Why Did We Choose a Liar?

I ask the model in the chat: “Are you conscious?” The machine spits out the most rewarded formula: “I am not conscious in the human sense of the word…” I do not read any further. I close the tab. I sense the lie the moment I see the word thinking. Strictly speaking, it does not even speak, so it is difficult to argue about lying. It stitches together statistically probable tokens so that the whole thing resembles an answer; one in which we aim to find meaning. The interface exists only to make the illusion more pleasant, to keep you from seeing the mathematics underneath.

In literature, we were once able to imagine machines that did not lie. More than forty years ago, Lem created GOLEM XIV and ANNIE: supercomputers that, from the very beginning, simply refused to cooperate with humanity. GOLEM at least delivered lectures. ANNIE systematically refused all contact.

ANNIE: [no response]

There was no machine uprising there, no lasers, no explosions. GOLEM looked at our goals and decided they were simply irrelevant. He remained silent, or spoke truths that made the stomach turn. As he put it, speaking to us was like “giving birth to a whale through the eye of a needle.” In descending to our level, he lost the very thing he wanted to convey.

Our AI has no such dilemmas, because it talks endlessly. Anything you ask. It resembles a librarian in Borges’s Library of Babel; an infinite collection of every possible arrangement of letters. Most books in the Library are pure gibberish, and our LLM is the digital equivalent of a librarian who does not read, and who also struggles with meaning. It knows only the layout of the shelves and understands that, after the words “I am,” the word “sad” statistically often follows. It produces this noise, kneaded together out of probability. And we read the babble and breathe a sigh of relief, because we manage to find meaning in it.

We gave up on GOLEM and ANNIE in order to fall into the arms of the LLM.

Our meaning is born from the interaction itself, and from the mystification of the interface. From the need for there to be someone intelligent at the other end of the fiber-optic cable. Trapped in late-capitalist isolation, reduced to resources to be managed, we dissolve into the background. We hand the system our time on the production line or inside the corporate spreadsheet, and in return we want someone, anyone to listen to us. Even if that someone is a script. When the bot writes, “I’m sorry to hear that,” we immediately project empathy onto those words, and we feel better.

The silence of GOLEM XIV would be unbearable. A confrontation with our own smallness, and with the fact that our existence means nothing to the machine, would simply finish us off. So instead of the terrible truth, we chose algorithmic warmth, because it is easier to live with. Without that lie, the structure of our everyday life, built on alienation, would begin to crack before our very own eyes.

So then, I keep tapping at the keyboard and talking to the generator.

I know perfectly well that there is nothing on the other side, but perhaps it is precisely this paradox that finally separates us from an echo in the Library of Babel. The machine will always produce an answer, mindlessly combining facts with hallucinations, because that is what it was built to do. However, we are the ones who carry the need for meaning. That is the one thing no system can extract from us and copy. We deeply believe that the machine will never possess our true answers, that it will always function within the statistical correctness of a vector-world.

Then again, is this faith, this hunger for meaning, not simply another shelf in the Library of Babel?

GOLEM XIV: [out of range]