If your near-and-dear one was having a health issue, who would you go to?
• 𝘛𝘩𝘦 𝘣𝘦𝘴𝘵 𝘈𝘨𝘦𝘯𝘵𝘪𝘤 𝘈𝘐 𝘋𝘰𝘤𝘵𝘰𝘳 𝘰𝘶𝘵 𝘵𝘩𝘦𝘳𝘦
• 𝘈 "𝘷𝘪𝘣𝘦-𝘤𝘰𝘥𝘦𝘥" 𝘋𝘰𝘤𝘵𝘰𝘳
• 𝘈𝘯 𝘦𝘹𝘱𝘦𝘳𝘵, 𝘢𝘤𝘵𝘶𝘢𝘭 𝘥𝘰𝘤𝘵𝘰𝘳
• 𝘈𝘯 𝘢𝘤𝘵𝘶𝘢𝘭 𝘥𝘰𝘤𝘵𝘰𝘳 𝘸𝘪𝘵𝘩 𝘈𝘐 𝘵𝘰𝘰𝘭𝘴 𝘢𝘵 𝘵𝘩𝘦𝘪𝘳 𝘥𝘪𝘴𝘱𝘰𝘴𝘢𝘭
One could say this is an extreme case and perhaps not worth a comparison, but I want to drop this here as food for thought.
I work in the AI/ML domain and I see its potential. I am all for change and adoption, but I am not yet fully bought into a 100% replacement.
Checkout my article I wrote 2 year back on LLM -- Sandy’s take on LLM and RAG so Far
𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗗𝗲𝗯𝘁
I am somewhat starting to like the term "Technical Debt." Let me use myself and some of my colleagues as examples.
I have been using GitHub Copilot for almost six months now. In the early days, I would prompt it, refine it, and let it create entire repositories and solutions for me—my day-to-day tasks and everything else. Once in a while, things would go wrong, and it would take me days to fix them. Also, when I started looking closer, I realized there could be some unnecessary code blocks mixed in with some really smart ones.
After a learning curve, I now go function-by-function or block-by-block, and I have my own way of testing the accuracy of the outputs. The majority of my tasks involve processing large chunks of data and making inferences from them—sometimes processing and passing them to domain experts or top management for decision-making. In this case, I have to be triply sure of what I deliver, so I go step-by-step.
For sure, a week's worth of tasks can now be done in a day or two, and I can generate code that is much more scalable and reusable. The point is: I wouldn't let it run on "auto-pilot" for my entire task just yet.
𝗩𝗶𝗯𝗲 𝗖𝗼𝗱𝗶𝗻𝗴
"Vibe coding"—well, for sure, some have successfully done it. Having gone through it myself, I would classify those successes as 0.01% or even less. If you think about it, no matter the field or the task, you will always find some outliers—those who defy the norm. As for the majority (including myself), we either aren't sure what we are doing or need more practice with the tools.
I like to use the example of Excel a lot. Corporate employees know how to use Excel, but how much one achieves with it depends on their skills and the effort they took to master it. I remember in my Quantitative Finance course, I was doing heavy Python coding for some bond pricing, and the instructor—an expert—did it in Excel right then and there with us. It's the same with my cousin; in five minutes or so, he made an entire loan repayment and amortization sheet for me in Excel.
LLM tools and agents are getting much smarter and faster, but without knowing the basics of the task at hand, things might spiral out of control, and that "Technical Debt" would keep on growing.
𝗜𝘁 𝗶𝘀 𝗔𝗹𝗹 𝗔𝗯𝗼𝘂𝘁 𝗡𝗮𝗿𝗿𝗮𝘁𝗶𝘃𝗲𝘀
Apple, now Anthropic—even recent politics and whatnot—throughout history, it has always been about the narratives one sets and how fast people catch up to them. Yes, you then need a product to support it.
This reminds me of Freedom 251 (India). It was advertised as a smartphone for ₹251. It looked like a scam, but the narrative it set got tons of bookings based on that story alone.
𝗧𝗵𝗲 𝗘𝗰𝗼𝗻𝗼𝗺𝗶𝗰 𝗟𝗼𝗼𝗽
I read this somewhere—imagine AI and robots can do everything and take over most jobs. If people don't have jobs, they have no income to buy stuff—both essentials and non-essentials. If that happens, who will these AI companies sell to? Who will buy the robots?
It is said that an equilibrium will be reached, but one cannot expect everything to go "all-in" in an instant. The world runs on consumerism.
Lastly, in one of his interviews, I heard Jamie Dimon, JP Morgan Chase CEO, saying (based on what I recollect):
We have autonomous driving—does that mean you take 2 million drivers out of work and the next best job they have pays only $25,000 a year? No, you can do it gradually, or have the government pitch in to say, "No, you can't do that," or "Let us do it sensibly."
Anyway, don’t get me wrong—I am all for AI, the change, and the new ways of working, as well as the new skills and job openings that will be brought about by it.
Thinking that AI can do it all? I am still not sold on it.
