'We have to reject that with every fiber of our being': DeSantis emerges as a chief AI skeptic

[DallasNE]

8. I Am A Retired Mainframe Computer Professional

Back in the mid-1990's, I worked on a phone directory assistance application. The backbone was a product called "Soundex", and what it did was to convert the word string the operator keyed in from alpha to numeric. I don't recall the percent variance that was the cutoff point, but for the sake of argument, let's say it was within a 95% match. Let's say they wanted the phone number for William Jensen. It had the intelligence to know that William and Bill would be a high value match, and the same for Jensen, Jansen Jenson, Janson, Jesson, etc. It would do the same with the street address.The operator might say we don't have an exact match at that address but we show an E William Jansen at the street address and the person would say yes and get the phone number. I'm guessing that AI is just a more robust Soundex. That would sometimes return odd results, too. Two different strings of words could convert to non-related "matches". For instance, it might think that "tractor tires" and "brothel services" were a close match based on the numerical conversion.

I can see how AI is a nice tool for apps like my directory assistance example and other similar uses, such as search engines, but how to you convert your legacy system for business-specific reporting requirements, such as a company's monthly report of operations based on a specific Chart of Accounts? I don't see how AI is useful in a situation such as this.