Like many right here, I’m a programmer. I studied programming from 2003 to 2008, though earlier than that I used to be already drawn to tech, coding, and related pursuits.
Earlier than diving into my reasoning, I need to make clear just a few factors.
Chapter Zero: Defining Two Key Phrases
First — and crucial — level: Stupidity is whenever you didn’t know what was alleged to be there within the first place.
Think about you stroll right into a home to hold your keys on a hook… besides the hook isn’t there as a result of somebody eliminated it. Your keys land on the ground and also you look silly. Not too hurtful — and later it’s sort of humorous when you understand you didn’t know the hook had been moved. However one can provide a extra expensive instance: standing at a automotive service middle, listening to somebody telling you your automotive is wrecked since you stuffed a diesel engine with gasoline — you didn’t even know your engine was petrol-powered. That’s costly, painful stupidity. Each time you’ve ever felt silly, it’s as a result of somebody assumed one thing, and also you didn’t know.
Second, equally vital level: One thing solely counts as working if it features by itself, with out supervision. You shouldn’t must prop it up, nudge it, patch it, or grease it so it runs.
Sure, most machines require upkeep — like altering oil or working antivirus on a server — however these are routine recognized actions. If to print a web page on a “new” printer you have to open a brand new app, click on “Exit” 3 times, change to the outdated app, then reopen the brand new app — that isn’t a working utility. It’s damaged.
Then — right here’s one thing: my diploma says I’m a software program engineer. The phrase “engineer” implies I can invent, construct, keep, and handle software program and {hardware}. I can set one thing up, create one thing new, or take away one thing unneeded. Whether or not it’s a mouse driver for Home windows 95 or a distributed system on Kubernetes — I can deal with it. If I don’t understand how, I’ve the instruments to search out the directions, be taught, perceive — after which I’ll sort out the mouse driver or Kubernetes.
Lastly — the primary level: A pc is a tool that gathers, shops, processes, and transmits info. Processing means executing a exact sequence of instructions encoded into the machine.
Chapter One: What Programming Really Is
Programming is a precise science. I can exactly predict what’s going to occur when a command runs, and I can estimate system reliability inside outlined possibilities. There are well-known strategies to construct fault‑tolerant methods.
We are able to construct methods the place any node might be swapped on the fly. Actual‑time methods assure a end result inside a strictly bounded interval — utilized in rockets, plane, telecom stations. Banking methods deal with numbers with excessive precision to trace balances and compute charges so precisely that purchasers don’t discover.
Computer systems can repeat the identical job infinitely with out deviation — and that’s why we’d like them. No person needs to take a seat in Excel and add up numbers manually. In actual fact, the time period “laptop” initially referred to individuals doing arithmetic calculations for house and army companies within the Fifties.
Chapter Two: And Right here Comes… AI!
It’s 2025. We now have so‑referred to as “synthetic intelligence.”
It’s touted as an unbelievable system — as a result of it does issues not like conventional computer systems. Underneath the hood, it’s largely matrix multiplications with random noise added to responses, letting us shuffle information in novel methods.
However right here’s the difficulty: introducing uncertainty into a pc’s output basically violates the definition of “laptop.” A pc ought to all the time give a precise reply.
I’m seeing large sums poured into AI — actually a black gap of hype drawing in {dollars}. It’s overhyped to absurd ranges. NVIDIA went from making GPUs for gaming, to promoting to cryptocurrency miners — and now AI people throw cash at them nonstop. The NVIDIA AI processor is $35,000 a chunk. Bulk consumers pay as much as 150,000 card in a single lot.
Power? That’s one other nightmare. Within the U.S., buyers are dumping money into energy crops simply to feed AI farms. I’m useless severe. I can assist — when you have an organization that builds energy crops — contact me. We have to spend $250 billion by yr’s finish.
All that’s to maintain AI working.
Which is okay! NVIDIA hasn’t collapsed — the market is booming.
However the query stays: What’s the level of AI? What’s the aim? What’s the precise output?
Certain, it’s apparent: AI helps course of giant quantities of knowledge quick. That’s nice. I don’t have to jot down code by hand — bots do it for me.
However the issue: no AI mannequin can ever assure 100 % correctness. That severely limits its use in programming. Test the dialogue within the subsequent thread — we are saying that irrespective of how effectively you engineer it, ultimately, one thing’s gonna go haywire and it begins spitting nonsense. Your AI assistant isn’t a working system. It wants care.
When working with AI, you need to strike a exact stability. It’s worthwhile to know precisely how a lot work you possibly can offload and the way a lot you need to do your self. If you happen to ignore AI, you’ll be too sluggish. If you happen to overuse it, you’ll waste hours debugging AI‑generated junk.
Putting that stability can notably velocity up work. How a lot relies upon. However you possibly can velocity up.
The hazard is rampantly over counting on AI — it’ll rapidly result in absolute stupidity. Mistaking rm ./* -rf (A command that deletes all of the recordsdata in a folder) for rm . /* -rf (a command that deletes all of the recordsdata in your laptop) could be deadly. Misunderstanding AI‑generated instructions can value jobs. Manufacturing will get deployed irrespective of how messy. AI‑written code goes stay unreviewed.
I see this from the banking world. While you recommend utilizing AI, bankers take a look at you want this 😒.
Chapter Three: Who Truly Wants AI?
Think about a monetary accounting system written within the Nineteen Eighties. Nobody modifications it — there’s no want. What issues is that it produces in 2025 the identical outputs it did in 1990.
Plus, AI methods suck at any sort of specialised programming.
Apparently, AI methods excel at very slim duties — predictive modeling and large information dealing with. Take a look at Google’s climate‑prediction fashions. “Predicting.” We nonetheless aren’t certain of the outcomes. The query “Will there be a hurricane in Florida?” issues. For now, it’s 50/50 — both it occurs or not. We don’t know.
However right here come LLMs and large‑information analytics: we will supply extra nuanced — however nonetheless probabilistic — insights.
AI textual content processing is beneficial. “Rewrite this textual content and substitute all ‘you’ in casual kind with formal ‘you’.” Straightforward. “Rewrite in an archaic fashion.” Performed. You then simply proofread it. It’s serviceable.
LLMs work effectively for duties the place actual precision isn’t required. You may open Phrase and do a world discover‑substitute — however the end result could be terrible. Because of LLMs, we will substitute “you” to “you” with 99.95 % accuracy.
Translations? Certain, should you don’t want an ideal literary translation. The LLM would possibly use incorrect shades or idioms, but it is going to be higher than any google-translate you may get.
That saves hours making an attempt to decipher a bizarre translation. Nevertheless it’s ineffective whenever you want a movie translation or a poetic rendering.
Chapter 4: The Human Issue
The toughest hit zone is the human issue. Suppose you run a enterprise and wish a programmer. You put up a job itemizing someplace and use an LLM to filter resumes.
If you happen to solely have three programmers, you received’t see the fourth rapidly. However when you have ten thousand candidates and 1,500 resumes per day per job, you possibly can’t rent with out AI.
Right here’s the factor: any “no” that AI sends on somebody’s resume has some official purpose. You’ll be able to’t simply rent by way of AI. It’s worthwhile to discuss to the candidate in individual.
That’s whenever you be taught: AI would possibly admit the worst monster you’d by no means need to work with — or reject a superb one who doesn’t memorize excessive‑stage theoretical math.
Conclusion
What they are saying: AI is a gold mine with no draw back.
Actuality: AI is simply one other expertise you need to examine, know, and be taught to use.
Any concept, nevertheless good, will get ruined if used too little or an excessive amount of. AI all over the place is stupidity. AI as soon as a month in ChatGPT is falling behind.
Don’t fall for each advertising pitch. As a substitute, ask for gross sales charts and earnings figures from these corporations pushing “AI have to be all over the place.”