Issue #134 | Feb. 15, 2008 | by Oliver Hedgepeth
In the mid-1980s, during the gold rush days of Artificial Intelligence, AI and the expert systems software were considered the magic bullets that would solve all our data problems with an increasing avalanche of data coming from automated sources.
At least, that was the view of many Army generals who came calling on the Army’s AI Centers or Knowledge Engineering Groups, and one USAF logistics AI Center. The Army’s logistics and transportation data base experts were selling the next AI systems as a replacement for the old COBOL data crunching software and the old HASP operating systems.
The problem was we were building homeostatic solutions for problems the generals did not know they had. The sad, but true, fact was they would not know of these problems for several years in the future.
There were only a few books on AI and expert systems and only one annual conference. Now, I sit here looking out over the landscape of books that have sprouted like wild orchids in the tropics. The RFID industry knows there is a short season to get your book into print before the public realizes you are just writing the same thing as the other books. But, like AI and expert systems, RFID is marching full speed with solutions into the work place.
AI lives today, all around you. But, we don’t use the term AI in that way anymore; in fact, that was the whole point when we started the military AI movement – get people thinking about new solutions, call it AI, and then fade the term away, as new software innovations emerged. It seems to have worked.
The real issue of AI was elephantine data bases. How do you get all that data for logistics systems around the world into the hands of the right decision maker in your company? The global data market today is causing many of the early AI pioneers to rethink what we did to bridge data mining, data searching, and data storage issues.
We have lots of storage today, but what data do you store? Do you know what metric to use to capture that data? The old metrics are not going to provide you with an answer. The data volume is growing at light speed. Yes, I know you have heard this story before. But, how do you approach the solution of aggregating billions of data pieces streaming at you daily, that you have now really decided to keep somewhere, on the off chance you may need them.
The RFID data issue today is a panoply of deciding what course to choose, with that data. It is how to measure that data. How to develop that yet unknown metric that will merge with or stand beside the metrics taught in accounting classes and supply chain management courses. How to live with the behaviors of so much data after you realize it will never stop coming, nor slow down.
Maybe it’s worthwhile rethinking the six-step approach we used then; but the point is to better understand the metrics needed for this rush of RFID data. At that time, this was our approach to developing metrics:
- Define the measure
- Decide the dimensions of the measure
- Know the limits of the range of this measure
- Understand the rationale for this measure
- Know the decisional relevance of even having this simple metric
- Know what other metrics were associated with this lot estimate metric
Sounds simple, but try to live with each one of these in equal portion; challenge your assumptions as you think through each metrics based on these six steps.
Everything old is new again.
Oliver Hedgepeth, Ph.D., afwoh@cbpp.uaa.alaska.edu is author of RFID Metrics published by CRC Press and former Director of the AI Center for Army Logistics.
