The Acting Commissioner of Social Security recently made the decision that the agency will, for now, have no involvement with Generative Artificial Intelligence (AI). While I have a healthy skepticism that AI will ever have much role in taking or adjudicating Social Security claims, there's one area where Generative AI may be of considerable assistance and that's dealing with the COBOL problem. Social Security's most important computer programs are written in the ancient COBOL programming language. Many other agencies and private companies are similarly dependent upon COBOL. Schools are no longer teaching COBOL. Computer programmers don't want to work with it. Few programmers are available to work on COBOL programs and those programmers have long since gone gray.
IBM, and probably others, have decided that Generative AI may be the cure for COBOL problems. There would be no need to laboriously rewrite all the old programs in newer programming languages. AI would quickly rewrite the old COBOL programs in modern computer programming languages. Programmers would be more widely available for modern computer languages. Corrections and improvements to existing programs would become easier. That would be a godsend to Social Security, if it works.
Retrain those with disabilities to program in COBOL.
ReplyDeleteI know thats not popular, but somewhere in the millions and millions and millions of people on DIB, somebody has to be able to do it.
Given that Chat GPT has passed a number of bar exams and some medical licensing exams, your skepticism about AI never having much of a role in taking or adjudicating Social Security claims may not be justified.
ReplyDeleteOn the other hand, from what I've read and from some personal experience, AI does make programmers much more efficient and using it to help convert the old COBOL programs seems doable. I am guessing that SSA's contract programmers are already making extensive use of AI.
Could AI be a tool? Sure. But put COBOL in on the left and java or upgraded COBOL comes out on the right ain't happening soon. COBOL is still around because its so good at what it does. A blessing and a curse.
ReplyDeleteFrom Ars Technica -
But there's a recurring argument that COBOL is good at managing business-specific systems and exchanges in ways that (some might argue) present fewer attack vectors. Or you might argue that AI-generated and restructured code might look proper and seem test-ready, but without the people around who know exactly why the code does the things it does, AI-upscaled code could have just as much noise as AI-upscaled video.
ReplyDeleteMicrosoft Excel regards 1900 as a leap year for compatibility reasons. 1900 was not a leap year. Relevance? AI has to understand the program as a whole to understand the elements of each job. This is one of COBOL's key failures. The inability to assess the part as a whole.
Sure AI maybe able to assist in interpreting and scouring through it, but while it works, it will be used. Some 40 percent of banks still use it. Until that decreases and a clear mandate from Congress and/or the President comes through (with ample funding), COBOL is here to stay in government, AI be damned.