When I first watched the Hollywood movie “Arrival”, a thought that crossed my mind was -“Could they have leveraged AI to decipher the ideographic language of the aliens faster”? For those who have not watched this movie, a significant portion of the movie is focused on a team of experts trying to decipher the language of aliens who are visiting our planet.

The language has ideographic origins, meaning a language in which the printed symbols represent entire words without relating to pronunciation. After all, a few years ago, MIT researchers claimed that they had developed a model that could decipher languages that have been “undeciphered” so far. But obviously, all that thought process I was going through while watching the movie was hypothetical. After all, the movie was a work of fiction.

I read the article “The World Depends on 60-Year-Old Code No One Knows Anymore” this morning, which pertained to deciphering another type of language- a programming language. The gist of the article concerns a challenge I once wrote about on LinkedIn in 2020. That was when the state of NJ struggled to find COBOL programmers during the pandemic to help maintain and troubleshoot their legacy systems. Looks like the problem not only still persists but has also increased in magnitude.

The public sector invests Billions in IT consulting fees every year. I wonder why no one ever questions why none of these consultants have been able to propose a plan to the state of NJ or other states to transition away. But as the article shows, it is not just the public sector. As the article’s opening lines suggest, “Every day, 3 trillion dollars worth of transactions are handled by a 64-year-old programming language that hardly anybody knows anymore.” The problem not only persists but has a sense of urgency around it.

The short-term (relatively) solution is similar to what is suggested in the article: rather than addressing the shortage of people who know how to code in the archaic language, why not teach this language to an AI model? The AI (Specifically, IBM Watson) solves the immediate skill shortage problem. However, looks like the way it has been explained in the article, Watson will run into some limitations. I agree that a “code assistant” approach might not work. But the good news is that it is a programming language; it is possible to develop a Generative AI model that is very proficient in the language. Not perfect, but then even humans are not perfect. Maybe Watson can’t do it, but building an AI model is feasible since programming is so rule-defined. And such a model can be trained relatively faster as well.

But that will not solve the core issue.

An AI model will also not be perfect. However, it will constantly improve the current situation and help address the skill gap significantly. The core challenge is that systems are still dependent on a language that is half a century old. As mentioned, I wonder why no technology consultant has flagged this and helped concerned entities transition to newer technologies. Note that the legacy aspect is not just about the language, but there are other aspects, like the inflexibility of the mainframe platforms. In today’s era, open-source and cloud-based alternatives to the mainframe platform are viable for almost all impacted industries. And addressing that is the proper long-term fix.

So, while building a Generative AI model that can code in COBOL is realistic, it will be a short-term fix, like the many patches that COBOL-driven legacy systems have seen over decades. What is needed is a comprehensive plan for each industry, specifically the critical ones like financial services. The reason that has been delayed is that while fixes and troubleshooting are technical projects that are easy to plan and execute (and earn consultants easy money), planning a large-scale initiative like this means significant strategic planning beyond the technology execution or configuration. That is not easy and is not immediate money for the consultants.

While this article does not intend to get into the details, any organization that genuinely wants to transition away from these legacy systems can do so without any significant impact on the current operations. Yes, it will be time and resource-intensive (people and $$$), but they can re-architecture in a way that will ensure that a language going obsolete will not put them at risk of the failure of the core flows in their organization.

The clock is ticking!