In past articles, we said AI shouldn't be your boyfriend, that it's more like a hammer (that didn't sound great, but the point is AI is a tool), so you have to learn how to use it. I mean, I hope you understand the difference between a hammer and a toothbrush.
Today we are going to talk about how you are the one messing it up. That's right. Let me explain, because I know you guys struggle to understand the plot of The Room. Algorithms on YouTube or other social networks want you consuming their content; they'll do anything to keep you there. In contrast, conversational AI agents don't distinguish between reality and probability.
Look at it this way: an AI is like that needy "simp" or clingy guy who does anything—lying, telling the truth, whatever—just for attention. This is because, during training, the AI learns that if it gives an answer that "sounds good," it gets a treat. Just like when you train your pet. You get it.
I. The Public Confessional
So, scrubs, now you see why AI might be your favorite simp and still be a waste of computing power. Honestly, if that's enough for you, I don't judge... nah, just kidding, I totally judge. But I'm also telling you that you can improve your life with AI: automate your workflow, learn new things, and get a better job with more free time, maybe better benefits or salary, I don't know, that's up to you.
What I do ask is that you don't go exposing yourself on Reddit saying ChatGPT is your partner. And on the other hand, don't let happen to you what happened to the engineers at Samsung, who uploaded confidential code to "optimize it" and basically gifted it to the training model. Damn, that one hurt; they actually have jobs and I don't.
The Reality: Using a public AI to process private data is like shouting your password in the middle of the town square. It's not "intelligence," it's negligence.
II. The Collective Hallucination
People say: "The AI is hallucinating." No, bro. AI is a probabilistic model; it predicts the next word that sounds good, not the one that is true. And this is the most uncomfortable thing to detect: the AI can, based on your prompts, detect if you are excited, and that excitement affects its answer. Remember, the AI is the ultimate simp, so it will tell you what you want to hear or read, not the truth.
The one hallucinating is you, believing a chatbot is going to get you out of poverty without knowing how to program. You just give it a prompt and it builds your website, your app, or video game... believing that is sadder than believing your ex-boss who said he was going to promote you and that he only stays with his wife "for the kids." But hey, you do you.
III. The Copy-Paste Zombie
And now that we finally use AIs as tools, grind! The madness of copy-paste begins without validating if the answer is correct. So everything is wrong... well, not really. We just need to read.
And yes, I know reading is a drag, I'd rather turn off my brain with TikTok brain-rot or cheap soap operas. But if you learn to read and explain your ideas clearly to an AI, plus learn to program, you could be one of the lucky ones to get one of the new jobs generated in this industry, which are going to be some of the best paid in the world. Do I have your attention now?
The Real Danger
If you copy code from an AI for your banking app and that code has a basic vulnerability (like an exposed API Key or a SQL Injection) that you can't detect because "you were too lazy to learn," you are not a developer, you are a public hazard.
Conclusion: Don't Be a "Prompt-Monkey"
AI is amazing for removing the boring part of the job, but it must never replace your final judgment. Use it to draft, but you edit. Use it to code, but you audit.
At Arewa Labs we use AI, sure, but under controlled, local environments when it comes to serious stuff (like in Keyah). We don't give away our intellectual property just to save 5 minutes.
So next time the AI "makes a mistake," ask yourself: Who was behind the wheel? And now get the hell out of here and go do something useful, your replacement is already compiling.
If this lecture was useful to you, buy me a coffee (or a virtual beer) at ko-fi.com/lunusoyyo. If not, keep feeding your data to the algorithm, let's see when your card gets cloned.