COULD AI REALLY DESTROY HUMANITY? Let’s Talk About This Without the Hype
I have been following a story coming out of the United States, and I will admit that it caught my attention.
The headline is frightening.
Artificial Intelligence could become so powerful that it poses a threat to humanity, possibly within the next decade.
Now before anybody starts throwing away their phone, shutting down their computer or blaming ChatGPT for everything wrong in the world, let us understand what is actually being said.
Because there is a difference between saying AI WILL destroy humanity and saying there is a risk that future AI could become dangerous if humans lose control of it.
And that difference is important.
The warning is coming from people inside the industry
What makes this story worth paying attention to is that these warnings are not simply coming from somebody sitting in his bedroom watching science-fiction movies.
Some of them are coming from researchers who have actually worked on advanced AI.
Former Anthropic and OpenAI researcher Jacob Coxon recently raised concerns about where this technology is heading and whether humans will be able to reliably control much more powerful AI systems.
Another Anthropic researcher, Evan Hubinger, has publicly put his own estimate of the risk of an AI catastrophe killing everyone within the next decade at greater than 10 percent.
That doesn't mean there is a scientific calculation proving we have a 10 percent chance of extinction.
It is his assessment of the risk.
But think about it another way.
If an engineer told you there was a 10 percent chance that a bridge would collapse while you were crossing it, would you drive across?
I wouldn't.
And that is why I don't believe we should simply laugh at these people.
But how can a computer destroy the world?
That was my question too.
My little laptop sitting on my desk cannot suddenly get up and attack anybody.
Neither can your cellphone.
And the AI systems we are using today are not the super-intelligent machines these researchers are worried about.
The concern is about what comes next.
Suppose AI eventually becomes capable of performing many intellectual tasks better than humans.
Suppose it can write its own computer programs, find weaknesses in computer networks, operate other machines, manipulate people, conduct scientific research and make decisions at a speed humans simply cannot match.
Then suppose we begin giving those systems more independence.
That is where the concern begins.
The danger doesn't necessarily require some evil robot waking up one morning and saying, "I hate humans."
The problem could be much simpler.
We give an extremely powerful system a goal.
It finds a way of accomplishing that goal that we never anticipated.
Then we discover that stopping it is much harder than starting it.
We have seen a smaller version of this problem before
Human beings have a habit of inventing first and dealing with the consequences afterwards.
We built cars before we developed the road-safety systems we have today.
We embraced social media before fully understanding what misinformation, addiction, algorithms and mass manipulation could do.
We developed plastics without imagining that decades later we would be finding microplastics practically everywhere.
The invention comes first.
The rules usually come afterwards.
With ordinary technology, we sometimes get a second chance.
The people sounding the alarm about advanced AI are asking a troubling question:
What if this is one technology where getting it badly wrong does not give us a second chance?
That is the part of this discussion that bothers me.
Senator Jon Ossoff has entered the discussion
This concern has now reached American politics.
U.S. Senator Jon Ossoff has called for stronger oversight of advanced AI research, including inspectors at major AI laboratories and international cooperation.
And whether you agree with him or not, the basic question deserves an answer.
Who is watching the people building the world's most powerful AI systems?
Right now there is enormous competition between technology companies.
There is also competition between countries.
Everybody wants to be first.
Everybody wants the most powerful system.
There are billions, perhaps eventually trillions, of dollars at stake.
That creates another problem.
Imagine two companies racing towards the same finish line.
One company says, "We should slow down and make sure this thing is safe."
The other says, "Fine. While you slow down, we'll get there first."
You can see the difficulty.
And where does little Saint Lucia fit into all this?
This is where I believe we in small countries need to start paying attention.
Most of the powerful AI systems being developed today are not being built in Saint Lucia, Dominica, Barbados or the rest of the Caribbean.
We are mainly consumers of the technology.
Yet if AI transforms employment, banking, education, journalism, cybersecurity, elections, warfare and the global economy, we will live with the consequences too.
Nobody will ask Saint Lucia whether the world should develop super-intelligent AI.
The decision will largely be made elsewhere.
But whatever comes from that decision will eventually arrive on our shores.
Already our children are using AI for schoolwork.
Businesses are using it.
Governments will increasingly use it.
People like me use it to research, write, create and understand things that previously might have taken hours or days.
There is tremendous good in this technology.
I can see it myself.
But something can be extremely useful and still require rules.
Electricity is useful.
We still put breakers in our houses.
Cars are useful.
We still put brakes on them.
Medicine saves lives.
We still regulate drugs.
So why should the most powerful AI systems humanity has ever created be treated differently?
Are we all going to die in ten years?
Nobody knows.
And anybody telling you that scientists have proven AI will wipe out humanity within ten years is going too far.
They haven't.
There is enormous disagreement among AI researchers about how likely these extreme scenarios really are.
Some researchers believe the extinction warnings deserve urgent attention.
Others think those scenarios are too speculative and distract attention from problems AI is already causing today.
And those present-day problems are real enough.
Scammers can use AI.
Criminals can use AI.
Governments can use AI for surveillance.
Fake photographs, voices and videos are becoming harder to distinguish from genuine ones.
Workers could find some jobs disappearing or changing.
Students can become dependent on AI instead of learning how to think.
And powerful technology concentrated in the hands of a few companies or governments creates problems of its own.
We don't have to wait for some super-intelligent robot to see reasons for concern.
So should we be afraid of AI?
I wouldn't use the word afraid.
I would say we should be awake.
AI could turn out to be one of humanity's greatest tools.
It could help discover medicines, improve education, increase productivity, assist people with disabilities, improve disaster forecasting and solve problems we presently struggle to solve.
I don't want us to throw that away because somebody frightened us with a headline.
But neither should we become so fascinated by what AI can do that we stop asking what it should be allowed to do.
That is the conversation I believe ordinary people should be having.
Not:
"Will robots kill us next Tuesday?"
But:
Who controls increasingly powerful AI?
Who decides how far it goes?
Who makes sure it is safe?
Who is responsible when something goes wrong?
And perhaps most importantly:
Are governments putting the brakes and steering wheel on this thing while there is still time to do it?
I use AI.
I appreciate what it allows me to do.
I also believe that when some of the very people helping to build this technology start warning the rest of us about where it could eventually lead, we should at least listen.
Don't panic.
Don't swallow every frightening headline either.
But don't laugh this one away.
Sometimes the sensible thing to do when somebody who understands the machinery tells you there might be a problem under the hood is to open the bonnet and have a proper look.
Choiseul on the Move





