Thursday, September 14, 2023

David E. Jakstis

David E. Jakstis and his support was seminal to the development of "truth engineering" which is twenty-three years old and becoming more apropos to the situation of computing than before. I will be getting into details as I expand upon the subject. But, first this:

David E. Jakstis  

  1 May 1952 - 13 Sep 2023   
Casting/Forging expert    
Boeing and Spirit Aerosystems

David Jakstis (LinkedIn) 



Over time, we will get into more details about the circumstances that brought David and I together. For now, we can briefly discuss a Knowledge Based Engineering (KBE) project whose accomplishments are apropos to evaluating the new world of AI. The past ten months have brought attention worldwide to the potential for computers to be smart. Though, lots of other reactions have been observed, many of these are not without a basis. Troubling reports come about daily, now. 

In the U.S., the business leaders were appealing in D.C. for discipline. Like kids with their hats in hand after mischief making (see older movies). And, not unlike the bankers sitting at a rogue's table in the context of the 2008 downturn. 

Part of the problem with computing is a lack of grounding in the true philosophical sense. We will get to that. David, on the other hand, worked in an environment that had to produce metal parts with defined properties. Skipping details, again, the era of thirty years ago was seeing lots of new approaches being done via computer with resulting issues causing people to tear their hair out. Now, the talk is of algorithms. Even worse issues, folks (take from an old guy who has been involved for a long time). 

Say, in AI, the "I" really relates to algorithms in action (very sophisticated ones, to boot). 

The context then was computer modeling across engineering disciplines and included aspects of physics and the necessary computational mathematics. David Jakstis and Bil Kinney had developed a means to generate a model of a forging die through specifying a few parameters in a graphical/textual mode. These inputs guided an "intelligent" approach on the computer that resulted in the geometry needed for a closed volume which represented a die (tool). That tool was then built and used in operation. The result was an entity that was very near to the net part. 

Meaning, some operations on the forging after it was created resulted in the part to be used. Skipping some detail, there were many steps in this process where outputs were not mathematically optimized. That is where my project came into the scope. Its title was "Multiple Surface Join and Offset," but that long name involved lots of different aspects of creating a usable computer model. 

In this type of affair, "truth" can have many meanings that are situationally determined. People can balance this type of thing; computers need more homogeneity. Heard of data science? One huge problem for that wish deals with data not being nicely configured for the operations that are desired. That topic will be addressed later. 

Similarly for casting, there were steps to create the model for the casting form to be used to create an almost-net part. There were common themes, as geometry was being handled by NURBS that are a standard type of modeling. But, the data differed as well as the process. Forging operations use heated ingots. Casting deals with control of flow and cooling. 

And, measurement was a common theme. Casting and forging provided plenty of examples with which to find problems, understand issues, and make changes to meet the overall goal. But, there are many other type of parts and materials. Needless to say, KBE methods became commonly used. 

And, the knowledge base approach, itself? This technique has been in use since the inception of the method. That will be discussed further. That is, lots of work has been done under the cover over the years. These deserve honorable mention. Too, lessons from these techniques need to be adopted. 

The larger picture is how do computers relate or ought to, to the world (phenomenon)? I will miss discussing this with David. In 2000, I gave him a white paper on the subject that I had written. After reading, he noted that it sounded like "truth engineering" which had the proper ring. So, the concept stuck. 

The reality of the situation? I can talk about twenty years of watching the world and the involvement of computing, across the board. Every five years brought more and more examples of problems being poorly understood. In fact, this recent set of events that has AI associated with it brought to bear several issues that are untenable without serious intervention. By whom? And how? There are many other questions. 

What David and I were working on decades ago can apply. From my perspective, there are many other  activities over the past two decades that need attention. A major change? The internet came to fore while David and I were doing the early work; it has matured enough to enable greater efforts than we could have imagined. 

As I proceed, I will regularly mention David's contributions as the basis for doing the proper analysis. 


Note: Work extending truth engineering in terms of computational modeling. 

Remarks:   Modified: 01/15/2024

09/18/2023 -- Added photo and link to ongoing work. 

09/21/2023 -- Added link to obituary.  

09/30/2023 -- David as an honorary member of the Thomas Gardner Society, Inc

11/11/2023 -- Using forging examples as a motivation for discussing the multi-pronged nature of truth. 

01/15/2024 -- To follow the work, see the TruthEng blog


Larry Walker said...

It is always interesting to read stories that show the integration of computing with other areas of expertise. I think of these as Expert Ssytems. The challenge lies in how the 'expertise' is extracted from the expert, so it can be represented on the computer for use. Both parties in the conversation have to be good listeners as they hear what the other side is saying. When mutual understanding takes place, wonders are produced.

AJSwtlk said...

Appreciate your support for the "integration" focus which applies across the board, almost (there are always exceptions ;>). In KBE, the "experts" were the developers in many instances. One might say that one appeal of the xNN/LLM is that the data has/have (either) the focus which goes back to the expert's (user's) realm.

In KBE, we took along important step: results were scrutinized and signed off by the expert. Thereby, the search for a solution was an interactive affair for the end user. Even if there were others assisting technically (engineering has no simple systems), the user made choices related to guidance along the way.

As opposed to the immature vision (my opinion of years of software integration work) of xNN/LLM where they seem to want to pursue "omniscience" or, even worse, oracle stature.