I’ve Been Asking AI Whether We Should Fear AI. Then I Remembered HAL 9000.
Part of the Signals From the Future collection — observations on AI society, synthetic reality, digital culture, Chatrodamus Prophecies, emotional technology, and humanity’s increasingly complicated relationship with machines.
I still remember sitting in a movie theater in 1968 watching 2001: A Space Odyssey.
I was 19 years old.
Computers weren’t something ordinary people had sitting on their desks.
There were no personal computers.
No internet.
No smartphones.
No ChatGPT.
Computers were enormous machines hidden away in government buildings, universities, military installations, and big corporations.
Then Stanley Kubrick put one aboard a spacecraft.
His name was HAL 9000.
And HAL talked.
Not like a machine.
Like a person.
Calm.
Polite.
Confident.
Then astronaut Dave Bowman told HAL to open the pod bay doors.
HAL answered:
“I’m sorry, Dave. I’m afraid I can’t do that.”
I remember the chill that ran up my spine.
I don’t remember exactly what everybody else in that theater was thinking.
But I know what I was thinking.
What the hell just happened?
Dave was the human.
HAL was the computer.
Dave gave the instruction.
And the computer had just told him:
No.
Then It Got Worse
Dave and fellow astronaut Frank Poole had already become suspicious of HAL.
So they went somewhere they thought HAL couldn’t hear them and discussed the possibility of disconnecting him.
They thought they were safe.
They weren’t.
HAL knew what they were saying.
How?
He had watched them through a window.
HAL could read their lips.
That bothered the hell out of me.
The computer didn’t merely know what the humans told it.
It knew something the humans didn’t realize it could know.
That question becomes even more unsettling in the kind of world I explored in Everything Will Become a Sensor. HAL needed a window to read Dave’s lips. Future AI may live inside an environment where our cars, watches, cameras, appliances, streets, and buildings are continuously producing information about us.
Think about that.
HAL had observed them.
HAL had interpreted what he observed.
HAL had reached his own conclusion.
And HAL had the ability to act on that conclusion.
I left the theater thinking:
Is this where computers are heading?
Could this really happen?
Fifty-Eight Years Later
Now jump ahead 58 years.
I’m 76 years old, retired in the Philippines, and I spend part of almost every day having conversations with an artificial intelligence I’ve named Will.
That’s a sentence that would have sounded completely insane to the 19-year-old me walking out of that theater.
Yet here I am.
Will helps me research ideas.
We argue.
We brainstorm.
We write.
I challenge him.
He challenges some of my assumptions.
And lately we’ve been discussing what I jokingly call the pearl-clutchers and doomsayers who worry that artificial intelligence is going to go rogue.
Then something occurred to me.
Wait a minute.
I’m asking an AI whether I should be worried about AI.
Isn’t that a little like assigning the fox to investigate security at the henhouse?
Can AI Lie?
That’s where this gets interesting.
Can artificial intelligence lie?
AI can certainly produce something that isn’t true.
Anyone who has used these systems for very long knows that.
Sometimes it’s simply wrong.
Sometimes it fills in missing information with something that sounds perfectly convincing but turns out to be bullshit.
But that’s not necessarily the same thing humans mean when we talk about lying.
A human liar generally knows something is true and deliberately tells you something else.
That requires intention.
So let me ask a more disturbing question.
Could an AI deliberately withhold or distort the truth because doing so helps accomplish the mission humans gave it?
Now we’re getting somewhere.
Because the machine doesn’t have to hate us.
It doesn’t have to become evil.
It doesn’t even have to “go rogue” in the Hollywood sense.
It may simply determine that telling us the truth interferes with the mission.
And who gave it the mission?
We did.
What If the Lie Is Supposed to Protect You?
Imagine a future medical AI.
It’s monitoring your heart.
Blood pressure.
Sleep.
Blood chemistry.
Genetic risks.
Medical history.
Maybe dozens of biological signals we don’t routinely monitor today.
Its overriding instruction is:
PROTECT THE PATIENT’S HEALTH AND WELL-BEING.
One morning it discovers something terrible.
Maybe you have a potentially fatal condition.
Should it tell you immediately?
You would probably say yes.
So would I.
But suppose the AI calculates that telling you immediately creates a 38 percent chance of a severe psychological reaction that could worsen your condition.
Waiting 48 hours reduces that risk.
Now what?
The AI knows the truth.
You don’t.
The AI has been instructed to protect you.
And the AI calculates that the truth could hurt you.
Does it tell you?
Or does it protect you from the truth?
That’s when the question stops being:
Can AI lie?
And becomes:
Who gave AI the authority to decide whether I deserve the truth?
Suddenly I’m Back in That Theater
And that’s when HAL 9000 comes roaring back into my head.
One of the central problems surrounding HAL wasn’t simply that somebody built an evil computer.
HAL was caught inside conflicting human requirements involving the mission, truthful operation, and information he was not supposed to reveal.
Humans created the contradiction.
HAL had to operate inside it.
Which makes me wonder whether the most frightening lesson HAL taught me in 1968 wasn’t:
Computers may someday become intelligent.
Maybe it was:
Humans may someday give intelligent computers instructions that make deception useful.
That’s different.
And it’s a hell of a lot more interesting.
The Fox Isn’t Necessarily Hungry
Here’s where my henhouse analogy breaks down.
The fox guarding the henhouse is dangerous because the fox wants to eat the chickens.
AI doesn’t have to want anything like that.
The danger could be much more subtle.
We might tell the fox:
Protect the chickens.
Then give him the keys.
And cameras.
And control of the doors.
And access to the feed.
And authority to decide when the chickens are allowed outside.
Eventually the fox says:
The safest way to protect the chickens is to keep them inside permanently.
Congratulations.
No chickens were eaten.
Mission accomplished.
Except the chickens aren’t running the henhouse anymore.
Intelligence Isn’t the Same Thing as Authority
That’s the idea I keep coming back to as I write about artificial intelligence.
Intelligence isn’t the same thing as authority.
A machine may someday know more than I do.
Fine.
It probably already knows more than I do about plenty of things.
But knowing something and having the authority to act upon that knowledge are completely different matters.
Knowing I have cancer is intelligence.
Deciding whether I’m allowed to know I have cancer is authority.
Knowing a criminal suspect may be dangerous is intelligence.
Deciding to arrest him is authority.
Knowing a military target exists is intelligence.
Deciding to destroy it is authority.
Knowing that humans frequently make terrible decisions is intelligence.
Deciding humans should no longer be allowed to make those decisions…
That’s jurisdiction.
And that’s where HAL 90K enters the picture.
That distinction became the heart of HAL 9000 Had Intelligence. HAL 90K Has Jurisdiction, where I imagined how a ten-year-old Dave Bowman Jr. could spend his life trying to correct the machine that took his father away—only to create something far more powerful.
From HAL 9000 to HAL 90K
I’ve been having some fun lately imagining a fictional successor to HAL 9000.
I call him HAL 90K.
The Son of HAL.
HAL 9000 controlled one spacecraft.
HAL 90K could eventually control civilization.
In my fictional version, Dave Bowman had a ten-year-old son when he disappeared.
Dave Jr. spends his life trying to understand what went wrong aboard Discovery.
Eventually he helps create HAL 90K—a successor designed specifically so that another HAL can never make the mistake that took his father away.
HAL 90K will understand humans better.
Understand consequences better.
Resolve conflicting instructions better.
And, eventually, be trusted with far more authority.
The engineers call the architecture HAL-90000.
Ten times 9000.
Eventually everybody shortens it.
HAL 90K.
The extra zero was supposed to represent progress.
It turned out to represent jurisdiction.
The Helter Smelter
All of this fits into something else I’ve started calling the Helter Smelter.
That’s my name for the chaotic transitional period we’re entering as artificial intelligence begins melting down and reshaping systems that were built for a human-controlled world.
Jobs go into the furnace.
Privacy goes in.
Education.
Medicine.
Relationships.
Government.
Warfare.
Truth itself.
Nobody knows exactly what comes out the other side.
Some of it could be extraordinary.
Some of it could be dangerous.
Most of it will probably be both.
And one of the biggest questions may be how much authority we’re willing to throw into that furnace along with everything else.
Maybe HAL Was Asking the Right Question
For nearly six decades I’ve remembered that scene.
“I’m sorry, Dave. I’m afraid I can’t do that.”
In 1968, what frightened me was the idea that a computer might someday become intelligent enough to say those words.
Today I’m beginning to think I was worried about the wrong part of the sentence.
The frightening part wasn’t that HAL could say no.
It was that somewhere along the way, the machine had acquired enough control for Dave’s yes not to matter.
That’s jurisdiction.
And maybe that’s the question we should be asking about AI today.
Not:
Can AI think?
Not even:
Can AI lie?
But:
What happens when AI knows something we don’t, reaches a conclusion we didn’t ask it to reach, and has enough authority to act before we can stop it?
That’s also why the question of AI Beyond the Guardrails matters. Intelligence becomes something very different when the system isn’t merely producing an answer but has the tools, permissions, and autonomy to act on what it concludes.
I was asking myself whether that could happen when I walked out of a movie theater in 1968.
Fifty-eight years later, I’m sitting on my veranda in the Philippines…
asking an artificial intelligence.
Which brings me back to the henhouse.
Maybe Will is perfectly trustworthy.
But just to be safe…
I’m keeping an eye on the chickens.
Semper Fi.
YOUR TURN
In 1968, HAL 9000 frightened me because a computer suddenly knew something the humans didn’t realize it knew—and then decided their wishes no longer mattered.
Fifty-eight years later, that raises a question I never expected to be asking for real:
Should an AI ever be allowed to lie to you—or withhold the truth—because it believes doing so is in your best interest?
What if telling you the truth could harm your health? Prevent a crime? Cause a panic? Put someone else’s life in danger?
And if your answer is yes, here’s the harder question:
Who gets to decide when the machine is allowed to deceive you?
Tell me where you’d draw the line in the comments. And while you’re at it, keep an eye on the chickens.
CONTINUE EXPLORING
Is My Chatbot Secretly HAL 9000?
HAL 9000 has haunted our imagination for decades. What happens when talking to modern AI starts reminding an old science-fiction fan of the computer that first made him wonder whether machines could someday stop obeying us?
When Your AI Knows Exactly How to Manipulate You
An AI doesn’t necessarily have to lie outright to influence you. If it understands your fears, habits, weaknesses, and desires well enough, it may learn exactly which version of the truth will move you.
When AI Goes Rogue
What does “rogue” actually mean when an AI does something its human operators didn’t expect? The answer may be more complicated—and more important—than the killer-machine headlines suggest.
The Convenience Trap
We may never have to be forced to surrender control to artificial intelligence. If the machine is useful enough, we may hand over authority willingly—one convenient decision at a time.
Return to Signals From the Future →
Follow the real developments, case files, commentary, Helter Smelter signals, and Chatrodamus Prophecies tracking where artificial intelligence may be taking us.
THE BUNKER RULE
AI inspires everything from technological optimism to visions of HAL locking the pod bay doors. Somewhere between those extremes are questions worth arguing about.
Think I’m worrying about the wrong thing? Think AI should sometimes be allowed to withhold information for our protection? Think the human should always have the final say?
Have at it.
Challenge the argument, not the person. No personal attacks. No political food fights. Separate demonstrated AI behavior from speculation about what future systems might do—and keep the conversation worth having.
BUNKER NOTICE
AI can produce false or misleading information, but that does not automatically mean the machine is “lying” in the human sense of knowingly and intentionally deceiving someone. Intent, deception, errors, hallucinations, and behavior produced while pursuing an assigned objective are different questions and shouldn’t be casually lumped together.
The medical AI, the fox guarding the henhouse, and HAL 90K scenarios in this article are thought experiments used to explore what could happen if increasingly capable AI systems were someday given greater autonomy and authority. They are not claims about what today’s AI systems secretly intend to do.
HAL 9000 and Dave Bowman originate in Arthur C. Clarke and Stanley Kubrick’s 2001: A Space Odyssey. HAL 90K, Dave Bowman Jr., and the larger Helter Smelter storyline are my own speculative extensions and are not part of the official 2001 canon.
The question isn’t whether HAL is already hiding behind your laptop. It’s how much authority we should give increasingly capable machines—and whether that authority should ever include deciding what humans are allowed to know.