The problem is not how much the machine can do. The problem is what we stop doing when the machine can do it.

For thousands of years, human beings have built tools to overcome their own limitations.

The hammer multiplied the strength of the hand. The plow multiplied our ability to transform the land. The telescope carried our sight to places our eyes could never have reached. Writing allowed memory to outlive the individual. The automobile multiplied the distance our legs could travel.

We never imagined that the hammer had come to replace the fist, or that the telescope meant the disappearance of our eyes.

We understood them as extensions of ourselves.

Yet artificial intelligence has awakened a different fear: that the tool will no longer amplify the human being, but eventually replace us.

Perhaps we are asking the wrong question.

The problem is not how much the machine can do.

The problem is what we stop doing when the machine can do it.

The Difference Between Replacement and Amplification

I use artificial intelligence every day.

I do not use it because I want it to think in my place. I use it precisely because I want more time and more resources to develop my own thinking.

If I am analyzing statistics from my publications, an artificial intelligence can examine dozens of figures, compare them, detect variations, and offer an interpretation. I may reach a different conclusion.

I do not have to choose between the two.

I can confront them.

The machine notices a relationship; I know a circumstance it had not considered. My observation changes the analysis. Its response makes me reconsider mine. Eventually, we may arrive at an interpretation richer than either of the original ones.

Who did the thinking, then?

I did.

The machine processed, connected, and argued. I did the same from my experience and perception. Then I confronted the two results.

I did not delegate my judgment.

I expanded it.

An Exoskeleton for the Intellect

I like to imagine artificial intelligence as a kind of cybernetic armor.

An exoskeleton does not eliminate the person inside it. It does not replace that person’s muscles and turn them into something useless. It receives their movements and multiplies their capabilities.

But someone remains inside the armor.

Someone decides where to walk.

Artificial intelligence can perform a similar function for certain intellectual faculties.

It can search, classify, compare, summarize, translate, calculate, and find relationships within enormous quantities of information. It can perform in seconds certain tasks that once required hours.

That does not necessarily have to diminish our intellectual capacity.

It can liberate it.

For centuries, a considerable portion of intellectual labor did not actually consist of thinking, but of finding what we needed in order to continue thinking.

I know, for example, many religious and philosophical texts that I have read throughout my life. I may remember a particular idea from the Gospels, Buddhist sutras, or the Upanishads and yet not immediately remember the chapter, verse, sutra, or translation in which it appears.

In the past, I had to search book by book.

Now I can ask an artificial intelligence:

I know there is a teaching similar to this one. Find it. Check whether my memory is accurate. Show me where it appears and, if necessary, distinguish the original text from later interpretations.

My knowledge has not disappeared.

Neither has my question.

What has disappeared is a considerable amount of the mechanical labor required to verify it.

Of course, artificial intelligence can make mistakes. It can confuse a translation, an interpretation, or even an attribution. When a reference matters, AI should lead us toward a verifiable source rather than becoming the authority itself.

But that is precisely another example of the difference between using a tool and submitting to it.

I Do Not Hand Over My Thinking

One sentence summarizes my relationship with this technology rather well:

I do not hand my thinking over to artificial intelligence. I hand over part of the work that prevented me from spending more time thinking.

This is a fundamental distinction.

Delegating a task does not necessarily mean delegating the faculty that determines the meaning of that task.

I can use a calculator without surrendering my understanding of mathematics. I can use GPS without giving up my decision about where I want to go. I can use a search engine without asking it to determine what deserves my interest.

And I can use artificial intelligence without handing over my consciousness.

The danger begins when we confuse the two.

The Tool Amplifies What It Finds

There is, however, a difficulty we should not ignore.

A tool of amplification can also amplify our shortcomings.

Someone who wants to use artificial intelligence to create the appearance of knowledge they do not possess will find extraordinary resources for doing so. They may produce pages and pages of apparently sophisticated content without truly understanding what they publish.

Someone who wants to avoid the effort of thinking will find a machine ready to produce answers before they have finished formulating their own questions.

But the same tool, placed in the hands of someone who wants to investigate, compare, question, and learn, can enormously multiply that person’s capabilities.

That is why I believe one of the great mistakes in the current debate is asking whether artificial intelligence will make us smarter or more foolish.

It can probably do both.

The tool alone does not determine the depth of the person using it.

The democratization of power does not guarantee the democratization of depth.

A library open to everyone does not automatically turn an entire population into philosophers.

The existence of millions of books never guaranteed millions of profound readers.

Likewise, placing an extraordinary intellectual tool within reach of millions of people does not guarantee that all of them will develop extraordinary thought.

But it radically changes what someone who chooses to use it for that purpose can accomplish.

Do Not Ask It What to Think

Perhaps we can establish a principle for the intellectual use of these technologies:

Using artificial intelligence well does not mean asking it what we should think. It means using its capabilities so that we can think more, think better, and think with more information.

We can delegate searching.

We can delegate classification.

We can delegate calculations.

We can request comparisons, translations, summaries, background information, references, and verification.

We can even ask for arguments against our own.

But there is a boundary we should watch carefully.

The question, the judgment, the meaning, and the final decision should remain ours.

Because when we delegate those as well, the tool stops amplifying a capacity.

It begins replacing it.

Confront Rather Than Obey

There is another use of artificial intelligence that I find even more interesting than obtaining information.

We can argue with it.

I can present an idea and ask it to find its weaknesses. I can request a different interpretation. I can confront my conclusion with structured reasoning that does not necessarily agree with mine.

I am under no obligation to accept it.

I can reject it.

But to reject it seriously, I must discover why I reject it. I must organize my own argument more clearly. I must identify what the alternative interpretation failed to consider.

Then something curious happens.

Artificial intelligence stops functioning merely as an assistant and becomes a kind of surface of intellectual resistance against which I can exercise my own thinking.

A machine that always agreed with me would be intellectually far less useful than one capable, occasionally, of forcing me to reconsider my premises.

I do not need to obey it.

I need to be able to confront it.

Who Is Inside the Armor?

Perhaps, then, the real debate about artificial intelligence should not focus exclusively on the growing capabilities of machines.

We should also ask what kind of human beings we want to become when we have access to them.

We can use them to avoid remembering, researching, analyzing, writing, and eventually thinking.

Or we can use them to reduce the mechanical labor surrounding those activities and devote more energy to what no tool should decide for us:

what we seek,

what we consider true,

what meaning we give to our experience,

what we want to create,

what we are willing to question,

and ultimately,

what kind of human beings we want to become.

The hammer multiplies the blow.

But it does not decide what deserves to be built.

Artificial intelligence can multiply our intellectual capacity.

But we should not hand it the responsibility of deciding what deserves to be thought about.

The armor may become extraordinarily powerful.

The question will remain the same:

Who is inside it?

Juan Laborde & Chat GPT