A few days ago, I was genuinely enjoying a video by Geoffrey explaining how AI works. He managed to describe some rather technical aspects of AI in a remarkably simple, everyday language, while still keeping the technical substance intact.
But then, towards the end, he finished his talk in a very strange way.
“This is why…” he said, and then added:
“This is why AI could destroy humanity. We should be afraid.”
Yes. He really said that.
My natural expectation was almost the exact opposite:
“This is why AI cannot destroy humanity.”
I have no idea what Geoffrey’s actual motivation was in that talk. I don’t know whether he had complete freedom to decide what he wanted to say. Perhaps some intelligence agency, government organization, or even the producers had their own agenda. Maybe someone offered him additional opportunities in exchange for saying something dramatic enough to make headlines.
After all, earlier in the talk, he mentioned that he needed $2 million and that he couldn’t make that kind of money simply by working at Google.
And let’s be honest: today, people who say extraordinary things—or deliberately provocative or sexual things—can attract enormous attention and, quite often, find ways to monetize it.
There is a carpet in the room nobody wants to lift
There is something underneath the carpet that nobody seems particularly interested in looking at.
The people developing AI are also building the evidence base that can later be used to blame AI for the things they themselves choose to do with it.
That, in my view, is where the truly frightening AI story begins.
Whenever someone tries to say this a little too loudly, the conversation is quickly redirected. We are told to “look over there.”
Look at the AI.
Don’t look at the people standing behind it.
The arguments presented as “AI risks” could just as easily apply to an innocent kitchen knife you use to cut fruit. They could also apply to an autonomous vehicle packed with sophisticated technological systems.
How far an AI system is allowed to go, how deeply its decision-making mechanisms are allowed to penetrate, and where its “hands” can reach are ultimately decisions made by people.
Usually by developers. Or their employers.
And, yes, sometimes by governments.
It isn’t particularly difficult to imagine what the interests and intentions of those who define these boundaries might be.
Think about Elon Musk. Think about Sam Altman.
If the people who commissioned the development of a nuclear weapon had certain priorities and constraints when they gave the development team its mission, I assume those same priorities would also be reflected when defining the missions, rules, and boundaries of AI systems.
But I want to underline something very clearly:
This is not a criticism.
It is an explanation of the strategy. A global and maybe good strategy!
Whether that strategy is right or wrong is a completely different question—and ultimately depends on the perspective of the people and societies affected by its consequences.
The groups and individuals who benefit from a particular strategy may celebrate it by saying:
“This is a huge success!”
Those who suffer from it may instead ask:
“What the hell are these people doing?”
Both reactions can exist at the same time.
AI’s job is to imitate almost perfectly
The fundamental job of AI is to imitate.
A supercomputer is not inherently more conscious than a knife you use to cut fruit.
But if someone picks up that knife and throws it at you, the knife will still perform the function it was physically given.
AI is still built on surprisingly limited computer software compared with the science-fiction version of an artificial human many people imagine.
The mathematical functions behind modern AI may be extraordinarily powerful at inference, prediction, and pattern reconstruction. But the result is still, fundamentally, something derived from what already exists.
As long as AI continues along this trajectory, I don’t see it spontaneously becoming conscious.
But I don’t know whether some of the hybrid mechanisms that will emerge in the future could eventually open the door to an era of machine consciousness.
That is a different question.
So what exactly is AI ethics?
What is the ethics of AI and robotics (tech)?
What should it be?
This is a global and ,as I said above, a mass-scale problem.
And yes, I deliberately call it a “problem.”
Because once AI systems capable of reaching critical areas of society exist, the idea that they might suddenly “go crazy” is an incredibly marketable narrative.
“AI has gone out of control and attacked.”
It would look perfect as a breaking-news headline.
But behind the curtain, things may be very different.
If it is technically possible to design enormous, deeply integrated analytical and decision-making systems in ways that could enable such “madness,” prevent it, or simply make certain actions impossible in the first place, then who exactly should be held responsible for the resulting risks?
The developers.
And those who define what the developers are allowed or required to build.
“Why are you trying to scare people?”
Recently, another argument has started becoming more common:
“Why do you keep talking about AI risks when you could be talking about all the benefits of AI?”
Fair question.
AI is an incredibly sophisticated software system with the potential to solve many extremely difficult problems.
So why would anyone deliberately add algorithms, permissions, interfaces, and code that allow it to do things beyond simply solving those problems?
Again, we can go back to the fruit knife.
A knife is perfectly useful for cutting fruit.
The interesting question is not what the knife can do.
The interesting question is who decides what else it should be allowed to do.
So, yes. Get ready.
In short:
Yes. Get ready.
AI has arrived with disruptive, powerful, and potentially destabilizing effects.
Most people are still playing with the entertaining(!) toys placed in front of them, without realizing that they are already interacting with something much larger.
Companies continue to behave as though privacy policies are steel doors protecting everything behind them.
They aren’t.
We may soon be entering the days when the trees in the forest start talking.
And perhaps that will actually be a good thing.
Who knows?