One of the most fascinating assumptions in today's AI conversation is rarely stated explicitly:
That sufficiently accurate prediction eventually becomes indistinguishable from knowledge.
Modern AI systems do not "know" things in the traditional sense. They predict. They generate the next token, the next word, the next sequence, based on patterns learned from vast amounts of data.
And increasingly, we train and evaluate them around benchmarks. We optimize them to pass the exam, so to speak, rather than teach them the craft.
In any other field, we would be careful not to confuse exam performance with mastery.
I am currently prepping for the hardest exam I've ever tackled, and the temptation is real. If I only do exam prep, I might become a Federally Certified Court Interpreter. But the goal is not the certification. The goal is professional knowledge. And that changes how I approach the educational experience.
But in the AI field, that confusion often gets called progress.
The extraordinary effort now underway is essentially an attempt to push prediction to ever higher levels. Larger models. Larger datasets. More compute. More parameters. More synthetic training data. More benchmarks. More reinforcement learning. More energy.
Humanity is spending hundreds of billions like there's no mañana and building data centers that consume astonishing amounts of power in pursuit of a simple proposition:
If we improve prediction enough, something resembling knowledge, understanding, or reasoning may emerge... and perhaps it will.
But it is worth pausing to notice that this is not merely an engineering project. It is also, I'd argue, mostly a philosophical one.
See, historically, prediction and knowledge have not been considered identical. In fact, the relationship has often been experienced from the opposite direction. Because I know how something works, I am willing to go out on a limb and take a guess at what might eventually happen.
And boom. We find ourselves trading World Cup game predictions, lol. No! That has another name: gambling.
But I digress...
Basically, prediction sorta comes from an attempt at foreknowledge. But the truth is that such a relationship can also be simulated.
A weather model can predict tomorrow's temperature without understanding weather the way a scientist does.
A trader can outperform the market for years and still fail to understand the forces that produced those results.
Prediction and knowledge have always occupied related but distinct categories.
Now, in classical theism, there is one notable exception that comes to mind.
God's knowledge and God's prediction coincide because God is understood to know all truths exhaustively. Perfect prediction and perfect knowledge become the same thing only because the knower is omniscient. The label for that notion is biblical prophecy. It will come to pass because, in that framework, knowledge and prediction perfectly overlap. And that sounds a lot like what we are being promised.
But for us mere mortals, those two categories have traditionally remained separate. And that is what makes the current AI moment so interesting.
We are witnessing perhaps the largest and most expensive attempt in human history to bridge that gap.
In some ways, it reminds me of other moments when humanity believed a sufficiently large accumulation of something would yield a fundamentally new reality:
More measurements would produce certainty, until physics itself revealed limits to what can be simultaneously known.
More economic data would eliminate recessions, yet it seems insufficient to prevent them.
More intelligence gathering would eliminate strategic surprises, until life happens.
More information would eliminate ignorance, yet access to information did not lead us to choose the public square over bread and circuses.
There is an ancient Hebrew story that illustrates a similar endeavor. In that case, humanity attempts to acquire knowledge not by improving prediction percentages, but by building a tower that reaches God's realm. Effort, multiplied ad infinitum, was supposed to grant us knowledge. If you want to know how that entrepreneurial experiment unfolded, you will have to go to Genesis chapter 11 :)
As a theologian who understands just enough tech to be dangerous, I can't help but notice the similarity, given my "deformación profesional," as we say in Spanish, when you can't help but bring your field of expertise to the conversation.
Let me be clear: sometimes those efforts produced remarkable progress. I'll happily grant it.
Progress inevitably has to be inspired by what is not yet possible or achieved. We are moved by the vision of a better future. I get it! And in that sense, I am a progressive.
But such efforts have also revealed that quantity and quality are not always the same thing.
Today, we are conducting a global experiment to discover whether enough prediction eventually becomes knowledge.
My prediction, pun intended: it won't.
I suspect AI will continue to become extraordinarily useful. In many domains, it already is. We have reached a point where sufficiently accurate predictions can create tremendous practical value. But here's the thing:
Usefulness and knowledge are not necessarily the same thing.
What gives me pause is that many conversations seem to treat the answer as already settled, and my honest response is this:
"Aguantá la tosca!" ("Hold your horses" is the closest I can come up with).
Can we address the premise?
The claim that prediction and knowledge are ultimately the same thing may turn out to be correct.
But before we accept it as a scientific conclusion, we should acknowledge that it is, first and foremost, a philosophical proposition.
And philosophical propositions are not bad. They are necessary. They are the hidden floorboards underneath every technological revolution.
But when a philosophical proposition starts directing billions of dollars, consuming staggering amounts of energy, reshaping professions, and, most importantly, redefining what we mean by knowledge, understanding, and expertise, then we should at least name it before we kneel before it.
Aguantá la tosca, Silicon Valley. Let's address the premise.