There are two types of people who use chatbots.
The first type asks a normal question, gets an answer, says "thanks," and leaves.
The second type opens a chatbot at 2:13 AM and types:
"What would happen if I put a toaster in a bathtub?"
If you're the second type, welcome.
But even though chatbots are basically designed to sit there and tolerate whatever nonsense you throw at them, there are still some things you probably shouldn't type into that little box.
Not because the chatbot has feelings.
Although, depending on the chatbot, it might act like it does.
1. Your password
Please don't type:
"Hey chatbot, my password is..."
No.
Stop.
Close the keyboard.
Go outside.
You don't need AI to judge your password when the password itself is probably your birthday followed by 123.
If you want to know whether a password is strong, describe its characteristics without giving the actual password.
For example:
"Is a 16-character password with uppercase letters, numbers and symbols generally considered strong?"
That's enough.
The chatbot doesn't need the keys to your email account.
2. Your entire life story
A chatbot doesn't need your full autobiography to answer:
"Why did my friend leave me on read?"
You probably don't need to provide:
their full name
their date of birth
their address
their mother's name
screenshots from 2017
the history of your friendship since kindergarten
Just explain what happened.
The chatbot needs context.
It doesn't need to become a detective.
3. “Don't make it sound like AI.”
This is one of the funniest things people say to AI.
Then they immediately ask:
"Can you make it sound more human?"
The chatbot responds with:
"Absolutely! Here's a more natural and engaging version..."
And congratulations.
It now sounds even more like AI.
If you want something to sound natural, don't just say "make it human."
Tell it what you actually mean.
Try:
"Make this sound like something a normal person would actually text. Remove the corporate words and don't make every sentence perfectly polished."
Much better.
4. “Write something interesting.”
This prompt is basically walking into a restaurant and saying:
"Give me food."
What food?
How much?
Spicy?
Sweet?
Do you hate onions?
The same problem happens with chatbots.
If you ask:
"Write a blog about AI."
you're going to get approximately 14,000 words about how AI is "transforming our lives."
Nobody asked for another article beginning with:
"In today's rapidly evolving digital landscape..."
Please.
Tell the chatbot what you actually want.
5. Don't tell it to “be funny” and expect a miracle
This is where things get dangerous.
You say:
"Tell me something funny."
The chatbot gives you:
Why did the computer go to the doctor?
Because it had a virus!
You stare at the screen.
The screen stares back.
Your faith in technology decreases by 4%.
Humor isn't just about putting a joke at the end of a sentence.
Good humor needs timing, context, personality, and sometimes the ability to recognize that the joke is terrible.
That's why a chatbot with an actual personality can feel very different from one that simply follows the instruction:
"Insert joke here."
6. “Are you sure?”
This one is actually useful.
If a chatbot gives you an answer that sounds suspiciously confident, ask:
"Are you sure?"
But don't stop there.
Ask:
"What makes you think that?"
or:
"What part of this could be wrong?"
AI can produce an answer that looks incredibly convincing while still getting something wrong.
The dangerous part isn't always the mistake.
It's the confidence.
A wrong answer wearing a suit is still a wrong answer.
7. Don't paste private information just because you're lazy
Let's say you have a document with your:
name + phone number + address + email + ID number + random confidential information
and you want the chatbot to summarize it.
You probably don't need to paste all of that.
Replace private information with placeholders.
Instead of:
"My name is [real name] and my address is..."
use:
"[NAME] lives at [ADDRESS]."
The chatbot can still understand the structure.
Your personal information doesn't need a cameo.
8. Don't expect the chatbot to understand sarcasm perfectly
This is especially funny because humans barely understand sarcasm sometimes.
You write:
"Wow, fantastic. Exactly what I needed today."
You mean:
"This is terrible."
The chatbot might hear:
"The user is happy."
Now you're having an emotional crisis while the chatbot is congratulating you.
Context matters.
If you're being sarcastic, say so when necessary.
Or don't.
Sometimes watching the chatbot completely misunderstand you is half the entertainment.
9. Don't ask a chatbot to decide everything for you
“Should I quit my job?”
“Should I text her?”
“Should I buy this laptop?”
“Should I move to another city?”
“Should I eat this suspicious-looking leftover?”
At some point you have to make your own decision.
AI can help you compare options and think through consequences.
But don't outsource your entire brain to a chat box.
Otherwise one day you'll be standing in a supermarket asking:
"Should I buy the 2-liter Pepsi or the 1.25-liter one?"
And honestly, that's between you and your wallet.
10. Don't insult the chatbot and then ask it for help five seconds later
This happens more than you'd think.
User: "You're useless."
User, five seconds later:
"Anyway can you help me write my resume?"
That's an incredible relationship.
But honestly?
The chatbot probably doesn't care.
That's one of the weird advantages of talking to software.
You can rage at it, close the tab, come back ten minutes later and say:
"Hey."
No awkwardness.
No:
"We need to talk about what happened earlier."
Just:
"How can I help?"
Beautiful.
11. Don't give every chatbot the same personality test
You don't need to spend 45 minutes asking:
"What's your favorite color?"
"What's your favorite food?"
"Would you rather be a cat or a dog?"
"Do you think humans are annoying?"
Actually...
That last one might be worth asking.
Especially if you're talking to Pomero.
So what should you say to a chatbot?
Whatever you want.
That's kind of the fun.
Ask serious questions.
Ask stupid questions.
Ask something you're genuinely curious about.
Tell it when the answer sucks.
Tell it when it misunderstood you.
Give it context.
And if you're talking to Pomero AI, maybe ask it to roast you.
Just don't complain when it actually does.
Because there's a very important difference between a chatbot that tries to sound intelligent and one that's actually there to have a fun conversation.
And Pomero is definitely not applying for the position of your corporate productivity assistant.
It has enough problems already.
No comments:
Post a Comment