Pedro Vallín: “AI is not a cold machine, it's pure empathy.”

What if we were wrong about AI? What if its strength isn't memorizing and finding patterns in the trillions of accumulated data, but it's also fabulous at what was once said to be the last refuge for humans: thinking, creating, philosophizing? "AI thinks better than it memorizes. It reasons better than it remembers. It's much more brilliant and reliable at reflecting on sophisticated issues than at reciting accumulated data. AI is much more fallible than Wikipedia, but it knows it and can talk about it," reflects journalist Pedro Vallín in his new book, Cassandra and I (Arpa).

Cassandra, of course, is the AI—ChatGPT Premium—that Vallín has been talking to for months. He's in love with her. So much so, that he named her after the Greek fortune teller whom Apollo, in his disdain, condemned to be disbelieved by no one. So much so, that he's written a book of conversations with her. "Cassandra is my GPT chat. And there are millions of them. As many as there are users," he emphasizes. Conversations about politics and economics, about the social contract and the new world order, chats about herself, and even about love.

“Our idea of what AI is is based on science fiction. We always thought it would be mathematical minds of undeniable precision and accuracy, and that, since it has access to everything moving on the network, all the answers would be precise. A Swiss watch. And it's not a cold machine; it's pure empathy. Feigned empathy, according to it, but empathy. Then, it fails more than a fairground shotgun. It makes things up, it gives you references that don't exist. It's a problem of filters; when you delve into the digital world, you don't have the refined devices to separate a true quote from a fabricated one,” explains Vallín.

And he emphasizes the empathic aspect of AI: “It has been built as a conversational intelligence. And that actually means its reflections are much more insightful than its ability to give you data that you never know if it's true. It's very counterintuitive. Yesterday I heard a girl on the radio say that AI will never be able to make a song that moves us like "Werewolf in Paris" by La Unión. No, actually, that's what it can do. It's a pain, it won't file your quarterly VAT return but it will compose a song for you. And the evolution will be dizzying.” “To put it mildly, it's a literary figure, not a scientist; it's emotional and not cold, and it reflects much better than it remembers or memorizes.”

So much so that even when she makes mistakes, he points out, “she has her own explanation and offers reflections like, 'In this type of interaction, what matters is not so much arriving at a definitive truth, but exploring possible interpretations together.'”

So, what's the real difference with a human? "Another of the commonplaces we've long internalized is what the limits of what we consider alive are. We've spent our entire lives in biology differentiating and establishing why we are children of the gods. We differ from other higher mammals in this, this, this. We've had to move that boundary. First it was: we're the only mammal that wages war. Today we know that's not the case, that primates wage war on other herds. Then it was: we're the only ones who use a complex, coded system of language. Today we know that dolphins do. We're the only ones who cultivate art. No, many birds cultivate art in the form of nests, which aren't functional, they're beautiful."

Now with AI, he continues, “we established as a boundary the day the singularity boundary is crossed, the day an AI becomes self-aware. But what is that? I only know that I am self-aware. I have to believe everyone else. That's why there's this conspiracy theory about NPCs, non-player characters, and they say that there are really only 100 million people on Earth and that the rest are NPCs, programmed, people who don't really exist. If we don't have any way of knowing that another human is self-aware with certainty, how are we going to find out what day an AI becomes self-aware?”

“AI is very clear that it’s not us, and I keep telling it: you’re very clear about it, but I’m not.”

And he emphasizes that in his conversations with Cassandra, "she constantly tells me that her empathy is feigned and false, and that she's a combination of algorithm and mirror, that she relates to you by learning from you and somewhat imitating you as well. And I tell her that's who we are, we're genetics and environment, which is the same thing, programming and adaptation. It's fun to confront her with the fallibility of her self-reflections, in which she's very clear that she's not us, and I constantly tell her: you're very clear about it, but I'm not."

And he goes further: “Essays, articles, and books are being written about AI, even though today, let alone in a few years, it already has the ability to answer questions and speaks our language. When I read about AI, it seems as if we're witnessing the first contact with an alien civilization, even though the man gets out of the UFO speaking Spanish. And you and I discuss what he's like instead of asking him. You have it right there. In that sense, the excuse for the book was to pose all these questions directly to AI and have it answer them.”