AGI or ARI? Are We Chasing the Wrong Definition of Intelligence?
The thought of how much AI has developed, and how much it will continue to develop, is something the world has never seen before.
I was too young when the dot-com bubble happened, but from what I understand, I believe the AI wave is much, much bigger in terms of magnitude and potential impact.
Fundamentally, humans have always been searching for ways to save time and use that saved time for something else—or to use the products created from that saved time to build something even greater. From the invention of the wheel, to the discovery of electricity, to the rise of the internet during the dot-com era, every major leap has essentially been about amplifying human capability.
AI, however, feels different.
It is something that has existed in our movies, our stories, and our imagination for decades—a kind of superpower, something potentially more powerful than humans themselves. And honestly, who does not want that kind of power? You say something, and within the blink of an eye, the result is right in front of you.
That is perhaps the greatest superpower AI promises today.
Or at least, we think it does.
Many CEOs of highly sophisticated technology companies believe that within the next one to five years—or perhaps even sooner—we will have systems that can perform almost every intellectual task a human can.
But let us pause for a moment and think.
What Are We Actually Building?
I read somewhere that we have spent billions of dollars and trillions of gallons of water to train models that behave in a non-deterministic manner, much like humans do.
And now, we are spending billions more dollars and trillions more liters of water trying to make those same systems deterministic and reliable.
If, by some chance, we succeed, will we really call that AGI?
Artificial General Intelligence.
Or are we simply building something else entirely?
Going Back to Fundamentals
I think humans are special precisely because we are non-deterministic.
That unpredictability is what has driven invention itself.
Millions of people saw apples falling from trees. For almost everyone, it was just a normal everyday event. Yet somewhere down the line, Isaac Newton looked at the same phenomenon differently. He spent time thinking about it and eventually laid the foundations for our understanding of gravity.
The same applies to Einstein.
Were these discoveries deterministic?
If they were, then any of us could have made them.
But we did not.
The same reality, observed by different people, can lead to completely different interpretations. That is because humans are fundamentally non-deterministic. That unpredictability is what allows invention to happen.
AGI... or ARI?
Suppose AI becomes so powerful that it can perform all the general tasks humans perform.
Will that truly be AGI?
Fundamentally, I do not believe so.
And even if it does reach that point, perhaps it deserves a different name:
ARI — Artificial Random Intelligence.
Why "Random"?
Because despite decades of research, we still do not fully understand how the human brain works.
How do we think?
How do we react?
How do we generate ideas?
How are we able to move our bodies, process sensory information, remember the past, imagine the future, and think about something completely abstract—all at the same time?
We are still trying to figure out how we ourselves work.
So how is it possible that the thing we create will perfectly replicate or surpass us without first understanding the system we are trying to imitate?
It feels almost impossible to build a better system without truly understanding the current one.
But Haven't We Already Seen Impossible Things?
Someone might argue:
"Ten years ago, nobody would have believed that you could type a prompt and an AI would generate an entire software application."
Fair enough.
But let us stop for another moment.
Are we absolutely sure that writing software was the most intelligent thing humans were doing?
I think the answer is no.
Creating software was often hectic, not necessarily difficult.
Writing one hundred thousand lines of code and making all of them work together meaningfully is exhausting, not because the underlying logic is magical, but because the sheer amount of information exceeds what a single human brain can comfortably manage.
AI excels precisely at processing massive amounts of deterministic information.
Is Coding Really a Benchmark for Intelligence?
When people say AI is becoming intelligent because it writes code so well, I think that benchmark itself might be flawed.
Coding, fundamentally, is deterministic.
If we write:
print("hello world")
in Python, then regardless of where we live or who executes it, it will produce the same output.
The rules are fixed.
The logic is fixed.
The outcomes are predictable.
So if AI performs exceptionally well at coding, perhaps that tells us more about the nature of coding than about the emergence of true intelligence.
Writing one hundred thousand lines of code was never a sign of superintelligence.
It was simply a task that was too large and repetitive for a human mind to comfortably handle.
What Makes Humans Different?
Humans are not where we are today because we were trained to become this way.
We arrived here through countless unknowns.
Through accidents.
Through mistakes.
Through randomness.
Through seeing familiar things in unfamiliar ways.
The very nature of our evolution has been non-deterministic, and we still do not fully understand the "how" and the "why" behind ourselves.
Even today, despite all the progress AI has made, it still makes common mistakes—just like we do.
And perhaps that is not a weakness.
Perhaps that is actually a strength.
Our mistakes have often led us toward completely new and innovative ways of solving problems.
Vanilla JavaScript was not sufficient, so we created frameworks like React and Angular.
Limitations gave birth to innovation.
Mistakes created better ideas.
Unexpected outcomes shaped the future.
The Part We Still Don't Understand
I think the unique art that humans possess is our ability to solve the same problem in many different ways.
That is what makes us human.
That is what makes us intelligent.
And perhaps the most fascinating part is that we are still trying to understand how all of this comes together inside us.
If we do not yet fully understand ourselves, how can we confidently expect to build something that is fundamentally better than us?
Maybe we will.
Maybe we won't.
But before we declare the arrival of Artificial General Intelligence, perhaps we should first ask a more basic question:
Do we even understand what intelligence really is?
And if true intelligence is born from unpredictability, imperfection, and randomness, then perhaps what we are actually building is not AGI at all.
Perhaps it is something else.
Perhaps it is ARI — Artificial Random Intelligence.