Monday, February 12, 2024

A year ago

TL;DR -- A year ago, the world was three months past the release of ChatGPT with millions signing up. What did that event pertail? We spent several months reviewing the various options as well as looking again at the history from Dartmouth onward. In this new year, one can sense a change. Will there be less hype? 


By this time last year, we were catching up on machine learning. Much later, we wrote the first of a series: Artificial intelligence, not solely machine learning (or AI, not solely ML) We put a "Papers" folder to publish articles with respect to xNN/LLM. That's refers to the neural net approach plus the large language model. John was involved with advanced computing his whole career and did lots of algorithmic work including those related to artificial intelligence and to what was then known as evolutionary computing including machine learning. His finals decades were oriented toward modeling with a focus on engineering support.   

So the following is a recap of the past year. After finding out about ChatGPT, we started discussions of the Director, Larry L. Walker, of the Sperry Univac Knowledge Systems Center (KSC). After some reflection, we planned a series of articles that would cover the history of computing and the related mathematics. The first thing that we published dealt with the decisions behind the KSC and with its accomplishments. The title was: Sperry Univac Pioneers Application of Artificial Intelligence – 1985-1987. 

After that, Larry wrote of his experiences with computing from the early days of Control Data and to his knowledge systems (expert systems) work. Meanwhile, Larry and John discussed the technology involved with Knowledge-based engineering (KBE) which was central to Boeing's 777 program. Throughout the year, the research looked into what happened with KBE at Boeing, in subsequent years, and in the world, in general. 

We have had some posts here on these subjects which we will collect. Late last year, the ACM Communications had an article by Thomas Haigh that looked at the history of AI and argued that there had been no AI Winter up to 1980 though the popular press seems to say that there was. In this regard, Larry talked of his having seen the AI winter of 1990. In the latest ACM Comm, Prof. Haigh published another article. This one is titled How the AI Boom Went Bust which agrees with Larry's assessment. 

So, letting John chime in, he has seen no AI Winter. KBE? It is still part of some CAD systems. Now, we can argue that it is not to the expectation that was there in the 1990s. John would agree. However, in doing a quick survey, there are many other names that KBE continued as. The theme with that is to not associate with a failed system. But, digging deeply, most of the continuals that John saw reference KBE as being in their history. 

Personally, for him? He worked KBE from the late 1980s until 2005 when he retired. Then, his focus became truth engineering which was spawned at Boeing and has continued in a research mode. Now, AI was created by mathematicians one of whom (John McCarthy) coined the term. Dartmouth was the site of a meeting on the subject in the 1950s. And, all along since the 1950s, there have been periods of excitement as people contemplate some creature emerging from the fog of computing ala Science Fiction's take on the matter. 

Being adult about the situation? A closer look, under the covers, shows the mathematics at play. Yes, it is sophisticated. Taming the hype cycles will require some maturity. Part of that will be grasping that mathematics and bringing it out to public view. 

Since that first paper on AI and ML, there have been two more which present the important factors that need to be discussed. A fourth one is in the works. After that, the focus will be on a discussion that is practical in scope. John will offer one example which comes from the KBE work. 

Essentially, some in the discipline of data science have been traveling the right path. Actually, there are many examples of computational systems providing benefits via services that we can respect as they offer results that have value. 

The theme of utmost importance will be determined by people as the technology progress. And, the universal themes seems to have been conjured up through the efforts of a few. Local themes have been given little attention. AI gives us a chance for a better balance; in fact, such a balance would improve all humans affairs, including science. 

Remarks: Modified: 02/13/2024

02/13/2024 --  Added image. 

No comments:

Post a Comment