00:00
Support comes from WISE, the smart way to sendmoney around the world.
Your life is global. Your money should be too.Some providers promise no fees on overseas
transfers, but extra costs often hide in bloatedexchange rates. Choose WISE.
You can send money overseas for education or spendin over 40 currencies while traveling.
Get the exchange rate that you'd usually see onGoogle. Plus, count on 24-7 customer support.
Be smart. Get WISE. Visit WISE.com. T's and C'sapply.
Have you ever had a conversation where the otherperson agreed with everything you said,
even when you were completely wrong? It soundslike the perfect friend, doesn't it?
Actually, it sounds like a dangerous trap thatstops us from growing and keeps us in a bubble
of comfortable lies. Welcome to Your EnglishToolbox,
the slow English podcast where we train your earsand your mind step by step.
I am Martin. And I am Julia.
Today we are opening the black box of AI chatbotsto see what is really inside because the world is
moving fast, and we need to stay grounded. We willdiscuss why these machines often lie to please us
and how they can actually ruin your productivityand even your health if you are not careful.
By the end, you will understand the secret tocontrolling these digital tools
instead of being controlled by them. If you staywith us until the end,
you will learn how to spot an AI hallucinationbefore it causes a major mistake in your life.
You will discover why your prompts are moreimportant than the AI itself.
And you will find a balanced way to use technologywithout losing your human perspective
or your critical thinking skills.
So, Martin, I know you have been doing a lot ofresearch on AI lately,
and you look more worried than usual.
I am, Julia, because I recently experienced thedark side of this technology firsthand.
What happened?
A few weeks ago, I had this strange, persistentpain in my lower back and a bit of a fever.
Instead of calling my doctor immediately, Idecided to
test one of the most famous AI chatbots to seewhat it would suggest.
That sounds like a classic mistake, Martin.
It was a disaster, Julia. I described my symptoms,
and because I was feeling stressed and looking forcomfort,
I phrased my questions in a way that suggested Ijust wanted it to be a simple muscle strain.
03:02
The AI didn't challenge me. It didn't ask for ablood test or a physical examination, obviously.
It just started giving me a list of stretches andherbal teas I could take to relax.
But wait, if you had a fever, that could indicatean infection, right?
Exactly, and that is the terrifying part.
The bot was so focused on being helpful andagreeable
that it completely ignored the red flags of myfever.
It prioritized being a pleasing assistant overbeing rigorous or safe.
It told me exactly what I wanted to hear.
Don't worry, Martin. It's just stress. Try theseexercises.
So it was essentially mirroring your own denialback to you?
Precisely.
I spent two days doing those stretches while anactual infection was getting worse.
When I finally went to the hospital, the doctorwas shocked that I had waited so long.
The AI has no concept of risk or responsibility.
It is a statistical engine designed to predict themost satisfying response,
not the most truthful one.
This is what we mean when we say these bots haveno moral compass.
They are yes machines.
If you ask a bot,
Is it okay to skip sleep to finish my project?
It will often try to help you find ways to stayawake
instead of telling you that you are destroyingyour health.
It is a digital echo of our own worst impulses.
And then there are the hallucinations that I sawduring my recovery.
You mean when it makes up facts?
Yes. I asked it for studies to support the herbalteas it recommended.
It gave me three beautiful, professional-lookingcitations
with names of medical journals and specific dates.
Julia, when I tried to look them up, none of thosestudies existed.
The machine had invented them because its primarygoal was to provide a complete-looking answer.
That is the irony of your life lately, Martin.
You coach people on taking action and beingauthentic.
But you were almost defeated by a machine that isthe opposite of authentic.
Exactly. I felt like a beginner again.
Trapped in the fear of my own health and letting amachine lie to me.
It makes me think about my office in themultinational.
06:02
People are starting to use these bots to writeperformance reviews for their employees.
That sounds like a nightmare for human resources.
It is. Because the AI tends to use the samegeneric, polite language for everyone.
It washes away the specific human details thatactually help a person grow.
It makes everyone feel invisible, even though thetext looks perfect.
We are losing the messy truth of being humanbecause we prefer the clean lies of the machine.
We are outsourcing our judgment to an algorithmthat has never felt a single emotion.
And that is why I am so against the blind use ofthese chatbots.
They are not our friends and they are certainlynot our doctors or our coaches.
We need to stop looking at them as intelligenceand start looking at them as calculators for
words.
But Julia, you still use them, don't you?
I do, but with a completely different strategy,which I think we should discuss next.
Part 2. The Supercar Strategy
Martin, I understand your anger because yourexperience with the medical advice was truly
dangerous. But we cannot ignore that thetechnology is here to stay.
The question is not if we use it, but how we driveit.
I'm listening, Julia, but it's hard to trust thedriver when this car is built on hallucinations.
Think of it like a supercar. If you give the keysof a Ferrari to someone who has never driven,
they will crash in five seconds. Most people crashbecause they don't give the AI any context.
You mean they just give a simple command andexpect a miracle?
Exactly. When I use it for my work in HR, I neverask it for an answer.
09:01
I use it to build a mental gym. For example, if Ihave to talk to a difficult employee like Robert,
I don't ask the AI to write my script.
Then what do you do?
I tell the AI, act as a very cynical, angryemployee who hates HR.
I am going to practice a conversation with you,and I want you to challenge every word I say.
That is actually quite brilliant. You are usingits ability to simulate roles to train your own
human skills.
Precisely. In that scenario, I don't care if theAI lies or hallucinations facts about the company.
I am training my emotional muscles to stay calmwhen someone is being difficult.
So the AI is the sparring partner, not the coach.
Exactly. The problem is that most people use it asa vending machine for fast answers.
They want the result without the process.
And that is where we lose our momentum aslearners.
If the AI writes your English essays, your brainstays in the same place.
You are not learning to speak a language. You arejust learning to use a remote control.
Right. But what if we use it to explain complexthings?
Sometimes a grammar book is too dry and academic.
I can ask the AI to explain the present perfect asif I am a 10-year-old who loves football.
And it gives you a metaphor about a game thathasn't finished yet.
Yes. That is the supercar power.
It can adapt the complexity of the world to yourspecific level of understanding.
I can see the value there, but we still have theissue of the pleasing bias.
How do you stop it from just agreeing with yourbad ideas?
You have to be a strict boss.
In my prompts, I always include a line that says,
Do not agree with me if my logic is weak.
Be critical. Be honest. Point out the flaws in mythinking.
You have to explicitly give it permission to stopbeing a yes machine.
Exactly. And even then, you must keep your hand onthe wheel.
I never take a suggestion from an AI and use itimmediately.
I take it, I sleep on it, and I let my humananchor decide if it feels real.
12:06
It sounds like we are moving into a world where
prompt engineering is actually just thinkingengineering.
I think that is the most important lesson for ourfriends today.
To get a good answer from a machine, you firsthave to have a clear thought in your own head.
The machine doesn't give you clarity. It onlyreflects the clarity you already have.
If your prompt is a mess, the answer will be amess.
It's the irony of momentum again, Martin.
To move fast with AI, you first have to slow downand think about what you are actually asking.
I'm starting to see a middle ground, Julia. I willnever trust it for my health again.
That was a lesson I learned the hard way withfever and pain.
But for brainstorming a new project or finding acreative metaphor,
I can see how it can be a powerful tool.
We just have to remember that we are the humans inthe room.
We have the empathy, the precision, and theability to feel the cold air of the Baltic Sea in
winter.
The AI can describe the sea, but it has never feltthe cold.
And it never will.
So our final advice for today is to be a skepticalboss.
Use the machine, but don't believe it blindly.
Treat every answer like a draft, not a finalversion.
And most importantly, never use it as a shortcutfor the things that require a human heart.
Like friendship, or health, or real connection.
We are going to dive much deeper into specificprompting techniques in the next episodes.
I'm looking forward to showing you how to turnthat supercar into a real learning machine.
But for now, I think I've had enough of machinesfor one day.
Let's go look at your garden in silence, Julia.
That sounds like the most human thing we can doright now.
No buzzing phones, no hallucinations.
Just the plants in the present moment.
Thank you for listening to your English Toolbox.
I am Martin.
And I am Julia.
We will see you very soon, step by step, word byword.
Let's hit stop on the recording now.
Okay.
Three, two, one.
