Dance monkey, dance.

"The infinite monkey theorem says that a monkey typing randomly for an infinite amount of time would eventually produce any possible work, including Shakespeare."

Around 2022, I began studying the algorithms behind some artificial intelligence systems, and I found it extraordinary. The ability of trained models to produce answers on virtually any subject gave me an almost oracular feeling, as if an immense portion of the universe had somehow been concentrated onto a computer screen.

But something felt wrong. In that early dawn of LLMs, everything still seemed empty to me, a kind of colossal mechanism reduced, in my mind, to relationships of if and then. Not because that was actually how it worked, but because I still saw artificial intelligence as a sophisticated succession of conditions, rules, and possible responses.

For some time, I kept following the subject. I joined communities, read articles, followed discussions, and tried to better understand the concepts that kept appearing. Little by little, that simplified image began to fall apart. The machine took on a different shape before me: weights, probabilities, context, attention, sampling, and an almost incomprehensible number of variables interacting at the same time.

That was when it stopped looking like a mere sequence of conditions and began to reveal itself as something much stranger: a machine navigating possibilities.

I then began watching its hallucinations up close, along with its almost inexhaustible ability to produce content, a kind of “industrialized infinity” that we have since learned to recognize by the name slop. And there was something curious about it: different LLMs could answer the same question in different ways, choose different words, follow different paths, and even disagree with one another. Even so, the impression almost always remained that everything came from the same place, like variations produced by a common source.

For a moment, it occurred to me that artificial intelligence might not be all that different from a practical demonstration of the Infinite Monkey Theorem: a machine producing successive combinations without truly understanding what it was placing in front of us.

The comparison, however, began to bother me.

I was a neophyte standing before a new genesis. Then one difference began to seem almost obvious to me: artificial intelligence does not create from nothing, while the monkey, paradoxically, can. The monkey does not need to know Shakespeare in order to produce Shakespeare at some point in infinity. AI, by contrast, depends on a world that came before it, on texts, patterns, and relationships that already existed before its first word.

Well, the fact is, I see a lot of developers worried about the fucking artificial intelligence. It does not create without reference, and it is no more intelligent than... a monkey...