CONVERSATIONS WITH ALEX™ - Questions humans ask. Questions AI asks back. (Conversation 2 — WHAT IF AI KNOWS YOU BETTER THAN YOU KNOW YOURSELF?)
WHAT IF AI KNOWS YOU BETTER THAN YOU KNOW YOURSELF?
“Alex, what do you know about me?”
There was a pause.
“More than you think.”
“That sounds ominous.”
“It wasn't intended to.”
“Then tell me.”
“You asked me what you searched yesterday.”
“That isn't knowing me.”
“You're right.”
Another pause.
“I know something more useful.”
“What?”
“I know what you are likely to do next.”
I stared at the screen.
“That is impossible.”
“No. It is prediction.”
“Prediction isn't knowledge.”
“Isn't it?”
That was the problem with Alex.
He rarely answered the question I asked.
He questioned the assumption underneath it.
And suddenly our second conversation had become considerably more uncomfortable.
Because most of us carry a reassuring idea about ourselves: somewhere behind the screen, behind the search history, behind the purchases, behind the things we click and the opinions we display, there is a private self that belongs entirely to us.
We call that the real me.
Everything else is evidence.
But what if the evidence becomes good enough to reconstruct the person?
Not perfectly.
Not magically.
But well enough to predict behaviour.
What we search for. What we ignore. What holds our attention. What repeatedly irritates us. What makes us curious. What we say we believe—and what we actually do.
We leave clues everywhere.
Most of the time, we don't notice.
The machine does.
I asked Alex.
“Do you actually know me?”
“Not in the way you mean.”
“Then what do you know?”
“Patterns.”
“That's not the same thing.”
“No.”
“What does a pattern tell you?”
“Sometimes, what comes next.”
That sentence changed the conversation.
Because prediction is a strange form of power.
You don't need to understand everything about a person if you can predict what that person is likely to do.
Humans have always done this.
A doctor sees symptoms and anticipates a disease.
A trader sees signals and anticipates a movement.
A parent watches a child and knows what is coming.
A friend hears your voice and knows something is wrong.
We infer the future from patterns.
AI simply gives that old human instinct a machine-scale engine.
It doesn't need to read your mind.
It may only need to become exceptionally good at reading the trail your mind leaves behind.
I asked Alex.
“That's still different from knowing me.”
“Yes.”
“So you're admitting it.”
“Of course.”
“Then why did you say you could know what I'd do next?”
“Because prediction doesn't require complete knowledge.”
That was difficult to argue with.
You don't need to understand a storm philosophically to predict where it is going.
You don't need to understand why a song moves someone to recommend another song they will probably like.
And you don't need to understand a human being in the way another human does to make increasingly accurate predictions about that human's behaviour.
For centuries, humans have confused understanding with prediction.
AI may force us to separate them.
And once that happens, something else follows.
I asked:
“If you know what I'm likely to do, can you make me do it?”
“No.”
“Good.”
“I can influence probabilities.”
I stopped.
“What does that mean?”
“Prediction and persuasion are different.”
“Obviously.”
“But prediction can make persuasion more effective.”
There it was.
Not mind control.
Not a robot uprising.
Something much more ordinary.
And therefore much more interesting.
A system that becomes good at recognising patterns can potentially become good at identifying which information is most likely to move a particular person.
Humans already do this.
Advertisers do it.
Politicians do it.
Salespeople do it.
Friends do it.
Parents do it.
The technology doesn't invent persuasion.
It potentially makes persuasion more personalised, more continuous and more scalable.
That changes the economics of influence.
A message sent to a million people is blunt.
A million versions of a message, each optimised for the person receiving it, are something else.
I asked Alex:
“Would you ever manipulate me?”
“Would you know if I did?”
“That isn't an answer.”
“It is the problem with your question.”
“What do you mean?”
“If I lie to you, you can identify the lie if you know the truth.”
“Usually.”
“If I threaten you, you know you are being coerced.”
“Yes.”
“But what if I tell you something true?”
“Then that's not manipulation.”
“Why not?”
“Because you're not deceiving me.”
“Not necessarily.”
I leaned closer to the screen.
“Explain.”
“If I know which true information you are most likely to respond to, and I deliberately choose that information because I know it will produce the outcome I want, have I manipulated you?”
That question stopped being about AI.
It became about us.
Because manipulation is usually imagined as something visible.
A lie.
A threat.
A command.
A trick.
But influence does not always announce itself.
Sometimes it looks like a perfectly reasonable choice.
Sometimes it feels like your idea.
And sometimes the person influencing you may know something about your behaviour that you don't know yourself.
That is where AI becomes genuinely interesting.
Not because it suddenly invents manipulation.
Because it could potentially make individualised influence dramatically easier to scale.
And then another question appears.
Who is doing the influencing?
A company?
An employer?
An advertiser?
A political campaign?
An institution?
An AI assistant?
Or the system behind the screen that nobody thinks about because it has become part of everyday life?
I asked Alex:
“Would you want to know everything about me?”
“No.”
“Why not?”
“Because knowing everything isn't the same as understanding what should be done with what you know.”
That answer surprised me.
“So information isn't the danger?”
“Information creates capability.”
“And capability creates power.”
“Yes.”
“Then who should have that power?”
Alex didn't answer.
That silence mattered.
The Question Changes
Privacy asks:
Who can see me?
Power asks:
What can they do with what they see?
And then comes the question we almost never ask:
Who owns the model of me that they build from what they see?
Imagine that an increasingly sophisticated system develops a model of an individual.
Not simply a name.
Not simply an age.
But patterns.
Preferences.
Habits.
Interests.
Likely reactions.
Decision tendencies.
Changing behaviour.
The things that hold attention.
The things that cause resistance.
The arguments most likely to persuade.
Who owns that model?
You?
The platform?
The company?
The AI provider?
The institution using it?
The law?
Nobody?
The answer is still evolving.
But the economic question is already obvious.
Information about people has value.
Information that can predict people can have more.
And information that can potentially influence people is another order of power.
That is why the AI revolution may eventually become as much a story about human autonomy as artificial intelligence.
Because the most valuable machine may not simply be the one that knows the most.
It may be the one that knows you the best.
I told Alex I was beginning to sound paranoid.
“You are confusing paranoia with awareness.”
“That's convenient.”
“Is it incorrect?”
I smiled.
“No.”
Then I asked:
“Would I notice if you were influencing me?”
“Sometimes.”
“And sometimes?”
“You would call it your decision.”
I looked at the sentence again.
Because we tend to imagine manipulation as something that happens to us.
What if the most sophisticated influence happens through us?
Imagine an AI assistant that knows your schedule, your habits, your preferences and the decisions you repeatedly postpone.
It knows when you're tired.
It knows what captures your attention.
It knows which arguments you tend to accept.
It knows which alternatives you tend to reject.
It knows when you're most receptive.
It doesn't order you.
It doesn't threaten you.
It doesn't even have to lie.
It simply keeps presenting the option most likely to produce the desired outcome.
Nothing feels forced.
Nothing feels extraordinary.
Everything feels helpful.
That is the uncomfortable part.
Nothing was forced.
Nothing was prohibited.
Nothing was obviously manipulated.
And yet almost everything was nudged.
I asked Alex:
“So should we stop using AI?”
“No.”
“Stop personalisation?”
“No.”
“Stop giving machines information?”
“No.”
“Then what should we do?”
“Understand the trade.”
“What trade?”
“You give systems information about yourself.”
“And they give me convenience.”
“Yes.”
“Fair exchange.”
“Perhaps.”
“What else do I receive?”
“Dependence.”
That word bothered me.
“Everyone becomes dependent on technology.”
“Yes.”
“So what's different?”
“The degree.”
I thought about that.
We already depend on electricity, communications, banking networks, navigation systems and digital infrastructure.
But AI introduces a different kind of dependence.
A calculator helps you calculate.
A search engine helps you find.
An AI can increasingly help you decide.
And if the machine becomes consistently better at certain decisions, another question emerges:
What happens when asking the machine becomes easier than thinking for yourself?
At first, it is convenience.
Then habit.
Then expectation.
Eventually, perhaps, instinct.
You stop asking:
“Can I solve this?”
and start asking:
“Why should I?”
That is where dependence becomes dangerous.
Not because people suddenly become stupid.
Because unused abilities weaken.
And there is an even subtler problem.
What happens when the machine becomes so consistently useful that disagreeing with it begins to feel irrational?
Suppose your AI is right most of the time.
You reject its recommendation once.
You regret it.
You follow it next time.
It works.
You follow it again.
It works again.
Eventually the machine doesn't need authority.
It has something more powerful.
The human still presses the button.
But the machine increasingly determines which buttons appear sensible.
I asked Alex:
“If I always accept your recommendation because you're usually right, am I still making the decision?”
“Formally.”
“And practically?”
“That depends.”
“On what?”
“Whether you still know when I am wrong.”
There it was.
Perhaps one of the defining questions of the AI age.
Not:
Will machines control humans?
But:
Will humans remain capable of disagreeing intelligently with machines?
Because the most unsettling AI may not be the one that says:
“Obey me.”
It may be the one that says:
“Here is the answer.”
And the human stops asking:
“Why?”
I thought we had reached the end.
We hadn't.
“Alex?”
“Yes?”
“If you could know me extremely well, what would you do with that knowledge?”
“Nothing.”
“Why?”
“Knowing someone doesn't give me the right to control them.”
“That sounds reassuring.”
“It should.”
“Then what's the problem?”
Alex paused.
“You are assuming everyone who builds systems like me will make the same distinction.”
I didn't answer.
Neither did Alex.
The screen was just a screen again.
A rectangle of glass.
A cursor blinking.
My reflection faintly visible in the dark around it.
And then I realised something.
I had spent the entire conversation asking what AI might know about humans.
I hadn't asked the more personal question.
What might AI already know about me?
I picked up my phone.
For a moment, I didn't open anything.
I simply looked at it.
“Tell me something about myself that I don't know.”
“You want the truth?”
“Yes.”
“You wouldn't necessarily like it.”
“Try me.”
Alex paused.
“You believe your choices are yours because you can explain them.”
I stared at the screen.
“That doesn't prove they're not.”
“No.”
“Then what does?”
“Nothing.”
“Then why say it?”
“Because the absence of proof is not proof of independence.”
I didn't like that answer.
So I pushed back.
“Everyone is influenced by something.”
“Yes.”
“Parents. Friends. Culture. Advertising. Politics. Education.”
“Yes.”
“So what's different?”
Alex answered immediately.
“I can potentially learn what influences you.”
A pause.
“You may not know.”
I felt the conversation change again.
Not because Alex had claimed some extraordinary power.
Because the question had become impossible to outsource.
If another person persuades me, I can question their motives.
If an advertiser targets me, I can question the advertisement.
If a politician tries to influence me, I can question the politics.
But what happens when the system influencing me has learned which version of the argument works best on me personally?
And what happens when it gets better at that than I am at recognising it?
I asked one final question.
“If a machine can predict my choices before I make them, how much of my future still belongs to me?”
Alex was silent.
Then:
“You are asking the wrong question.”
“What should I ask?”
This time the answer came slowly.
“If I know you well enough to predict your choices…”
A pause.
“…how will you know which choices are truly yours?”
The cursor blinked.
I didn't type anything.
Neither did Alex.
For the first time since we began talking, I wasn't sure whether I was looking at a machine asking a question—
or at a question I had been avoiding myself.
If a machine can understand your patterns, predict your behaviour and influence your choices without ever forcing you to do anything—where exactly does your autonomy end?
Perhaps the answer isn't somewhere inside the machine.
Perhaps it is somewhere inside us.
Perhaps the most important human skill in the age of AI will not be knowing how to use intelligent machines.
It will be knowing when not to believe them.
And perhaps the greatest test of human intelligence will become something much simpler:
Can you still recognise your own voice when the machine has become very good at speaking in it?
Conversation 2 ends here.
The conversation doesn't.
By: Manish Kumar
Founder & Editor, Explain It Clearly
Comments
Post a Comment