Analytics replacing AI?

Back in the 1980s, it appeared that the most advanced software applications were increasingly becoming the province of expert system shells -- i.e., of rule-based application development  tools.   The conventional wisdom is that this technology crashed and burned, but may be making a modest comeback now. 

On closer examination, however, a different pattern emerges -- what looked like expert system applications in the 1980s are largely turning out to be analytic applications today. 

Here are some examples:

1.  Predictive analytics -- data mining, etc. -- are an outgrowth of machine learning, which basically started as an attempt to automatically discover rules that could not easily (if at all) be extracted from human "experts".

2.  Modern dashboards look a lot like the famed STEAMER system, which was essentially the inspiration for the product design and core demo for Intellicorp's KEE.

3.  The best success stories for expert system technology were primarily in application areas being addressed by predictive analytics or other analytic technolgies today.  Some of the biggies were::

• Airline seat pricing -- United Airlines claimed $100 million in incremental profit from an expert system way back in 1984. Today there are pricing applications, with the key underlying technology being data mining.  (More on that soon.)
• Credit granting -- Inference's application for American Express was the consensus choice for Best Third-Party Expert System Ever.   Today, of course, antifraud is heavily based on data mining.
• Homeland security -- By the mid 1980s, the Secret Service and DEA had useful tools for tracking wannabe assassins and small plane flights over Florida, respectively.  Similar applications today are based on reporting and predictive analytic technologies.