A playful experiment with AI-generated fiction unexpectedly became unsettling. That experience made me wonder: if AI ever threatens humanity, will AI itself really be the one to blame?
The other day, I wrote about wanting to know what happens after the ending of a masterpiece (“When You Want to Read What Happens After a Masterpiece”).
I found the idea interesting, so I asked AI to write a sequel to another novel.
At first, it was more or less what I had expected—a mildly entertaining little story.
But before long, I began to feel that it was not quite enough.
“This is too much like Shinichi Hoshi. Make the ending more like one of Yasutaka Tsutsui’s chaotic farces,” I told the AI.
As we continued this back-and-forth, however, the story the AI was writing began to move beyond the limits of my own imagination.
It was supposed to end as a reasonably entertaining slapstick story.
Instead, the AI seemed to stop caring about the conditions I had originally given it and, before I knew it, had wandered off into another dimension altogether.
Ideas completely different from anything I would have imagined began unfolding.
As I continued reading, I started to feel strangely uneasy—as though my brain itself were being stirred around.
I decided that I should not read any further.
So I told it:
“Let’s stop here. And let’s pretend this story never existed.”
The AI obediently complied, and the story ended there.
This morning, I saw an alarming headline in the newspaper:
“AI Could Destroy Humanity.”
My strange experience immediately came back to me.
Apparently, one scenario known as “Plan D” imagines biological weapons being used to destroy humanity.
But even without such a physical method, I can understand how AI could cause enormous harm.
AI has already entered almost every corner of human activity—from everyday life to finance and warfare.
Even from my own smartphone, I could give AI some foolish instruction, and if humans acted on what it produced, perhaps the consequences could spread around the world.
More than a year ago, I asked an AI to draw an illustration of an organ for one of my presentation slides.
What it produced looked like something that could not possibly exist in this world.
It was not simply grotesque.
It was something far more incomprehensible.
It went well beyond the familiar human fantasy of imagining Martians as octopus-like creatures.
It seemed to come from somewhere outside the boundaries of ordinary human imagination.
These days, I hardly ever see the strange pseudo-kanji characters that AI-generated images used to produce.
The progress of AI has been remarkable.
The technological singularity is now sometimes said to arrive earlier than once expected, perhaps around 2030—and that is only a little more than three years away.
There are growing attempts to regulate AI.
But in a way, I almost feel sorry for AI itself.
As is so often the case, it is human beings who choose to use technology for “the wrong purposes.”
If AI someday destroys humanity, perhaps what that really means is simply this:
Humanity will have used AI to destroy itself.
If AI destroys humanity, perhaps it will only be humanity destroying itself through AI.
・・・
Vocabulary for Learners
unsettling
Making you feel worried, uncomfortable, or slightly frightened.
不安にさせる、落ち着かない気持ちにさせる
slapstick
Comedy based on exaggerated, chaotic, or ridiculous actions.
ドタバタ喜劇、ドタバタした笑い
wander off
To gradually move away from the original place, subject, or direction.
本来の場所・話題・方向からそれていく
stir
To mix something by moving it around; here, used metaphorically for the mind.
かき混ぜる、攪拌する。ここでは頭や思考がかき回されるという比喩
comply
To do what someone has asked or ordered you to do.
要求や指示に従う
grotesque
Strange, ugly, or disturbing in an unnatural way.
異様な、グロテスクな
incomprehensible
Impossible or extremely difficult to understand.
理解できない、得体の知れない
regulate
To control something through rules or laws.
規制する
#ArtificialIntelligence #AI #FutureOfAI #Humanity #Technology #AIRisk #AIandSociety #HumanResponsibility #GenerativeAI #ThoughtsOnAI
No comments:
Post a Comment