I recently saw an IBM presentation on how AI will help eliminate COBOL once and for all. Basically, they are trying to sell a tool that uses their WatsonX cloud service to convert COBOL to Java with the help of the current buzzword named generative AI. Let me tell you, COBOL has outlasted most of the tech that was invented since 1965 and I think it will outlast generative AI, too.
@nixCraft Awesome. AI is going to fail so big. If they have to double down from "let's replace search engines", "cure cancer" and "make human artists obsolete" to "let's replace COBOL", there is no limit to what they will promise.
@nixCraft There are legacy systems out there that use COBOL or MUMPS or Modula II or some qother supposedly dead languages
They run on obsolete hardware
And people pay a fortune for tech bods with the skills to maintain them
They do this because, often despite many attempts, they've failed to develop a modern system that will process, or sometimes even access, the data & give the same output
I will try not to point & laugh when AI also fails to solve the issue
@nixCraft It was offered at the college I went to and some just begged and begged me to tutor them until I relented. I didn't know COBOL at all but I was able to help them through...at great personal cost. Migraine is putting it lightly and it lasted for hours after.
It's not particularly hard...it's just tedious and I think you can only like it if you like filling out the original punch cards it was used for but on a screen. AI deserves it. I'm not sorry. I hope it can feel pain.
I recently saw a presentation where they said that Generative AI will solve the mainframe-COBOL problem for us by explaining what COBOL programs do, line by line, in plain English.
I'm like, ... "Excuse me, ... Have you looked at the COBOL program you're translating to 'plain English sentences'? What do you think a COBOL program *IS*?" (... and was always designed to be.)
The inability to distinguish between totally different systems means: They don't know 'man command' They are inept to use DuckDuckGo They can't write a 'hello world' in a new language
@nixCraft if it was an investor presentation then it makes perfect sense. Hype up the money coming in, hand pick a few test cases, put a case study on your website, then get more clients.
Once the contract is signed, hope sunk cost fallacy prevails, along with a small dose of can't turn back now.
Running systems in parallel never happens due to a "replacement" project, right?
@nixCraft Java is the next COBOL, so it should be skipped entirely. BTW, there is nothing wrong with COBOL, but there is profit in convincing people there is.