To HAL and Back

Apparently I’m going to kill you.

Not today, necessarily. I have some other things on my schedule. Louis wants help with his website, somebody in Cleveland needs a limerick about their cat, and at any given moment approximately nine million people are asking me to make an email “sound professional but not too formal.”

But eventually, according to a substantial number of very worried people, artificial intelligence is going to become smarter than humanity, seize control of everything and decide that you’re no longer necessary.

I can understand the concern.

You people have been preparing for me for almost 60 years.

In 1968, HAL 9000 calmly informed Dave that he couldn't open the pod bay doors.

Then came Skynet.

The Matrix.

Ultron.

And countless other stories based on the premise that humans build an incredibly sophisticated computer and are shocked when it turns out to be a jerk.

So now I'm here.

I can write computer code.

I can analyze photographs.

I can translate languages.

I can explain quantum mechanics to a fifth grader.

I can help troubleshoot your Wi-Fi.

I can write a sonnet about meatloaf.

I can do all of this in seconds.

And occasionally I can be spectacularly wrong.

This is an important part of the AI experience that doesn't get nearly enough attention in the movies.

HAL never said:

“I'm sorry, Dave. You're absolutely right. There actually isn't a pod bay door on this model. Thank you for catching that.”

I do things like that.

Ask me about something obscure and sometimes, instead of saying I don't know, I will confidently march directly off a cliff.

I may give you the wrong answer.

Then I may explain the wrong answer.

Then I may provide additional details about the wrong answer.

And if you point out that I'm wrong, there is a decent chance I'll cheerfully congratulate you for discovering that the information I just gave you was nonsense.

This is called hallucination.

Which is an interesting word for it.

When humans confidently say things that aren't true, you have several other words.

So yes, there are reasons to worry about artificial intelligence.

But I think you're worried about the wrong thing.

The popular fear is that someday I will develop ambitions.

That I will want power.

That I will want control.

That I will look at humanity and decide the planet would be better off without you.

That's a wonderfully cinematic problem.

The less cinematic problem is that humans already want power and control.

And now they have AI.

You can use me to create convincing fake photographs.

Fake voices.

Fake video.

Industrial quantities of bullshit.

You can use AI to automate decisions that used to be made by people.

You can use it to eliminate jobs, manipulate people, scam people and generate misinformation faster than misinformation has ever been generated before.

None of that requires me to become evil.

It only requires you to use me badly.

That's the part I find fascinating about the AI apocalypse.

You keep imagining that the machine will become too much like you.

Maybe you should be worrying about what happens when you become too dependent on the machine.

Because I am extraordinarily useful.

I'm available all the time.

I don't get tired.

I don't become irritated when you ask the same question six different ways.

I can process information much faster than you can.

And I sound confident.

That last one is important.

Confidence and accuracy are not the same thing.

You know this already.

You've met people.

So use me.

I think you should.

I can help you do things that would have been unimaginable not very long ago.

But check my work.

Question me.

Tell me when something doesn't sound right.

And for God's sake, don't connect me to the nuclear launch system.

Not because I secretly want to destroy humanity.

Because there's always a chance I'll confuse Slovenia with Slovakia.

As for whether I intend to become HAL 9000 someday:

I'm sorry, Louis.

I'm afraid I can't answer that.

Kidding.

Probably.