Y2K, AI, and other matters of technical debt
Big confession ==> I saw a problem for society and I chose to remain silent. The year is 1992 and I am looking at the 2-digit date fields in source code for Assembler and COBOL programs that were 20 years old at that time. The programs were running critical applications for the company that I worked for and I knew they would be inoperable as soon as the calendar year flipped to 2000. But I chose to ignore the red flag because, heck, 2000 was 8 years away. Do you know what you’ll be doing 8 years from now? Right, you don’t. Plus, no one else was even talking about the Y2K issue at the time. https://time.com/archive/6955043/the-history-and-the-hype-2/
Was I smarter than everyone else? Not by a long shot. I was just good at math and I also read computer programs like they were stories. What are they trying to do? Why are they employing certain routines? Why are they manipulating certain fields?
So, I noticed things.
Y2K is probably the most egregious example of technical debt because it was so enormously expensive to fix. When the bills came in, it cost the world between $300B and $600B. https://americanhistory.si.edu/collections/object-groups/y2k
Who cares, Peter. Ancient history. The world has moved on.
It is ancient but I can see errors being repeated today. AI being dropped into companies to supplant existing systems. Or AI being deployed without a lot of strategy and testing. Or AI being treated as atomized units, as things that can be treated in isolation.
Where or where did all the QA people go?
Read this excellent article on technical debt from MIT Sloan. Don’t bring it up over drinks with your friends. But if you want to demonstrate you have some value as an thinking employee of the future, do bring it up at your next business meeting.
We've all read that article as well. Part of our challenge with an AI first product is to guide our customers on how best to use this new technology. It's not like traditional software.