November 25, 2019 // By Paul Grizzaffi
In the Peabody’s Improbable History segments of The Adventures of Rocky and Bullwinkle and Friends, Mr. Peabody often directed, “Sherman, set the WABAC machine”. This WABAC machine, a computerized device for time travel, was clearly named as a play on the words “way back”, as in “a long time ago”. In each episode, Mr. Peabody and Sherman went back in history to visit previous historical figures and events. Due to a “quirk” in the machine, the situations were distorted in some way. It made for entertaining television as I watched the reruns at my grandma’s house after school.
Speaking of history, waaay back in the ’90s, I was fortunate enough to be selected for an internship at Oak Ridge National Laboratory (ORNL) in Oak Ridge, TN. There, I spent a semester working in, of all places, the chemical engineering department. Irvin Osborne-Lee and Juan Ferrada, my two advisors, were awesome to me. The taught me a lot about what users might want in an application and even more about how to be a professional (this may come as a shock to many of you, but I used to be a little rough around the edges). Also, I got to be a co-author with them on a paper about a project I was working on; I stumbled upon that paper on the web a while back and, recently, I got to thinking about it again.
So, about that project…Juan and Irv had a great interest in using computers in the aid of chemical engineering; could this be the genesis of my automation career? And the genesis of my speaking career? Perhaps, but I digress. As part of my duties, I was tasked with creating a software package that would help diagnose issues in rotatory systems. I was an undergrad in computer science; I didn’t even know what rotatory systems were, much less how to write software for them. Juan and Irv were very helpful in explaining what they wanted to see and how I might go about creating it. I’ll skip the gory details; you can read most of them in the full paper about hybrid AI systems that’s attached at the end of this blog post.
Why a hybrid AI system? Well, the hybrid here was a combination of a neural network and an expert system. The guidance was that each of these AI systems had a different strength and we should exploit the strength of each system to create a combined system that provided a better solution than either would have alone.
Why’d I write this post? Well, dated though it may be, I kinda felt like making this paper part of my “easily viewable” publications. Also, because Irv and Juan were so good to me as mentors and leaders; thank y’all!!! Don’t take the good ones for granted. In the middle of writing this, however, I found the parallel to some of my current automation beliefs: one tool may not be able to solve all of a specific organization’s needs and that’s OK.
Much as the WABAC machine distorted historical events, I expect that some of my memories are distorted as well; mea culpa. I also think our understanding of the application of AI tools may have aged to the point of distortion as well. What is not distorted, however, is that we should always use an appropriate tool for a job, even if it means we wind up with more than one tool.