame that discussion at least in part, in terms of outside data, insofar as that makes these theories open – and with a reconsideration of what evolutionary and disruptive mean there.

And as part of that still ongoing line of discussion, I have been addressing its Point 3 since Part 34. And to bring this preliminary orienting note up to date as a start to this posting and its discussion, I concluded that installment by stating that I would address the following more derivative points here, doing so in light of what I have been offering in response to Points 1 and 2 and my start to addressing the above Point 3 as well:

• Reconsider the issues of consistency and completeness, in light of the impact created by disruptive and seemingly disruptive examples, as for example raised in Part 33 (with its Higgs field example) and Part 34 (with its three physics sourced examples.)
• This will include, as noted in Part 34, my addressing the challenge of direct and indirect testing of potentially validatable assertions that would fit into an open general theory, and what those terms actually mean in practice: direct, indirect and validatable.
• Yes, that will mean reconsidering my Higgs boson/Higgs field physics example from Part 33 again too. And I will also mean my further considering what evolutionary and disruptive mean: terms and concepts that I have made at least case-by-case level use of throughout this blog. And I will also, of necessity connect back in that to my discussion of compendium theories as discussed in this series in its Parts 2-8.

And with that offered as orientation for what is to follow, I begin with the first of those three to-address topics points and with a bipartite distinction that I made in Part 34 as to how issues such as theory consistency could be considered: as overall and (at least ideally) all-inclusive body of theory spanning determinations, or as case-by-case determinations.

• For a closed axiomatic body of theory such as Euclidean geometry, any such effort, as conceived on a more strictly case-by-case, derivative theorem basis would be grounded entirely upon deductive analytical reasoning, and with a goal of determining adherence to and consistency with the set of underlying axioms in place.
• For an open axiomatic body of theory that begins with such an axiomatic framework but that also allows in (and even requires) outside evidentiary support such as experimental or other empirically sourced data, inductive reasoning becomes essential too – and the starting axioms that might be in place as this evidence is brought in, can be as subject to challenge as any newly proposed derivative theorem or lemma would be.

For purposes of this phase of this overall narrative, I simply point to closed bodies of theory for purposes of what might be considered simplest case comparison, as overall theories of physics, or of business, or of any other such empirically grounded reality are by their very nature, open. And that is where change and disruptive change enters this, where they might arise as a matter of planned intention and as conceived innovation, or without such a priori intention (even if inevitably so and as a result of a “law of unintended consequences.”)

What do simple change and disruptive change mean in this context? I would argue that the most salient general point of distinction between them lies in where they shed light on and differ from the routinely expected and predicted.

• Simple change does not challenge underlying theory, even if it can, or at least should be able to provoke the bringing of theory implementation into focus and even into question. For a physics example of this, I point to the so called paradox of radiation of charged particles in a gravitational field, as discussed in Part 34 (which at the very least had simple change elements to it.) For a business example, I would cite essentially any more-minor business process adjustment, as for example might become necessary if a manufacturer has to change where or how they source some essential supply that would go onto their own products and their production. This obviously does not in any way challenge basic underlying business theory. It most probably would not challenge that enterprise’s basic business model in place or its strategic or operational planning either – unless that is, this now needed change and any realistic accommodations to it would of necessity have disruptive consequences.
• But genuinely disruptive change, and certainly as an extreme for that, can challenge everything. In a physics context consider the accumulation of repeatedly verified and validated experimental findings that were both incompatible with classical physics and even incomprehensible to it, that led to the development of quantum physics. For a business and I add an economics context, consider the advent of electronic computers, and certainly as they transitioned from being rare “not for here” curiosities into their becoming ubiquitous business necessities – and for how that qualitatively transformed the role of and even the nature of information in a business context. Read something of business theory from right before World War II, and from today and look at the difference and at how much of that difference is explicable in information collection, organization, and usage terms.

And this leads me to two observational points, the first of which might be more obvious than the second and particularly for anyone who has been following this blog:

• The truly disruptive can raise red flags in how it sheds questioning spotlights on what was established theory. But it can be better at that than it is at presenting alternatives to those now-old and outdated understandings. Once again, consider quantum physics, or at last its early states where it was agreed to that those new and disruptive experimental findings were valid, but where there was no consensus as to what they collectively meant – no consensus or anything like it as to how they should be interpreted or understood, or of how a new generally stated axiomatically grounded weltanschauung could best be developed around them.
• And this old foundation-breaking without a corresponding at-least quickly following new foundation-making, can only lead to the arising of a (hopefully productive) compendium theory next step that would offer localized explanatory and predictive theory for specific types of disruptive outside data findings, that would at least hopefully serve as puzzle pieces for developing a new more general axiomatically grounded foundation around.

Note that while I have been discussing the above Point 3 here, I have also been explicitly discussing Points 1 and 2 here as well; they cannot in practice be separated from each other in any meaningful discussion of any of them.

And with this, I return again to my Higgs field example of Part 33. What axiomatic assumptions does the presumption of this mechanism for conveying the Higgs force carry with it, aside from the perhaps more obvious ones of its being selected and framed so as to be as consistent as possible with already established particle physics and related theory, and (hopefully) with a mainstream understanding of general relativity too? I have already cited one other axiomatic assumption in passing, with the notion of this serving as a reprise of the old theory of an aether for conveying the electromagnetic force. (OK, photons as force conveying boson particles do not need or use that type of mechanism, by maybe Higgs bosons do …. See this piece on the Michelson–Morley Experiment.)

I am going to continue this discussion in a next series installment where I will continue addressing the issues raised in the first three bullet points of this posting, with a continued focus that is (as here) largely aimed at the above-repeated topics Point 3. And I will continue addressing that, with Point 1 and 2 diversions, in terms of the three derivative topics points that I offered here. And with business theory and business practice in mind as I prepare to do so, I finish this posting by posing a few basic questions, the first of which is a reprise from the first general theory discussion of this series of its Parts 2-8:

• If a general theory, or rather an attempt to develop and organize one remains in effect, in perpetual turmoil from the consistent ongoing arising of the disruptively new and unexpected and from all directions,
• Does that effectively preclude any real emergence from a compendium model overall understanding there with the development of an overarching axiomatically grounded general theory?
• And what does this mean as far as overall body of theory consistency and completeness are concerned?

Addressing those questions will bring me directly into a discussion of the second of those derivative bullet-pointed topics, and a more detailed consideration of:

• Direct and indirect testing of potentially validatable assertions that would fit into an open general theory, and what those terms actually mean in practice: direct, indirect and validatable.

Meanwhile, you can find this and related material about what I am attempting to do here at About this Blog and at Blogs and Marketing. And I include this series in my Reexamining the Fundamentals directory and its Page 2 continuation, as topics Sections VI and IX there.

Reconsidering Information Systems Infrastructure 15

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on May 21, 2020

This is the 15th posting to a series that I am developing, with a goal of analyzing and discussing how artificial intelligence and the emergence of artificial intelligent agents will transform the electronic and online-enabled information management systems that we have and use. See Ubiquitous Computing and Communications – everywhere all the time 2 and its Page 3 continuation, postings 374 and loosely following for Parts 1-14. And also see two benchmark postings that I initially wrote just over six years apart but that together provided much of the specific impetus for my writing this series: Assumption 6 – The fallacy of the Singularity and the Fallacy of Simple Linear Progression – finding a middle ground and a late 2017 follow-up to that posting.

I initially offered a set of four tool kit elements in Part 13 that might arguably hold value in both thinking about and developing more open artificial intelligence agents: agents that hold capability for carrying out open ended tasks, as I have used that term in this series. And I began to elaborate upon those at-least conceptual tools in Part 14. In the process of doing so I raised and began discussing something of the distinctions that can be drawn between deterministic and stochastic processes and systems as they would arise in this context, and with a focus on deterministic and stochastic understandings of the problems that they would seek to address, and either resolve outright or manage on an ongoing basis. (Note: the open ended task that I have been repeatedly referring back to here as a source of working examples: completely open and free ranging natural speech, can in most cases and circumstances be seen as an ongoing task with at least some subsequent conversation more usually at least a possibility. Us humans generally identify ourselves as Homo sapiens – the thinking hominid, and occasionally as Homo faber – the tool making hominid. But for purposes of this narrative, Homo loquace – the talkative hominid might be a better fit.)

That perhaps exculpatory note offered and now set aside, I will continue discussing open ended tasks using natural speech as its working example. And I will discuss agents that can successfully carry out that task as having proven themselves to have general intelligence.

The four “tool packets” under consideration here are:

• Ontological development that is both driven by and shaped by self-learning behavior ( as cited in Part 13 in terms of antagonistically positioned subsystems, and similar/alternative paradigmatic approaches),
• Scope and range of new data input that might come from the environment in general but that might also come from other intelligent agents (which might mean simple tool agents that carry out single fully specified tasks, gray area agents that carry out partly specified tasks, or actual general intelligence agents: artificial or human, or some combination of all of these source options.)
• How people or other input providing agents who would work with and make use of these systems, simplifying or adding complexity to the contexts that those acting agents would have to perform in, shifting tasks and goals actually required of them either more towards the simple and fully specified, or more towards the complex and open-ended.
• And I add here, the issues of how an open ended task to be worked upon and goal to be achieved for it, would be initially outlined and presented. Think in terms of the rules of the road antagonist in my two subsystem self-driving example of Part 12 here, where a great deal of any success that might be achieved in addressing any overtly open-ended systems goal will almost certainly depend on where a self-learning agent would begin addressing it from.

I addressed a few of the basic issues inherent to deterministic and stochastic systems in general, in Part 14. My goal here is to at least begin to offer a framework for more specifically thinking about the above tools (approaches to developing tools?) in more focused deterministic and stochastic terms, taking those issues at least somewhat out of the abstract in the process. So for example, I wrote of deterministic systems and the problems that they could best manage, as being more limited in scope; they can in fact often be thought of and realistically so, as simple systems challenges. And I wrote of the increased likelihood of combinatorial explosions in the range and diversity of issue and circumstance combinations that would have to be addressed, and in real-time, as problems faced become more stochastic in nature.

• Quantitative scale increases there, and both for what has to be resolved and for what it would take to achieve that, can quickly become very overtly qualitative changes and change requirements too, in this type of context.

And with that noted, let’s consider the first of those four tool packets and its issues. And I begin so by repeating the key wording that I offered in an anticipatory sentence that I added to the end of Part 14 in anticipation of this posting to come:

• I am going to at least begin it (n.b. that line of discussion) with a focus on what might be considered artificial intelligence’s bootstrap problem: an at least general conceptual recapitulation of a basic problem that had its origins in the first software-programmable electronic computers.

The first bullet pointed tool suggesting, if not specifying element to discuss here, as repeated above, is all about ontological development and with distinct positive advancement presumed as an outcome from it. This of necessity presumes a whole range of specific features and capabilities:

• First of all, you cannot have ontological development: success and improvement creating or not, without a starting point.
• And it has to be a starting point that has a capacity for self-testing and review built into it.
• It has to have a capacity to make specific change capabilities built into it where an agent can modify itself (at least through development of a test case parallel systems capability),
• Test and evaluate the results of using that new test-capability functionality in comparison to its starting stage, current use version for what it is to do and for how well it does that,
• And then further modify or use as-is, or delete that test possibility as a result of outcomes obtained from running it – as benchmarked against the best current standards in place and in use.

Two points of detail come immediately to mind as I offer this briefly stated ontological development framework. First, the wider the range of features that have to be available for this type of test-case development and validation, the more elaborate and sophisticated the machinery (information processing code definitely included there), that would be needed to carry this out. And the more complex that has to be able to be, the more of an imperative it becomes that ontological development capability itself would have to be ontologically improvable too, and expandable for its range of such activity as much as anything else.

And that brings me directly to that bootstrapping problem. Where do you start from, and I have to add how do you start for all of this? The classic bootstrapping problem for software-programmable electronic computers: the question of how you might be able to “lift yourself up by your bootstraps” in starting an early generation computer of that type was simple by comparison. When you first turned one of them on it had no programming in it. And that, among other things meant it did not even have a capability for accepting programming code as a starting point that it could operationally build from. It had no already-established input or output (I/O) capability for accepting programming code or anything else in the way of information, or for providing feedback to a programmer so they would know the status of their boot-up effort.

So it was necessary to in effect hand-feed at least some basic starter code, in machine language form, into it. That of necessity included basic input and output functionality so it could read and incorporate in, the code that would then be entered in by more automated electronic means (such as punch tape, or later via punch cards or from magnetic tape storage media.) But first of all and before you could do any of that, you had to teach it how to accept at least some input data at all, and from at least one of these types of sources.

That was simple: basic early computer I/O code was straightforward and clear-cut and both for what had to be included in it and for how that would be coded and then incorporated in. It is at the heart of the challenges posed by open-ended tasks that we do not start out knowing even what types of coding capabilities we would need to start out with, and certainly with any specificity, in setting an effective starting point at a software level for either addressing those tasks, or for building an “artificial general intelligence embryo.”

• To clarify that, we have a basic, general, high level categorical understanding of the building block types that such a system would have to start with.
• But we do not know enough to really effectively operationalize that understanding in working detail and certainly for anything like an actual general artificial intelligence agent – yet.
• We know how to do this type of thing and even routinely now, for single task specialized artificial intelligence agents. They are tools, if more complex ones, and the tasks they would carry out are subject to at least relatively straightforward algorithmic specification. Gray area agents and their tasks become more and more problematical and certainly as their tasks shift further away from the more fully determined of that. Actually addressing open ended tasks as discussed here that would call for general intelligence to resolve, will call for disruptively new and novel, game changing developments and probably at both the software and hardware levels.

When the challenges that those agents would have to face, become more complex and more stochastic in nature as an at least indirectly causally connected consequence of their becoming more and more open ended, and as we face increased uncertainty as to what resolving their tasks would entail, the more difficult it becomes to know where to start: the more uncertain any postulated “I/O code” starting point for that would have to be.

And this brings me to two deeply interconnected issues:

• The issues of understanding, and deeply enough, the precise nature and requirements of open-ended tasks per se, and
• The roles of “random walk” and of “random mutational”, and of “directed change” and how they might be applied and used in combinations, and regardless of what starting points are actually started from, ontologically.

I am going to at least begin to address those issues in the next installment of this series. And in anticipation of that discussion to come, that will lead me directly into a more detailed consideration of the second tool packet as repeated above. My goal here is to complete an at least initial discussion of the full set of four tool packets as offered above. And after completing that I will finally explicitly turn to consider the avowed overall goal of this series as repeated in the first paragraph of all of its installments up to here: “analyzing and discussing how artificial intelligence and the emergence of artificial intelligent agents will transform the electronic and online-enabled information management systems that we have and use.”

Meanwhile, you can find this and related postings and series at Ubiquitous Computing and Communications – everywhere all the time and its Page 2 and Page 3 continuations. And you can also find a link to this posting, appended to the end of Section I of Reexamining the Fundamentals as a supplemental entry there.

Moore’s law, software design lock-in, and the constraints faced when evolving artificial intelligence 12

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on May 12, 2020

This is my 12th posting to a short series on the growth potential and constraints inherent in innovation, as realized as a practical matter (see Reexamining the Fundamentals 2, Section VIII for Parts 1-11.) And this is also my ninth posting to this series, to explicitly discuss emerging and still forming artificial intelligence technologies as they are and will be impacted upon by software lock-in and its imperatives, and by shared but more arbitrarily determined constraints such as Moore’s law (see Parts 4-11.)

The above paragraph and its basic message have applied as-is, to essentially all of my earlier postings here. But they require a new type of clarifying expansion with this series installment and for what is to follow it. That is because I have been writing this series, at least up to here, essentially entirely in terms of a single, overarching physical sciences based hardware technology and its use: electronic computers that have been build using integrated circuit chips, and software designed to run on that type of platform. And to focus on the software side of that, I have been writing this essentially entirely in terms of programming languages that can be implemented through computers that would parse and process binary data and ultimately just binary data, and as a core defining feature of that overall paradigm. And regardless of the nature or the nuanced complexity of the human-usable data that we enter into our increasingly powerful integrated circuit-based computers, and that we see and hear coming from them in return, all of the information processing taking place in those devices, takes place at a binary data level. All of it takes place as the manipulation of strings of ones and zeros.

I have in that context, also at least briefly discussed the conceptualized but never actually realized mechanical analytical engines of Charles Babbage’s visions: his dreamed of general purpose computers. And I have also made at least brief note of the intended software approaches that his assistant, Ada Lovelace envisioned and that she began to develop code for, for that context. Any transition from their understanding of computers and of automated computation per se, to that of an electronic context would have represented profoundly transformational change. But that never really happened, because mechanical computers that went beyond the simple single function in capability, as for example implemented as steam engine regulators, were never actually built.

The changeover from vacuum tube computers and from transiently-hardwired programming as a means of creating a more general computer capability in them, to the development of software-level programmable transistor and integrated circuit-based computers was transformationally profound. Both types of technology were built, and used and advanced and refined in the process. The solid state physics-based technologies of integrated circuit computers are still actively advancing, and that is where Moore’s law enters this narrative; the demands of maintaining it and its predicted pace of development have both advanced computer technology itself, and compelled fundamental technological advancement too, in order to make that possible. And that technology base has advanced in directions, and to degrees that would have been unimaginable to the people who developing the earliest transistor or integrated circuit-based computers, let alone those who developed and built those still earlier vacuum tube computers. None of them would have or even could have imagined the types of quantum physics-level considerations that would enter into essentially every next generation circuit design issue as pertain now, and at any of those earlier computer systems development stages.

But the basic principles of both those early vacuum tube, and of transistor component computers as well, were already both fundamentally understood and widely so. And they were commonly made us of in what were already well known products. And the same held for integrated circuit technologies, when computers built around them were first developed. The basic functional elements that went into making all of them were both know and widely employed and in both products for the consumer marketplace and for more specialized and business-oriented products as well. Consider those early vacuum tube computers as a source of case in point examples. Vacuum tube radios were commonplace in households and certainly in countries such as the United States with widely available electrical power grids to support their use. And they had been for years. And early vacuum tube televisions were coming onto the market too, even if as high priced luxury items more than as standard affordable household items. I cite these two vacuum tube innovations, that had already become mainstreamed for the first and that was destined to become so for the second, because of their already existing familiarity.

The same basic story can be said of transistors and of integrated circuits too, with at least some devices already on the market that used them that had already become at least relatively commonplace. But the same cannot be said of quantum computers or of the basic technologies that are required to support them. Quantum computing depends essentially entirely and certainly for its information processing elements, on low temperature physics and its technologies and on the superconducting state, where the highest temperature superconducting materials now known still require liquid nitrogen cooling if they are to function as such. No consumer marketed products and essentially no business marketed ones make use of any of that. And the logic of quantum mechanics systematically violates the common sense understandings that we all tend to automatically presume, based on our everyday experience too.

So in a fundamental sense, quantum computing represents a disruptively new technological advancement that is novel in ways that are qualitatively different than any that have come before and certainly in anything like a computer or automated information processing context. And I really begin this next stage discussion of that advancement with this held out as a core element for thinking about quantum computing per se.

I am going to continue this line of discussion by successively delving into a progression of basic issues that will include:

• A reconsideration of risk and of benefits, and both as they would be viewed from a short-term and a longer-term perspective,
• A reconsideration of how those considerations would be addressed in a design and manufacturing context,
• The question of novelty and the challenges it creates when seeking to discern best possible technological development paths forward, where they can be clear when simple cosmetic changes are under consideration, but opaque and uncertain when more profound changes are involved,
• And lock-in and certainly as that becomes both more likely to arise, and more likely to become more limiting as novelty and the unexpected arise.

I will add to that list as I proceed with this narrative. And I will discuss all of those points from both a technological and a business perspective. And with that noted, I end this posting by returning to its basic series title: “Moore’s law, software design lock-in, and …”, and to the issue of Moore’s law per se.

Moore’s law as more formally stated, concerns next generation advancement in integrated circuit based central processing units, and at most with just related computer chips included there. As such it would be declared over and no longer valid when integrated circuit technology ceases to advance at its predicted pace, and even where that technology per se is still being used and when it is still being technologically advanced, if more slowly. And as formally stated it would not even begin to apply in a non-integrated circuit technology context. That context is fundamentally built into it. But I would argue that the basic aspirational impetus behind that statement of intent, turned “law” was more fundamental than that, and that that underpinning goal and the belief that drives it, apply to wider contexts.

Moore’s law in its more fundamental sense is about change and advancement, and it is about a drive to achieve them and even on a more predictive basis as far as their pace is concerned. So I would argue a case for at least considering what might be considered a Version 2.0 of it that would apply to quantum computing too. I will, of course have more to say on that too in upcoming postings. And I will more explicitly discuss quantum computing and at both a hardware and a software level in the process of carrying out all of this discussion to come.

Meanwhile, you can find this and related material at Ubiquitous Computing and Communications – everywhere all the time 3 and also see Page 1 and Page 2 of that directory. And I also include this in my Reexamining the Fundamentals 2 directory as topics Section VIII. And also see its Page 1.

Some thoughts concerning a general theory of business 34: a second round discussion of general theories as such, 9

Posted in blogs and marketing, reexamining the fundamentals by Timothy Platt on May 9, 2020

This is my 34th installment to a series on general theories of business, and on what general theory means as a matter of underlying principle and in this specific context (see Reexamining the Fundamentals directory, Section VI for Parts 1-25 and its Page 2 continuation, Section IX for Parts 26-33.)

I have been discussing three closely interrelated topics points and their ramifications in this series since Part 31, which I repeat here for smoother continuity of narrative:

1. Provide a brief discussion of the generality of axioms, and of how a theory elaboration can lead to reconsideration of the axioms underlying it, where that can lead to accepted axioms already in place, now being reconsidered as special cases of more general principles,
2. Adding, subtracting and modifying such axioms in general, and
3. The whole issue of consistency, and even just within constrained sets of axioms. And I will frame that discussion at least in part, in terms of outside data, insofar as that makes these theories open – and with a reconsideration of what evolutionary and disruptive mean there.

I have focused on and primarily addressed the first two of those points since then, leading me to the third of them: my primary point of consideration for this posting. And looking beyond that, I am planning on addressing the challenge of direct and indirect testing of potentially validatable assertions that would fit into an open general theory, and what those terms actually mean in practice: direct, indirect and validatable.

I add in anticipation of that discussion, that I proposed a possible reconsideration of a currently untested assumption of modern physics in Part 33: the presumed existence of a Higgs field, by suggesting an at-least theoretically possible approach to testing for its actual presence – where no current or foreseeable technologies could make that possible – at least directly. This line of discussion explicitly connects back to the issues of that third topics point and in fact to the first two as well and I will have more to add in addressing all of them in this emerging context. Then after continuing that case in point example, I will discuss two overarching, closely interconnected (if at times opposing) perspectives that enter into essentially every area of consideration here: the challenges of reductionism and of emergent properties.

And I begin addressing all of this by considering what at least in principle, might appear at first to be a simplest case context for discussing the above Point 3: consistency in a closed axiomatic body of theory.

There are two approaches that I could take here in pursuing that “simplest case” line of discussion:

• The question of whether absolute consistency can even be provable in principle, for axiomatic theories of any real complexity, as addressed for example by Gödel’s incompleteness theorem for bodies of theory that would include within them a theory of arithmetic, and
• The impact of apparent emergent inconsistency at a specific instance level, in an already largely developed body of theory.

I at least briefly and selectively addressed the first of those points in Part 30 (and recommend that you review that in anticipation for what is to follow here.) And I made note in that posting of a closely related second issue too: completeness, or rather the possibility of establishing whether any such body of general theory can ever be completely, comprehensively developed in some absolute sense. Both sets of issues: both sets of fundamental constraints prove to be crucially important here and in the context of this posting too.

Let’s at least begin to think about and clarify, and expand upon that point as already discussed here, by considering the second of the above two consistency issue bullet points and in the specific context of a particular widely accepted body of general theory: physics. And to be more precise here, let’s consider the discovery of and theory-based elucidation of some specific empirically observable, experimentally measurable phenomena as case in point examples.

• To bring this out of the abstract, let’s specifically consider solitons as a first such example.
• The unification of electrical and magnetic phenomena as mathematically described by Maxwell’s equations as a second, and
• The so called paradox of radiation of charged particles in a gravitational field as an historically relevant third.

I begin with solitons because their mere empirically observable existence served as direct challenge to some of the basic assumed tenants of classical physics, as then held sway.

• Solitons are stable waveforms that do not disperse or decay and even over extended periods of time. More precisely, if still somewhat loosely stated, they are self-reinforcing wave packets that maintain their general shapes and amplitudes while propagating at constant velocities, with their speed dependent upon their primary frequencies. And they arise and are maintained by mechanisms that serve to cancel out expected nonlinear and dispersive effects, as they would normally arise in the mediums that they exist in, that would be expected to dampen them.
• Mathematically, these solitons present themselves as solutions of a class of weakly nonlinear dispersive partial differential equations so while they did not comfortably conform with the physics in place when they were first formally characterized, they can be and in fact were precisely mathematically described and early on.

And this story does go back a ways. A Scottish civil engineer named John Scott Russell is credited with first formally describing them in 1834 in the professional literature, from when he first observed this phenomenon in the Union Canal in Scotland and then replicated it in a controlled laboratory setting.

On the face of things, solitons appear to violate numerous basic scientific principles, as laid out in classical physics. To cite a specific example there, they specifically appear to violate the second law of thermodynamics, and certainly as that is classically perceived. And in fact the only way to reconcile their existence with the rest of physics as that would serve as an overarching general body of theory, was to completely reframe them and in quantum mechanical terms. (See R. Rajaraman’s 1982 paper on that, as a foundational document as to how that reframing was reached: Solitons and Instantons.)

Why do I include this example here? It arose as a matter of replicably observed reality and it continued to exist as such as an unexplained, or rather an unacceptably explained observed phenomenon that did not fit into a large and seemingly exhaustively tested and proven body of theory, creating an at least apparent paradox problem if nothing else. And it was the accumulation of more and more such “outsider” phenomena that led to the development of quantum mechanics and the more expanded and nuanced body of theory that was finally able to include in this phenomenon too.

Maxwell’s equations offer a fundamentally different, but still parallel lesson here. Electricity and magnetism were well known and studied phenomena in the mid-nineteenth century, when Maxwell and others began an effort to unify them in some way, that led to his conceptual and mathematically stated breakthrough for achieving that goal (see this History of Maxwell’s Equations for further details on that.) And when that unifying description of what would now be called electromagnetism and electromagnetic properties was first published, it was heralded as a fundamental, groundbreaking advancement to the then-prevalent understanding of the physical world that held sway at that time. And it did offer very real and significant descriptive and predictive value. But this body of theory, it turns out, was not as fully consistent with mainstream nineteenth century physics as was then presumed. And some key aspects of it turned out, upon further study to be effectively independent of that physical theory’s basic weltanschauung, and I view that prevailing theory in that way because it did represent the world view of physicists and of scientists in general, of the time.

In a fundamental sense, the electromagnetic theory that Maxwell encoded in his equations, was a forerunner to what in the early twentieth century became known as the special theory of relativity. And his basic theoretical model, as he developed it, translates quite nicely to a general theory of relativity context too, as briefly outlined in Maxwell’s Equations in Curved Spacetime.

Why do I cite this example here? Solitons represent an example of how a replicably observable phenomenon can demand new basic theory and new basic underlying axiomatic assumptions that would underlie that, if it is to be meaningfully included in anything like a coherent and at least overtly gap-free general theory. Maxwell’s equations somewhat fit into the basic, accepted general theory then in place, but they could never fully fit in there because they were ultimately to prove more compatible with a more expansively inclusive, still yet to be discerned alternative axiomatically grounded vision of reality. And with that, I turn to my last such example here, as promised above: the at one time, painfully apparent paradox of radiation of charged particles in a gravitational field, as cited above.

According to every prevailing understanding of how particles at rest should behave in a general theory of relativity context, they should radiate energy and on a largely continuous or at least highly temporally structured ongoing basis. But observationally, they do not. Did this reflect a fundamental error or gap in the general theory of relativity per se? Did this reflect more fundamental problems in the basic, axiomatic assumptions held in place there? Ultimately this paradox was resolved by reconsidering frames of reference, and how those axioms and a range of conclusions derived from them should best be understood and interpreted. So this example adds in the challenge of how best to understand and make use of the axioms that are in place too.

All of the basic issues that I have sought to raise and highlight with my physics examples here, have their counterparts in other large general bodies of theory, any realistic possible theory of business included. And that certainly holds true when such bodies of theory are thought of in terms of the more general principles that I just raised in the context of the above three examples.

I am going to continue this narrative in a next series installment, as briefly outlined above. This will mean reconsidering the issues of consistency and completeness, in light of the impact created by disruptive and seemingly disruptive examples, as for example raised here. And this will include, as noted above, my addressing the challenge of direct and indirect testing of potentially validatable assertions that would fit into an open general theory, and what those terms actually mean in practice: direct, indirect and validatable. Yes, that will mean reconsidering my Higgs boson/Higgs field physics example from Part 33 again too. And I will also mean my further considering what evolutionary and disruptive mean: terms and concepts that I have made at least case by case level use of throughout this blog. And I will also, of necessity connect back in that to my discussion of compendium theories as discussed in this series in its Parts 2-8.

Meanwhile, you can find this and related material about what I am attempting to do here at About this Blog and at Blogs and Marketing. And I include this series in my Reexamining the Fundamentals directory and its Page 2 continuation, as topics Sections VI and IX there.

Reconsidering Information Systems Infrastructure 14

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on March 16, 2020

This is the 14th posting to a series that I am developing, with a goal of analyzing and discussing how artificial intelligence and the emergence of artificial intelligent agents will transform the electronic and online-enabled information management systems that we have and use. See Ubiquitous Computing and Communications – everywhere all the time 2 and its Page 3 continuation, postings 374 and loosely following for Parts 1-13. And also see two benchmark postings that I initially wrote just over six years apart but that together provided much of the specific impetus for my writing this series: Assumption 6 – The fallacy of the Singularity and the Fallacy of Simple Linear Progression – finding a middle ground and a late 2017 follow-up to that posting.

I have been discussing artificial intelligence tasks and the agents that would carry them out in this series, as categorically falling into three at least roughly distinguishable domains that would fit along a single continuum of complexity and subtlety (and difficulty for actually developing them into operational forms):

• Fully specified systems goals and their tasks that could be carried out by what are essentially still just tools, even if complex and sophisticated ones,
• Open-ended systems goals and their tasks (e.g. natural conversational ability with its lack of corresponding fully characterized performance end points or similar parameter-defined success constraints) that could only be fully carried out by agents that are developed to the point where they would arguably qualify as persons, and
• Partly specified systems goals and their tasks, that would fall in between those two extremes, for what agents would be able to carry them out (as in self-driving cars where they can be programmed with the legal rules of the road, but not with a correspondingly detailed algorithmically definable understanding of how real people in their vicinity actually drive and sometimes in spite of those rules: driving according to or contrary to the traffic laws in place.)

I have primarily discussed gray area, pertly specified goals and tasks and the agents that would be required at minimum for being able to carry them out, up to here in this series, citing and discussing their simpler fully specified counterparts as needed. Then I began setting a more explicit stage for delving into the issues and challenges of what can only be considered open ended tasks and goals and their agents, in Part 13 where I listed in for there-summary form, a set of four possible sources of tools for developing such agents. And I stress that word: develop here rather than build as I at least start out with an axiomatic level presumption in my thinking about this, that a true artificial general intelligence as would be called for, for genuinely open ended tasks, would have to have a self-learning based, ongoing ontological development capacity built into it and with all of the autonomy and potential for it that that would imply.

Those four “tool packets” if you will, are:

• Ontological development (as just cited here) that is both driven by and shaped by self-learning behavior ( as cited in Part 13 in terms of antagonistically positioned subsystems, and similar/alternative paradigmatic approaches),
• Scope and range of new data input that might come from the environment in general but that might also come from other intelligent agents (which might mean simple tool agents that carry out single fully specified tasks, gray area agents that carry out partly specified tasks, or genuinely general intelligence agents: artificial or human, or some combination of all of these source options.)
• How people who would work with and make use of these systems, simplify or add complexity to the contexts that those agents would have to perform in, shifting tasks and goals actually required of them either more towards the simple and fully specified, or more towards the complex and open-ended.
• And I add here, the issues of how an open ended task to be worked upon and goal to be achieved for it, would be initially outlined and presented. Think in terms of the rules of the road antagonist in my two subsystem self-driving example of Part 12 here, where a great deal of any success that might be achieved in addressing any overtly open-ended systems goal will almost certainly depend on where a self-learning agent would begin addressing it from.

One of the perhaps-defining features of my overall narrative as offered in this blog, as has repeatedly been noted here, is that we all make assumptions in our thinking. And I add that we particularly do so and with a more maximized impact from that, when we seek to think and plan in as organized and analytical a manner as possible so as to gain as much meaningful consistency and surety from our efforts as possible. That definitely applies here. And to continue this brief connecting and organizing note, I add that I tend to be just as active in pointing out the underlying assumptions that I have just been making, so I can justify them, expand upon them, reduce them to special case-only status, or otherwise challenge them – and with a goal of prompting a wider ranging thought process about the issues at hand. I pursue that approach here, when I begin this main line of discussion for this posting by noting and at least partly challenging an assumption that I have built into this series up to here, and certainly when addressing gray area challenge tasks such as the development of self-driving vehicles. That presumption can be contained for the most part in one word: deterministic (as opposed to stochastic.)

To clarify that, let’s reconsider the admittedly gray area problem of those self-driving cars: an area of artificial intelligence systems development that is very actively being pursued as of this writing. And since this is a posting about open ended tasks and their agents, let’s focus here on the more open ended half of this, that would for its presence and significance render it a gray area problem in the first place: autonomous self-driving vehicles having to operate in the presence of average skills level and at times distracted or impaired human drivers (e.g. drivers who are too tired to drive safely but who are doing so anyway, drivers who are too actively involved with their cell phones while supposedly driving attentively, drivers who are drunk or on other drugs, or who have innately poor judgment to begin with.)

• Deterministic systems follow more consistent, simpler cause and effect patterns with what are usually simpler sets of A’s following B’s, following C’s as ongoing occurrence chains.
• Stochastic systems with their more likely or less likely (higher or lower probability) cause and effect event patterns, add significant additional complexity to this and even when there is available actuarial or similar a priori data available for accurately predicting the likelihood of occurrence of essentially any and all of the possible situations that might be faced at any given time.
• And when low probability cause and effect outcomes have to be considered in any meaningful risk management calculations taken: when even low probability actions might be taken by other (human) drivers in this example, that can lead to what amounts to a combinatorial explosion in the number of possible events and situations that a self-driving car or truck might have to be ready to respond to and immediately in its here-and-now.

Let’s take that out of the abstract by posing an admittedly fictionalized possible decision making internal conversation on the part of such an autonomous vehicle, that I would pose here for illustrative purposes:

• Is that blue SUV being driven by another self-driving agent that for whatever reason I am not real-time communicating with even though we are in close proximity to each other, or is that a human driver?
• Is that driver performing erratically, and if so, are they driving in a way that would create additional meaningful risk or uncertainty for themselves and other vehicles in their vicinity?
• It appears there is a human behind the wheel, but many autonomous vehicles have human passengers who sit in the “driver’s seat” and even in fully autonomous vehicles still.
• Is that vehicle drifting left and right beyond a range of what would nominally be within-lane motion, in a manner that would be risk creating?
• How likely is it that that vehicle will cross the center line of this road and enter a collision-risking pattern with this vehicle, as we approach each other in our opposing direction lanes?
• That vehicle has just crossed over the center line, partly into oncoming traffic and then veered back into its own lane again, and without passing another vehicle or swerving to avoid hitting anything. What action should I take in the event that this happens again and particularly now, when it has moved much closer?
• (Nota bene: the later in this process, and the closer those two vehicles are to each other when the self-driving vehicle of this scenario decides to take evasive or other self-protective action, the fewer the options it will have left as available to it for doing so. At the same time, the earlier it begins any corrective protective action, the more likely it is that it will affect surrounding traffic unnecessarily, slowing down all other vehicles in the area as they in turn have to make adjustments to their driving too.)

The more a proposed task, or the realization of some goal shifts from the simple fully specified, to the genuinely open ended, the more significantly stochastic it becomes for how acting upon it would be planned out and executed, and the more complex any ripple effects become as other agents and participants have to adjust their actions to accommodate what might be the unexpected as that plays out around them. Think of this as a matter of low probability events and responses made to them as adding friction to entire multi-agent systems.

• And the more open, or close to that that a task becomes as issues such as friction have to be accommodated in its being carried out, and the wider the range of long-tail, low probability events that might arise that would be of importance in that planning and execution and the more of them that might be unpredictable, the more complex and difficult the task becomes of developing agents that can perform it and at any given risk-management level of performance.

Having made note of that, let’s reconsider this overall problem from a strictly risk and benefits perspective, and with a focus here on the risk side of that. And to take this out of the abstract, let’s do so at least to start, in terms of a specific real world scenario.

If a pedestrian were to run out in the street from between parked cars as for example when a child rushes forward without looking or thinking, to retrieve a lost ball while playing:

• That might be a very low probability event and certainly in a non-residential neighborhood, where children would not more routinely be expected to play,
• But the consequences of an autonomous vehicle hitting and killing such a child would be so significant that a single such occurrence might lead to all autonomous vehicles running the same basic artificial intelligence software that that vehicle does, being taken off the road until a comprehensive fix can be developed.

Now let’s consider that type of risk and benefits determination in a natural conversation context, where human people can and do miscommunicate at times and where that can have trivial impact at worst, or severe consequences – and even at the level of the vehicle versus pedestrian example that I just made note of. And I begin addressing that challenge by suggesting a “minimally acceptable” though not complete solution, or rather minimally acceptable approach to carrying out artificial agent-based conversation, as the larger and more comprehensive goal of true open ended natural conversation is still being worked towards:

• An artificial intelligence based natural speech capability need only be good enough to satisfy the basic needs of the circumstance that it is being deployed in, without creating extraordinary undue risk – here, risk that would exceed that found in human to human conversation under similar circumstances.
• And this becomes open ended when “similar circumstances” as cited there, expands to cover wider and wider ranges of potential shared communications need, and particularly as risks arising from miscommunications there becomes greater.

Consider Alexa and Siri and other artificial intelligence digital assistants that are currently available now, with their current communications limitations and their current rates of realized and predictable miscommunications. Those two bullet points are in fact consistent with how they and other currently available artificial intelligence digital assistants are being developed – a point of similarity that is not in any way coincidental.

I am going to continue this discussion in a next series installment where I will reconsider the four point tools set that I made repeated note of at the top of this posting, there framing them in stochastic terms. And in anticipation of that discussion to come, I am going to at least begin it with a focus on what might be considered artificial intelligence’s bootstrap problem: an at least general conceptual recapitulation of a basic problem that had its origins in the first software-programmable electronic computers.

Meanwhile, you can find this and related postings and series at Ubiquitous Computing and Communications – everywhere all the time and its Page 2 and Page 3 continuations. And you can also find a link to this posting, appended to the end of Section I of Reexamining the Fundamentals as a supplemental entry there.

Moore’s law, software design lock-in, and the constraints faced when evolving artificial intelligence 11

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on March 10, 2020

This is my 11th posting to a short series on the growth potential and constraints inherent in innovation, as realized as a practical matter (see Reexamining the Fundamentals 2, Section VIII for Parts 1-10.) And this is also my eighth posting to this series, to explicitly discuss emerging and still forming artificial intelligence technologies as they are and will be impacted upon by software lock-in and its imperatives, and by shared but more arbitrarily determined constraints such as Moore’s law (see Parts 4-10.)

In many respects, my Part 10 installment to this series can be seen as an anticipatory note for a discussion of what can be considered:

• An approaching end of both Moore’s law as a still valid metric for predicting and pacing the development of next generation integrated circuit chips,
• And of integrated circuit technology per se as a best and even only approach possible for developing next generation, cutting edge computer systems: artificial intelligence agents and their systems included.

And I ended that posting by offering a nod in the direction of quantum computing as a next probable cutting edge technology base (after noting more towards the beginning of that posting that I had intentionally excluded it from consideration in this narrative up to then.)

Moore’s law will fail and for a variety of synergistically interacting reasons, some of which I have touched upon in this series. The basic laws-of-nature level limitations on what can be done using semiconductor materials and solid state physics, will be reached. And I note that, while acknowledging that this rule of thumb “law” has held true, albeit with a great deal of effort, over an amazing span of years and decades – and in spite of an ongoing flow of what have turned out to be its premature obituaries.

The age of Moore’s law will end. And the pace of development of next generation semiconductor technology-based information processing devices that it has been predicting so successfully, will drop as the underling technologies behind creating them reach true maturity. And looking beyond the narrow constraints of integrated circuit chip development and its evolution per se, to consider all of the fruits of semiconductor technologies, that larger arena of development will eventually mature too, as its fuller capabilities are reached in practice (and as new cutting edge alternatives are pursued, reducing and then removing much of the pressure and incentive to make better, from basic silicon and related materials as per our current integrated circuit paradigm.)

And with that noted, I add what might be considered a third basic driver to the two that I have already been focusing upon here in this series: Moore’s law and design lock-in. And that third fundamental piece to the puzzle that I have been discussing here is hope. And it, in many respects, is the most fundamental of the now-three.

• Moore’s law is fundamentally grounded in hope.
• The basic decision-making processes that lead to lock-in, as well as to successful and even game changing innovation through development of the New, are all driven by hope.
• And hope infuses all of our thinking, as to the realizable capabilities and potential of quantum computing too, as that has increasingly come to be seen as the next new technology development arena that would supplant the semiconductor technologies of today as a source of dramatically more powerful computers and computer systems …
• … and even as that technology is still in its essentially embryonic form.

I find myself thinking in terms of a wider historical timeframe as I write this, going back at least as far as Charles Babbage and his only embryonically realized analytical engines – his only partly realizable efforts to create what would have been a true general purpose computer using gears, shafts, escapements and related components, and the technology of mechanical engineering.

And I find myself thinking of those early electronic-age, vacuum tube technology computers that among other things helped to win World War II for the Allies by helping them to break Nazi German military codes. And that technology continued to advance after that war as the fundamental source of all new computer development and advancement – until the advent of the transistor and of the first crude integrated circuits. (As a side note, it still amazes me how long the British government held as top secret, the development of what was undoubtedly the first fully electronic, vacuum tube based computer, that was used in at least large part as just noted, to routinely rapidly decode German military communications. The existence of that by-then historically quaint device was not publically acknowledged until well into the age of the integrated circuit-based computer, decades later!)

That brief digression aside, the ENIAC: perhaps the best known large scale manifestation of this basic computer technology, at its most expansive point of development in late 1956 when it was finally retired from service, “contained 20,000 vacuum tubes; 7,200 crystal diodes; 1,500 relays; 70,000 resistors; 10,000 capacitors; and approximately 5,000,000 hand-soldered joints. It weighed more than 30 short tons (27 tons), was roughly 2.4 m × 0.9 m × 30 m (8 ft × 3 ft × 98 ft) in size, occupied 167 m2 (1,800 sq ft) and consumed 150 kW of electricity” (with those specifications quoted from that just cited Wikipedia link.) And its computations played a role in the design and development of the first hydrogen bomb, among other things. It is easy and common to make note of how much more computational power, even simple consumer goods have now with their embedded integrated circuit technologies. But the more limited technological reach of that earlier basic vacuum tube technology was world changing and not just in a military context or from that perspective.

Gear-driven computers of any scale and capability that would be required in order to actually build one of Babbage’s analytical engines, were not possible given the basic limitations of that technology. It was not possible to build that complex a system with the per-part and larger-assembly tolerance limitations that it would call for (where in the real world there are always margins of error in precise size and shape of mechanical parts produced, and where the cumulative effect of temperature based metal expansion and contraction and of friction, among other factors, enter in too.) But Charles Babbage and his programmer Augusta Ada King, Countess of Lovelace, has bigger visions, and dreamed of building to achieve them.

The invention of the vacuum tube, opened up whole new worlds of opportunity there, and when that potential was sufficiently realized the first vacuum tube computers were built, all arriving as unique one-off efforts and all developed and build with specific pressing problems to solve in mind. (I intentionally skip over the issues of mechanical, or electromechanical computers, as what in many respects could be seen as a final attempt to fulfill Babbage’s dream in what was fundamentally still a would-be mechanical, pre-electronic computer world.)

And then came the transistor and the integrated circuit and the scale and scope of the dreamed-for grew accordingly – from striving to build true general purpose computers and ones of vast capabilities, to the dream of creating true artificial intelligence, and ultimately true artificial general intelligence. And as that basic technology appears to be approaching its fundamental laws of nature-driven outer boundaries (albeit with what is most likely still a significant way to go there, as of 2020), a next possible future successor is starting to take shape: quantum computing.

Quantum computing is still in its embryonic stage of development, even if it has proven itself to hold a capacity for significantly further development. And I turn back to Babbage’s proposed analytical engines as envisioned but as proven impossible to actually build, as a clarifying counterexample, as I offer that assessment. Babbage’s pre-electronic age computers could not be developed from initial conception into actual devices capable of solving complex problems, let alone specific benchmark problems that could not be solved in any realistic timeframes by other available means – even when deploying and using the very best, most capable currently available older technology systems. A quantum computer has now achieved what is referred to as quantum supremacy: the capacity to solve in minutes, a specific benchmark problem that even the fastest of our current integrated circuit-based supercomputers would require thousands of years to solve. See:

Quantum Supremacy Using a Programmable Superconducting Processor, and
Google Claims a Quantum Breakthrough That Could Change Computing.

I am going to continue this discussion in a next series installment where I will return to the issues and perspectives of the adaptive peak model that I have borrowed from the biological sciences, for discussing patterns and potentials for ongoing innovative development in this type of context (as already cited here in Parts 3, 5 and 6 of this series.) And as part of that, I will reconsider both Moore’s law and lock-in for what they are, when considered from a longer time frame perspective.

Meanwhile, you can find this and related material at Ubiquitous Computing and Communications – everywhere all the time 3 and also see Page 1 and Page 2 of that directory. And I also include this in my Reexamining the Fundamentals 2 directory as topics Section VIII. And also see its Page 1.

Some thoughts concerning a general theory of business 33: a second round discussion of general theories as such, 8

Posted in blogs and marketing, reexamining the fundamentals by Timothy Platt on March 7, 2020

This is my 33rd installment to a series on general theories of business, and on what general theory means as a matter of underlying principle and in this specific context (see Reexamining the Fundamentals directory, Section VI for Parts 1-25 and its Page 2 continuation, Section IX for Parts 26-32.)

I have been discussing three topics points and their ramifications in this series since Part 31, which I repeat here for smoother continuity of narrative:

1. Provide a brief discussion of the generality of axioms, and how a theory elaboration can lead to reconsideration of the axioms underlying it, where that can lead to accepted axioms already in place, now being reconsidered as special cases of more general principles.
2. Adding, subtracting and modifying axioms in general, and
3. The whole issue of consistency, and even just within constrained sets of axioms. And I will frame that discussion at least in part, in terms of outside data insofar as that makes these theories open – and with a reconsideration of what evolutionary and disruptive mean there.

And to be more specific in this connecting text lead-in to this posting and its intended discussion, I have addressed that complex set of interconnected topics up to here with a specific focus on the first of them. My goal here is to turn to the second of those topic points and address the issues of generality and of special case rules. “Axiom” as such and what that word actually means, is often thought of in a manner that can only be considered to be axiomatic, with that very concept more usually taken as an unexamined given than anything else.

The concept of axiom only really comes to the forefront of thought when specific axioms and the conceptual systems that they serve as intended foundations for, come into question. And that can happen in closed axiomatic theory contexts, as in the case of alternative approaches to geometry as briefly discussed in Part 32, or in open axiomatic theory contexts as in the case of Newtonian and Relativistic general theories of physics as also touched upon in that posting. (See Part 32 for summary definitions of axiomatically closed and open bodies of theory, as I will continue making use of that point of distinction in what follows.)

I begin addressing the above-repeated second topics point of the list of them that I am working on here, by repeating a couple of what would probably be fairly obvious details that I have already made note of:

• When axioms come into doubt and are challenged in an axiomatically closed body of theory, as for example when parallelism and its axiomatic description has been challenged, they are most often replaced in total with new alternatives, or eliminated as to type entirely. They are not reduced to special case or similar status and retained as such; they disappear – and with their replacement forming part of the axiomatic underpinnings of a new, alternative general body of theory.
• And when axioms come into doubt and are challenged in an axiomatically open body of theory, as happened in a physical sciences context when observations began to be made that did not fit into a Newtonian axiomatic explanatory model, they tend to be reduced to special case rules and rule of thumb approximations, even if very useful ones.

Newtonian physics based calculations of all sorts are still routinely used and essentially wherever their theoretically more precise relativistic calculation alternatives would offer corrections that are so minor in scale as to be dwarfed by any possible measurement errors that might obtain in a given empirically based setting. So classical Newtonian physics can be and is used as a font of special case wisdom and utility. But that noted, theory opening outside data (such as experimental data) can also lead to the complete repudiation of and elimination of what was once considered an inviolable axiomatic law of nature principle too. Just consider the theory of the ether, and particularly in the context of electromagnetic energy and its propagation. It was axiomatically assumed that there had to be an all-pervasive space-filling medium that electromagnetic energy would travel through as an essential requirement for its propagation from any point A to any point B in space. Then the Michelson–Morley experiment: initially intended to detect and characterized this assumed medium, demonstrated that it did not in fact exist.

• That led to the repudiation and elimination of an axiom-level assumption, of the possible and even likely existence of necessary all-pervasive space-filling medium in space and time, and apart from space and time per se that would play an essential role in energy transfer and in fact in causal connectivity insofar as that would depend on energy transfers.
• And then Peter Higgs developed his theory of mass as a consequence of the action of a special categorical type of boson (subatomic particles with whole integer spin), now known as the Higgs boson, that serves as a conveyer of and a conferrer of mass and for all subatomic particles that can interact with it, and for all matter that would be comprised of those particles. And that particle was detected and in replicated experiments, proving its actual, empirical existence. And that particle and its action, at least according to current theory, requires that another form of all-pervasive space-filling medium exist instead: the so called Higgs field (see above-linked piece on the Higgs boson for a quick introduction to this too.)
• So the electromagnetic ether disappeared from physical theory, but an essentially-ether based theory came back to replace it – at least for now, absent any realistically possible way to validate or disprove its existence as a next step counterpart to the Michelson–Morley experiment.

Electromagnetic energy and photons (as boson particles that contain and convey it) experientially exist as validated empirical realities. The concept of the ether in their context was taken as a (not yet) testable axiomatic truth, and a de facto part of the non-empirically based side to physical theory as held, until that experiment moved its possible existence into the open side of that body of theory. Mass, and more specifically the Higgs bosons that manifest it at its most basic subatomic level have now been experientially proven to exist too. I would argue that its ether: the Higgs field might or might not exist, and certainly as it is currently understood. But either way it is part of the as of now at least, closed side to current physical theory as far as understanding mass at that fundamental level is concerned.

• So a fundamental axiom can go away in an axiomatically open body of theory – and then come back again as has happened here with presumptions of all-pervasive space-filling mediums that might exist as separate from space and time itself.

Nota bene: As an idly considered alternative, or rather as a source of such an alternative to the Higgs field as an all-pervasive medium for transmitting the effects of Higgs bosons, it crosses my mind that a reconsideration of the nature of space-time itself might be in order. And I see possible value arising there, in addressing this problem by carrying out this reconsideration with a particular focus on a scale approaching that of the Plank distance as a spatial unit, and for time, the interval that it would take a photon to traverse one Plank distance unit, or at least some small number of them. I cite this as a possible source of insight while thinking through the story of electromagnetic energy and its photonic bosons, and in the light if you will of the Michelson–Morley experiment and its confirmation, and with a tight linkage established between minimum paths between points in a general relativistic space-time, and the paths that photons in fact follow there. Light, to focus on one band of the electromagnetic spectrum, travels through space itself and in a way that both shapes and in many respects defines space-time and certainly as a source of observable phenomena. Does the same basic principle that I loosely articulate there, apply to the phenomena that we associate with mass too?

That, of course, leads me to the question of experimental validation and the issues of moving what amounts to axiomatically closed conjecture into the realm of empirically testable axiomatically open territory. And I would suggest that given the tools available and in use in modern physics, that testing anything like my conjecture here would call for call for a particle accelerator that would at minimum be orders of magnitude larger than any that have ever been built or even seriously considered. Developing a mathematical theory that would descriptively and predictively model a direct, aspect of space alternative to the Higgs field (and that would prove or disprove the existence of that too), would be the easy part and certainly when compared to actually carrying out this type of direct test.

And that leads me to the issues that I would at least begin to address here as I conclude this posting: the issues of what constitutes direct tests, or indirect ones in general in an axiomatically open body of theory context. I will simply note for now that any determination of direct, as that term would be used here and in similar contexts, would depend on what is axiomatically assumed and on what is routinely done, and on what types of tools are available and in routine use.

Setting that side note aside, at least for now, I am going to continue this discussion in a next series installment where I will turn to consider the third and final topics point of my above-repeated list:

3. The whole issue of consistency, and even just within constrained sets of axioms. And I will frame that discussion at least in part, in terms of outside data insofar as that makes these theories open – and with a reconsideration of what evolutionary and disruptive mean there.

Then I will consider two areas of consideration: the challenge of reductionism and of emergent properties, and the challenge of direct and indirect testing and what they actually are.

Meanwhile, you can find this and related material about what I am attempting to do here at About this Blog and at Blogs and Marketing. And I include this series in my Reexamining the Fundamentals directory and its Page 2 continuation, as topics Sections VI and IX there.

The challenge of strategy precluding tactics, and vice versa

Posted in reexamining the fundamentals by Timothy Platt on January 20, 2020

Strategy and tactics are often presented and discussed as if representing somehow-opposites of an either/or, starkly differentiated dichotomy. But for a variety of practical purposes and in a variety of contexts it can make more sense to see them for how they would fit together along more commonly shared continua that are scaled along organizational-structural, and action-and-response based analytical scales.

I began addressing the second of those two approaches to thinking about strategy and tactics as business planning tool sets in a recent posting, that I decided to include in this blog as a supplemental entry to an ongoing series on general theories per se, and on general theories of business in particular; see Going from Tactical to Strategic and Vice Versa, as an Emergent Process (as can be found at Reexamining the Fundamentals 2 and its Section IX.) And my goal for this posting is to continue its basic discussion, but from a distinctly different perspective than the one that I that pursued there. But that intended divergence from my earlier posting on this noted, I begin again with what are essentially the fundamentals that I built that posting from:

• The presumption of distinction and separation that can be drawn between strategy and tactics, that I just made note of at the beginning to this posting (and in my above cited earlier one), is both very real and very commonly held.
• And it shapes how tactics and strategy are understood and how they are actually carried out.

To cite an example context where those points of observation most definitely hold, consider how often managers or leaders are judged to be taking a strategic (and therefore not a tactical) approach to understanding and organizing their efforts and those of their teams, or a more tactical (and therefore not a strategic) one. As a reality check there, ask yourself what you would read into and simply assume about the two people generically cited by first name here:

• Mary is a real strategist, always focusing on and grasping the big picture.
• Bob is a tactician and plans and acts accordingly.

But for any real Mary or Bob, they probably wear both of those hats at least occasionally and certainly as circumstances would demand that they do, if they ever actually effectively wear either of them.

This, up to here, simply continues my line of discussion of my earlier above-cited strategy and tactics posting. But now let’s add one more factor to this discussion: mental bandwidth and the range and diversity of information (along with accompanying metadata about it as to its likely accuracy, timeliness, etc) that a would-be strategist or tactician can keep in their mind and make immediate use of at any one time. Consider the consequences of a single individual being able to hold at most some limited maximum number of details and facts and ideas in general, in their mind at once, with all of that actively entering into a single at least relatively coherent understanding that they can plan and act from.

Think of this as my invoking a counterpart to the social networking-reach limitations of a Dunbar’s number here, where in this case different individuals might be able to juggle more such data at once, or less so than others, but where everyone has their own maximum capacity for how much information they can have in immediate here-and-now use at any given time. Their maximum such capacity might expand or contract depending for example on whether they are rested or exhausted but they would always face some maximum capacity limitations there and at any given time. And I presume here that this limitation as faced at any one time and for any given individual, remains the same whether they are pursuing a more tactical or a more strategic planning approach.

• An alternative way to think about strategy and tactics and of differentiating between them, is to map out the assortment of data details that they would each be based upon, for where they would be sourced and for how closely related they are to each other from that.

If you consider a tactics versus strategy understanding in that context, it can be argued that good strategists hold more widely scattered details that collectively cover a wider range of experience and impact, when so planning. And their selection and filtering process in choosing which data to plan from is grounded at least in part on what amounts to an axiomatic presumption of value in drawing it from as wide an organizational context as possible so as to avoid large-scale gaps in what is being addressed. And in a corresponding manner, and according to this approach to understanding strategy and tactics, good tacticians can and do bring a more immediate and localized context into clearer focus in their understanding and planning and with a goal of avoiding the trap of unaddressed and unacknowledged gaps at that level. But the same “content limit” for their respective understandings holds in both cases.

• According to this, micromanagement occurs when strategic intent is carried out with a tactical focus and with a tactician’s actively used data set, and with this done by people who should be offering strategic level insight and guidance.
• And its equally out of focus tactical counterpart would arise when a would-be tactician is distracted by details that are of lower priority in their immediate here-and-now, even if relevant to a larger perspective understanding – where their focusing on them means crowding out information and insight of more immediate here-and-now relevance and importance to what they should be planning for.

And with that, let’s reconsider the above cited Mary and Bob: strategist and tactician respectively. According to this, Mary more naturally throws a wider net, looking for potentially key details from a fuller organizational context. And Bob more naturally focuses on their more immediate here-and-now and on what is either in front of them or most likely to arrive there and soon. And actually thinking and planning along a fuller range of the continua that more properly include both strategy and tactics as cited here and in my earlier, above-noted posting, means being able to switch between those types of data sets.

In most organizations, it is usually only a few of the people there, if any who can effectively strategically plan and lead. It all too often it is only a few who can really effectively tactically plan and lead too. And it is one of the key roles of organized processes and systems of them in a business, that they help those who would plan and lead and either strategically or tactically, to do so more effectively and with better alignment to the business plan and its overall framework in place. Good operational systems make effective tactical and strategic management and leadership easier.

It is even rarer that an individual in a business or organization be able to effectively take either a strong and capable tactical or a strong and capable strategic approach and that they be able to smoothly transition from one to the other and back as circumstances and needs dictate. And ultimately, this scarcity probably dictates the strategy versus tactics dichotomy that I write of here, more than anything else does.

• This discussion up to here, of course, leaves out the issues of how those working data sets that strategists and tacticians use, would be arrived at and updated and maintained, as an ongoing understanding of the context that any such planning would have to take place in. I leave discussion of that to another posting.

Meanwhile, you can find this and related material at my Reexamining the Fundamentals directory and its Page 2 continuation, as topics Sections VI and IX there, and with this posting specifically included as a supplemental addition to Section IX there.

%d bloggers like this: