Are these two important?
- 1st – AI, not solely ML
- 2nd – Knowledge and truth (see below)
- 3rd – Physicalness and mathematics
- 4th – ML’s emerge/surge and data/decisions
- 5th – Where do we go from here? (pending)
By: John M. Switlik, January 2024; Context: Sperry Univac Knowledge Systems Center.
This article continues a series that started with “Artificial Intelligence, not solely machine learning” (Linkedin, TruthEng). What we discussed before pertains to how ML highjacked (or attempted to highjack) AI. We, the humans involved, want AI back. The 1st article briefly looked at a few of the important topics: AI and Lisp; Knowledge-based engineering (KBE); and how the Boeing 777 development project involvement with KBE contributed to the start of the discipline. Basically, the notion is that ML needs KBE, and its like, for many reasons which we will discuss.
From its start, KBE proved to offer useful supplemental resources and processes to the computing of the time. Through continued use, sometimes, with other names, KBE progressed tremendously over the past two decades. We can use a look at the current situation in the light of KBE and discuss the future from a few perspectives of truth.
Coming articles will address these topics: Physicalness; More on mathematics; ML’s emergence and surge; and Data and decisions. That last topic will use a patent related to data science as its framework for discussion.
Now, in the following sections (below and in subsequent articles), the material is independent of the rest but ties cohesively with the others. The necessary relations between the topics will be discussed later. Now, we are setting up the framework for future discourse that has meaning and actionable attributes.
Continuation of the series – #2
Today, we will look at these three:
Twins and more – We know nature having been, as a species, here for a long time. Individually, we have learned. Collectively, we used science for knowledge. The computer? It was an artifice of our prowess with artifacts. But, it was all natural. Now, we have something that comes to “life” on the computer that requires us to change our thinking. We have not done this yet. It’s about time.;
Engineering – We change the world. Usually, it’s by hook or crook. Engineering is a discipline that changes the world with our approval. Sometimes we enjoy the change; other times not. But, there are mechanism to handle the change. We just witnessed two decades of uncontrolled change with impacts that we have not understood yet. The latest? Open AI’s little sandboxy thing. So, engineers might know nature to an extent; they don’t know computing as it ought to be known.; and
KBE now – This is a framework to apply computing to the processes and thinking of engineers. It’s newish, having started in the 1980s/90s. Some thought it had a winter with the other AIs in the 1990. We are here to show that KBE did not and is quite alive and doing well. Albeit, KBE is not the only “moniker” being used. So, we will look at that.
In the context of this article which extends several threads related to AI (as it ought to be discussed), a useful concept to consider is “digital twin” which has widespread usage, nowadays. Though, at the same time, some might be recoiling from the use of “digital” as we used it three decades ago. And, the word has been overused and misused. We can ignore that controversy for now and put the concept to good use.
Essentially, we are talking about things that are physical (in the world of nature as are we humans) having corresponding computational models related to them. That thing on the computer can be called the twin whether in the aggregate or with respect to some part. There are many varieties of twins to consider, but we will look at the types provided by KBE.
Skipping over a bunch of details, we could say that photos of a person can be the start of a twin on the computer. We all know that would be very superficial, but one might wonder given some uses today whether people see more to the situation than there is (photos given by generative means is an example). As we know, a “digital” twin of a person would be a huge affair. The question to keep in mind, how close could such a twin come to the reality of that which it is twinning?
Okay, let us take one of these cars that are self-driving. It would have a twin (let’s say, somewhere in a conglomerate of devices that compute) which knows a lot of information about the auto that its twinning. The use of the plural here is to acknowledge that we are in an age of disparately located devices where distributed means help provide a type of coherent whole.
Let’s make two things clear. The auto is in the world. It is one thing. And, it is our creation and has properties that we can assess. We will call it, henceforth, that which is twinned: Twt. Now, to get to showing our points, we have to talk about Twt and the twin{s) that have been created for it. We would like to use Din. So, we will be talking about Twt and how it came about as well as at the Din(s) and what they entail. For ease, let us just use T and D, respectively, for these two. Again, the former is the one with natural physicalness. The latter is about our abilities to create artifacts on the computer. But a twin could have access to information about physical aspects of the T and even have physicalness itself. For now, we will keep it simple and punt the more complex variety to later.
Both concepts are new as opposed to our human selves (for which we will use H) which are old hat to many. At some point, we will have to address further issues related to the issues of philosophy (grounding), psychology (many factors) and general humanness (with western civilization as a focus). There is the need to provide a reasonable framework for the discussion which will anticipate the coming downturn and work to keep KBE properly going into the future.
So continuing, the two artificial entities (T and D) offer an interesting mix. The former is like any auto but has additional equipment on board: sensors; actuators; and processing power to handle local computing needs. Too, it is tied operationally to something on the cloud (C) that supports the situation of trying to handle incoming data (all sorts) plus make decisions. On the other hand, the latter (computing milieu) is that which provides to the C piece the various facilities which support operational situations. But it is more as it can mimic (digitally) properties of the T thing. As well, C is an entity of vast proportions in its own right. Pieces of C get into action via various schemes that go back to the beginning of computing. Thank us for avoiding that right now.
The main crux now? The auto (T) is real enough to run you over. That’s sufficient to know. The twin (D) is via BoBs various ways. And, H can have several roles, such as driving the T, playing around with the interface with T and D, being the main brain behind the scheme, and such. Or, H can be like the author and having observed the idiocy of some trucks on the road wants to tell people about it. Oh, yes, the thing would not pass a test if we designed the test correctly. Too, those brainy types on board? Please, go talk to some million-mile, safe driver and, perhaps, a lot more. This stuff was not Silly-Con valley at its finest.
Back to business, one can see that the mixtures are vast. Each needs some attention which is what engineering and science will bring to the table. So, this piece of the puzzle? Engineers have been doing their part, from what I can see. Thank them. Even do so for those involved with the auto-auto, as they’re dealing with sensors, actuators, and other essential parts needed for progress.
In the meantime, we have to talk about change and its agents. Engineering is all about change. We all do that sort of thing. However, for public affairs, we have assigned the duty to degreed engineers, usually accompanied with a requirement for licensing. Naturally, there are others who make change, but engineers are related to the KBE that we are discussing. And, we need to get engineering in all aspects into AI; they can tame ML’s overreach, one would hope.
Finally, there are uses of engineering that are not kosher. Financial type? This was not a gift from MIT.
Formally, engineering deals with applied science, at its core, abetted by mathematics. In the common sense, there are several other connotations of the term. As said, we all engineer when being creative (Note 1). And, engineering is the one discipline that regularly changes the world in ways that have allowed us to enjoy (or not) the results. In the practice of engineering, “trades” (some use trade-offs) are a common occurrence dealing with the complexities of choice. Most decisions are not “yes or no”, nor do they reduce themselves down to mere consideration of easy alternatives. As we will see, KBE was (still is) a superb framework for supporting many types of decision.
Business deals with tradeoffs, too, with different jargon and motives. However, we address these from the underlying framework of human decision making as it pertains to all aspects of technology. We saw, in the past couple of decades, PhDs in Physics (and the like), who were grounded more into mathematics than most business types, come over to finance. Want to know the truth? It is a mess now. One word is all that we need about causal issues: ca-pital-sino (neologism created by the author). Take that as a cursory mention for later discussion: “engineering” had a big hand in the mess.
Okay, after that introduction to the complications in store, we are limiting ourselves to core issues and who knows what else that is related. Is there anything more core than quantum mechanics (QM) outside of human issues? In that realm, which we will get to in the future, mathematics plays a heavy role. In fact, it has the heaviest of roles for the gloried discipline. The history of QM illustrates that early experiments allowed some observations of the traditional type. That is, those of the eye-ball types were possible, with crude instruments. But technology got better, as did instrumentation. Behind technology, we can discuss the mathematics and how it (with group theory being a very good example) turned out to be important to QM.
We might say that the use of “D” came to be, whether mathematical or, later, digital. QM demonstrated the cojoining of the nature (being researched) with its twins (our devices). We all know of the remarkable results of this pairing. We may have seen some of the downside. Mostly, those are on the planet of our existence (species loss, as noted by biology). So there is a limit to how far we can address the bad effects. Along another line, computation is just beginning and is too new to show us its true value assessment (good and the bad).
Now, nature and its twins? Another term came to fore by necessity. What we are referring to is “physical state” which denotes that which the twin is not. The twin? It is not physical. What nature provides is so definitely. Stepping up to the cloud, and the AI (or ML), there is a physical piece, say buckets-of-bits (BoBs is what we can use). The main output is not physical, except by side-effect as we have noted. It has no mass, does not really use energy, and, seemingly, is ephemeral (try to touch it). The final things of computing are of the mind (that is, the human). And, psychology will come into play, later (again, stay tuned). May we take this further?
In the prior article, we mentioned that getting involved with an ICAD project was our introduction to KBE. The common trait with earlier work was the Lisp language of John McCarthy. However, the problem set of engineering was a delightful surprise. Not only is the discipline broad, mathematics plays a huge part. As well, there is a long history of solutions that come along from those who worked in the discipline. At the time, only some of this had been brought to the computer.
The situation with engineering and computing was a lot different than we think about now. It was not long before, when slide rules were seen everywhere.
At the time, mainframes were the big servers. There were many other types of computing that were specialized. In terms of engineering, lots of effort had gone into creating workstations. These were mostly targeted to specific types of engineering. For instance, electrical engineers had sophisticated systems due to their type of work dealing with the intricacies of analysis and design. Computer-aided design at the time required a workstation that was tied to the data source on a mainframe. The workstation had facilities for handling geometry and performing actions on such.
On the other hand, a civil engineer later said that the business office of his company had better equipment than did his engineers involved with the construction work of his company, some of which was quite involved (commercial buildings of large proportion).
Be that as it may, in 2022, the focus on KBE seemed to have switched to overseas as searches returned little U.S. activity; yet, quick searches pulled many international papers to the fore. The puzzlement was that we observed KBE as being basic to truth engineering research (Note 2) a couple of decades ago, for which computing’s role was a key discipline. Getting to know the “generative” ways, as proposed and demonstrated by the ML crowd, in 2023, had an impact on priority. KBE was back to being on the table plus there was concerted effort to dig deeper to see what had transpired over the last two decades.
There is a lot to report; we can quote Scott Heide who worked for ICAD while at the MIT AI lab. He was there in the day of the Lisp machine as an early employee. Scott’s current firm, Engineering Intent, exemplifies ICAD leanings (Design automation), as we can see with their rule focus (Rule authoring in Knowledge Bridge). In Scott’s words: While nobody has come up with the perfect name for the technology, it has had a number of monikers: rules-based engineering, knowledge-based engineering, engineering automation, engineering decision-support.
After seeing this quote, further searches about the current status of KBE used the other concepts. Generally, in these additional sources, KBE was mentioned but was not prominent. Basically, given the amount of activity, one might argue that AI never had a winter since there had been lots of progress over the past two decades. Recently, U.S. DOD presented its guidelines for digital engineering. A recent conference was loaded with details about the modern ways, some of which emphasize the importance of good models of physics and the advantages of simulation. To be thorough, we mention some firms dealing with matters related to KBE: Infosys, Ansys, and Paramcs. This is a very small sample. There definitely was no “AI” winter; albeit, we have seen several economic downturns related to other factors which would cause a digression at this point. So, we’ll punt that.
With respect to the above discussion about the D and T (remember them?), there are several things to discuss further. Simulation has really improved and is now a major factor in analysis. Of course, what of the D types are we referring to, since nature itself is being mimicked with computing. We could use this definition: computing is using our ideas about electromagnetic and other phenomena via our configuring matter and controlling energy related to such to bring about side-effects that we find useful.
Along this same line, the topics of sensors and actuators, especially with regard to robotics which may be augmented with AI, will be both interesting to watch but, at the same time, will require more critical analysis using types not seen yet. That’s on the long list of future topics.
Coming topics, again
In short, there is no end to the associative links to other topics. KBE has always been within the framework for research in truth engineering. As such, it is operationally demonstrative. Additionally, we must bring in the factors that go beyond mere show and tell. Coming articles will address some of these as they are very many – say, the total sum of knowledge associated with a university and its intellectual heritage (H talents/knowledge plus artifacts). The claims of those pushing the “generative” modes seem to suggest that there might be a search for “omni” status (science, potence, etc.). We want to establish a reasonable basis for why the claim might be (is) suspect. Further topics to set the stage for discussion are: physicalness; more on mathematics; ML’s emergence and surge; and data and decisions. The latter is related to demonstration with a patent as the initial focus (Note 3). But, to be mentioned everywhere will be KBE.
Context
Boeing Commercial Aircraft
Sperry Univac Knowledge Systems Center
————————————–
Notes:
- Education — traditional school of liberal arts and sciences (University) with a major in quantitative/mathematical economics. Though not an engineer, I have worked in engineering support via advanced computational systems all of my professional life (was always using expensive equipment). So, I can reference things observed, professionally and culturally, in enough of a manner without being too specific. Necessary details can be filled, if required. Then, during the past two decades, I have studied what I call “truth engineering” which came out of my work experience, especially KBE. My focus for self-study (autodidacts offer educated views outside of the peer structure – analog is custom made versus off-the-shelf as we see with software) was the basis for western civilization and its current glory which is engineering. In this context, “study” means a scope of top to bottom with a special emphasis on mathematics and how it has evolved. At the same time that I was doing my study, of course, the world changed into being heavily reliant upon computing, perhaps inordinately. And, watching this was part of my activity. The changes led to problematic issues of major proportion. This series will consider how KBE can help us handle our current dilemmas.
- AI or ML — After OpenAI’s release, I started to look at ML. Of course, issues were recognizable, immediately. After all, it was merely a continuation of the past two decades. I contacted my old employer (Larry Walker) at Sperry Univac’s Knowledge Systems Center (KSC). Larry filled me in on his computing history which goes back to the early systems in roles that were both technical and managerial. We both agreed that circumstance ruined a good thing. Actually, we are both of the mind that the evolution of AI would have been different had KSC continued its work. I had gone to Boeing after the KSC and discussed my experiences with KBE with Larry. Obviously, my work was a continuation with the different twist of being directly involved with engineering.
- Systems and methods for filtering and smoothing data, US7139674B2, Switlik and Klein