AI Is a Backhoe, Not a Magic Wand

It’s been a couple of years since May 23, 2023, when I posted my first rant about AI in general and ChatGPT (properly pronounced “chat-jippity”) in particular. In that blog, I compared AI to a calculator: In the hands of a novice, a scientific calculator is far more likely to be used to spell out “fart” than it is to find the proof to Fermat’s theorem. (Use hex mode, put in “46415254”. How I know this is a story for a different day.) Like the calculator, the spreadsheet, and the internet, AI-driven LLM tools are likely to change HOW we do our work, but not the fact that humans will still be the ones doing the work in the first place. Today I want to offer another perspective, inspired by a LinkedIn post (you can find it here) by the inimitable Dave Mangot. He started by referencing someone who experienced the consequences of their poor choices, when it came to “vibe coding” with an AI: Then Dave followed up with this story: A long time ago I started at a new job and everyone was logging to the Unix hosts as “root”. I asked a friend of mine if I was just out of the loop and this was an ok thing now. He replied “You know what we call people that do that? Amateurs”. Dave continued with his own observation: Despite what all the AI vendors tell you, it’s not there yet and no one really knows when it will be. I’ve spend the last few days coding with an AI Assistant and it’s great! I love it. Never going back. But it also: wanted to put the DB password in plaintext in the code wanted to hardcode the DB config in the code told me the wrong module name for requirements.txt I know enough to instruct the Assistant otherwise so it’s fine. >Building a kit car (prototype) you ordered through the post does not mean you are ready to join an F1 team. Maybe someday. So as I pondered all of this, another analogy occurred to me: The shovel vs the backhoe Both a shovel and a backhoe do the same job – they dig holes. The backhoe is certainly faster and more powerful, but in the hands of a novice a shovel can only cause a relatively small amount of trouble, while a backhoe can cause much bigger headaches (both literally and figuratively). While both could, conceivably dig up a rock, or the sprinkler system, or the main internet feed to your business, the backhoe is much more likely to do so before the operator is aware of what’s happening. Just to be clear, my point isn’t that automated tools are inherently bad. My point is that, IN THE HANDS OF AN INEXPERIENCED PERSON, an automated tool is much more likely to cause trouble and damage instead of doing good. This is due to the fact that the automated tool (whether the construction or computing kind) requires experience and expertise to be effective. It’s a labor-saving device only insofar as the operator understands the labor to be done. In the intervening 2 years since my original AI post, I have heard from many colleagues who’ve used AI tools to great effect. But I’ve seen far more randos who think they can hop up in the cab of the shiny AI-driven excavator and dig a foundation in an afternoon, who only end up causing destruction and confusion.