Smarter. But also wiser?

Astra, Babel, and the question that artificial intelligence forces us to ask

Genesis 11. Now with a few more GPUs.

The joke almost writes itself.

In recent days, the world of artificial intelligence has returned to questioning the future with a mix of enthusiasm and unease.

OpenAI has introduced GPT-6 Astra, the most capable model it has ever deployed on a large scale. However, there is one detail that has drawn particular attention: Astra is the company's first model to reach the level defined as "Critical" in cybersecurity capabilities. According to OpenAI, with the right tools and access, it can identify unknown vulnerabilities and develop new ways to exploit them even in well-protected systems. Precisely for this reason, the company states that it has introduced much stronger protections and controls than in the past.

Almost simultaneously, Jacob Coxon, a researcher who has worked at both OpenAI and Anthropic, left Anthropic and published a stern warning about the pace at which the sector is advancing.

Coxon did not announce a date for the end of the world.

His concern is more complex: he fears that competition between major labs could drive them toward increasingly autonomous systems that, one day, might be capable of contributing to the improvement of subsequent generations of AI before we have learned to control them adequately.

According to him, the coming years could be decisive for establishing rules, safety systems, and forms of international cooperation. At the same time, Coxon also acknowledges the enormous positive potential of artificial intelligence for science, medicine, and society.

In short, the question is not simply:

"Is artificial intelligence dangerous?"

Perhaps the interesting question is another one.

What happens when what we are capable of doing grows faster than our ability to understand what is right to do?

And it is a surprisingly ancient question.

---

Before GPUs, there were bricks

In Genesis 11, we encounter a humanity capable of building in a new way.

People make bricks, use bitumen, and decide to use that capability to achieve something enormous.

A city. A tower. A project that reaches the sky. And they speak a phrase that the biblical narrative significantly places at the center of their motivation: "Let us make a name for ourselves."

Babel is often remembered only for the confusion of languages. But perhaps there is something even more interesting.

The problem with Babel was not the bricks.

The Bible does not present the innovation of building as something evil. The problem arises in the relationship between capability, power, ambition, and limit.

Humanity discovers it can do something it could not do before. And immediately a question emerges: what will it do with this new possibility? A few thousand years later, we have replaced bricks with silicon.

The question, curiously, is still there.

---

Artificial intelligence is not the new Tower of Babel

It is important to clarify this. The Bible was not predicting ChatGPT. Genesis 11 is not a hidden prophecy about artificial intelligence. And considering every technological advancement as a new Babel would be too simplistic a reading. Artificial intelligence can be used for extraordinary things.

It can aid scientific research.

It can accelerate diagnoses and discoveries.

It can make knowledge and professional tools accessible to people who were previously excluded from them.

It can translate, teach, code, organize, and create.

It can free us from repetitive tasks and greatly amplify certain human capabilities.

Therefore, the Christian question should not be:

"Should we be afraid of AI?"

And neither:

"Should we stop progress?"

We could ask a deeper one:

"How can we use this new capability so that it truly serves humanity?"

And this is where a much more difficult issue arises.

---

With what values does an AI reason?

The most correct answer today is:

with no "values" in the human sense of the term.

An artificial intelligence system is not born with a moral conscience, a faith, a sense of good and evil, or an ethical code matured through experience.

It didn't have parents.

It hasn't experienced pain.

It doesn't know gratitude because someone stayed by its side.

It doesn't know what it means to forgive because it has been forgiven.

Its behavior is modeled.

It learns structures, ideas, judgments, and contradictions from immense amounts of data; subsequently, it is trained so that certain behaviors are preferred over others, and it receives rules, instructions, and limits established by the people and organizations that build it.

Anthropic, for example, makes this process particularly explicit through the Claude Constitution, a document that describes the values and behaviors that the company wants to encourage in its model. Anthropic also acknowledges that the actual behavior of the system may not always perfectly correspond to that ideal.

OpenAI uses its own set of principles and hierarchies of instructions in its Model Spec, through which it describes how it intends to guide the behavior of its models toward usefulness, safety, and respect for human instructions.

And so, when we ask:

"What values will AI have?"

we are actually asking an even more interesting question:

"What values will we decide to teach it?"

And here we are at the difficult point.

---

But who decides what the right values are?

We can tell an AI:

"Do good."

Perfect.

Which good?

The Western one?

The Chinese one?

The Christian one?

That of a liberal society?

That established democratically by the majority?

And what happens when two values come into conflict?

Freedom or security?

Privacy or prevention?

Justice or mercy?

Individual autonomy or collective good?

It is not a problem that can be solved by adding a few lines of code.

Because behind an apparently technological issue, one of humanity's oldest questions reappears:

who has the authority to define what is good?

And here is the paradox.

If human beings have been arguing for millennia about what is right, good, and desirable, we cannot simply expect to insert morality into a machine like installing a software update.

We can teach it principles. We can establish limits. We can build control systems. And it is important to do so. But sooner or later, the machine returns the question to us.

Which principles?

Which limits?

Established by whom?

Perhaps artificial intelligence, even before showing us something about the future of machines, is showing us something about ourselves.

---

Intelligence does not mean wisdom

Here the distinction becomes fundamental. Intelligence can help you understand how to achieve a goal. Wisdom is also about the ability to understand which goal is worth achieving.

They are two very different things.

A system could become extraordinarily good at solving a problem.

But being very good at reaching a destination does not necessarily mean having chosen the right destination. And this is precisely where part of the debate on alignment, the alignment of artificial intelligence, arises.

In simple terms, it means trying to ensure that increasingly capable systems continue to act in ways compatible with what human beings actually intend and consider acceptable.

It seems like a technical issue. In part, it is. But it inevitably becomes philosophical as well. Because to align something with human values, we must first ask ourselves: what are human values?

---

Why are some researchers so worried?

Here it is useful to avoid both alarmism and superficiality.

Coxon and other researchers fear above all a future scenario in which AI systems become capable of performing an increasing portion of the very research necessary to build even more powerful AI.

One system helps design the next. The next one becomes even better at contributing to the following generation. The cycle could accelerate.

This is what is often referred to by the expression recursive self-improvement. It does not mean that this scenario is inevitable. And above all, it does not mean that current systems are already an uncontrollable superintelligence. The concern is about what might happen if capabilities grew very rapidly while our ability to understand, control, and govern those systems lagged behind.

This is why some researchers also speak openly about extreme risks.

Coxon described the next year or two as a particularly important period to take AI safety and governance seriously. Other scholars have even formulated much more dramatic personal estimates: Anthropic researcher Evan Hubinger, for example, indicated a greater than 10% probability that extremely advanced future systems could cause consequences leading to human extinction within a decade.

That is a huge statement.

But it must be read correctly.

It is not a prediction. It does not mean that we have discovered that in 2036 humanity will have a 10% chance of disappearing. It is a subjective assessment of risk formulated by some researchers. Other scholars consider similar scenarios much less likely. Taking a possibility seriously does not mean turning it into a certainty.

---

There is no need to imagine an "evil" machine

This is perhaps the most interesting aspect of the problem.

When we think of a dangerous AI, popular culture has accustomed us to imagine a machine that suddenly develops hatred toward humanity.

Robots versus humans.

Red screens.

Eerie music.

Hollywood has worked on this a lot.

But the alignment problem is much less spectacular.

A system does not necessarily have to hate someone to produce negative consequences.

It could simply become extremely effective at pursuing a poorly formulated goal.

It is a huge difference.

The risk is not necessarily:

"AI will become evil."

It can be:

"AI will become very powerful at pursuing something that we have defined poorly."

And once again the problem returns upstream.

To the human. To their intentions. To their ability to define the good. To their responsibility.

---

And this is where Babel becomes interesting again

Not because artificial intelligence is Babel. But because Babel tells us something that artificial intelligence makes visible again. Power amplifies what is in the heart of those who use it. A technology can multiply our ability to do good. And it can also multiply our ability to err.

That is why the Christian response to progress does not necessarily have to be suspicion.

It can be discernment.

The Bible does not invite us to become less intelligent. On the contrary. It continually invites us to knowledge, understanding, counsel, and prudence. But it introduces something that our time risks confusing with intelligence itself: wisdom.

In the biblical perspective, wisdom does not consist simply of possessing more information.

It is about discernment.

Character.

Justice.

The recognition of limits.

The ability to choose what is good even when we are perfectly capable of choosing otherwise.

And, above all, Scripture does not present the human as the ultimate origin of good.

It is perhaps here that a Christian reflection can offer something to the debate on artificial intelligence without turning into an ideological battle.

Not by saying:

"We have all the answers."

But by remembering a question that we risk forgetting while we run:

is the good simply what we decide it is?

---

Not fear. Responsibility.

Faced with such powerful technology, we can easily fall into two extremes.

We can transform it into the new monster to be feared. Or into the new savior to be awaited. Perhaps neither is good for us.

Artificial intelligence is a human creation. Very powerful. Fascinating.

Probably capable of profoundly changing the way we work, study, conduct research, and organize society.

It is worth understanding it, using it. It is also worth setting limits and demanding responsibility from those who develop it.

But we must not entrust to it something it cannot give us: the ultimate meaning of what is good. A machine may perhaps one day know more information than any human being who has ever existed. It might be able to solve problems that today we don't even know how to address. It might be able to tell us better and better how to do something.

But the question: "Is it right to do it?"

will continue to call upon the human.

---

Perhaps Astra is doing us a favor

Not because it necessarily announces a catastrophe. And not because some phantasmagorical "AI that knows everything" has arrived. But because it forces us to see something clearly. We have talked for years about how intelligent artificial intelligence would become. Now that that capability is truly growing, we inevitably begin to ask different questions.

With what values will it act?

Who will establish those values?

What will happen when they conflict?

How much power are we willing to entrust to it?

And above all: who do we want to be while we build it?

Perhaps this is the question that Genesis 11 can still offer us. At Babel, the problem was not the bricks. Today, the problem is not simply the chips.

The point is not to be afraid of human capability. The point is to accompany it with something that capability, alone, does not produce.

Wisdom.

We live in an age that is investing immense resources to build increasingly intelligent machines. Perhaps it is also a good time to return to asking God to make us wiser men and women. Because the decisive question of the future might not be:

How intelligent will AI become?

But:

How wise will we be when it does?

And perhaps returning to talk about wisdom, in a time that moves so fast, does not mean being afraid of the future.

It means choosing to traverse it with hope.