By EUGENIO BUCCI
The machine invites us to play a supporting role in our own history. And talks through silicon elbows
Yes, you know what ChatGPT is. Of course you know: you've read about it, you've seen news on the news and maybe you've even played with it. O chatbot developed by Open AI Institute exploded into the preferences of the interconnected masses. Version 3,5, which works by combining 175 billion parameters simultaneously, has already hit the mark of 120 million users. Version 4,0, on the way, has 1 trillion parameters.
People don't talk about anything else. The machines too: they talk about nothing else. That's right: now, machines talk, write about abstract topics, far beyond “take the second exit on the left” or “please wait in line, your call is very important to us”. Computers and cell phones were promoted to talking beings, and already with the air of thinking beings. The guy goes there and asks: “What line of Bocage can I quote to my girlfriend at dinner tonight?”. The business responds in writing. “How do I make an orange cake?” He teaches, in a second. “What is the difference between an emir and a caliph?” “Who was Aeschylus?” “How to reverse global warming?” “Is Vladimir Putin moved by the sad passions that Espinosa spoke of?” “Will trigonometry one day be expendable in geometrical calculations?”
The temples of knowledge are in an uproar. The cyber-oracle has addictive aspects, we already know, but it is the vicious aspects that most agitate the academic community. Students turn to him to write their homework. How is the teacher? How do you know if that text really belongs to the person who signed it? School evaluation methods are in check. Plagiarism has changed levels. Copyright for robots is on the agenda.
Classroom tests, based on the old pen and paper technology, are reborn. The most prestigious scientific journals in the world are quick to announce urgent editorial rules: they do not accept papers written by Artificial Intelligence (AI), although they admit to using the infamous Artificial Intelligence to better distribute “content” on the networks.
From now on, everything will be different. “What is coming is a flood of innovations and nothing in the past compares to what is about to happen”, warned Professor Glauco Garbix, from the Department of Sociology at USP, at the seminar ChatGPT: potential, limits and implications for the university, which took place at the Institute of Advanced Studies at USP, in São Paulo, last Tuesday. One of the most influential Brazilian researchers in Artificial Intelligence, Glauco Garbix says that the technologies in progress are not neutral “tools”: “they are not a mere screwdriver”.
In summary: the great mutation has already begun in the way human beings relate to each other and to knowledge, work, consumption and culture. Machines have not yet begun to learn to be people, but they have already begun to behave like subjects of language.
Problems in sight. If you were to ask psychoanalysts what distinguishes humans from other animals, they will say that only humans are subjects of language, unlike earthworms and calculators. An anthropologist will sketch an answer along the same lines. The man-animal stands out because he speaks and, by speaking, activates abstract representations and chains together propositions guided by moral values.
Well, ChatGPT does all that – or at least simulates it very well. Not that he's human, that's not it, but the differences between human and non-human are getting more and more blurred. If machines are beings of language (and if there are already people in the United States using conversational apps to do psychological therapy), what, after all, separates a person of flesh and blood from a chattering algorithm?
There are those who avoid debate and take refuge in technical claims. ChatGPT makes mistakes, they say, with relief. In fact, in the words of programmers and computer scholars, the gadget goes into hallucinations: it makes mistakes, induces errors, lies – and all this without blushing.
But is that actually a criterion for guaranteeing that the organism is not human? To err is inhumane? Since when? The other day, in an answer given to a doctoral student at USP, the digital prodigy came out with an “à Deus”, with crasis and all. What could be more human? We are dealing with a machinic Rolando Lero, a character who has the stamina to sustain obvious wild assertions. And he writes “to God” with crasis.
Others say that Chat shouldn't worry us because it's not really smart, it just pretends to be. For these, the artifact gives the impression of logical coherence, but does not think anything. Maybe they're right. However, the world is full of people who flaunt the intelligence they don't have. Exactly like GPT. Are they less human?
And so we are. With algorithms that speak (and, even worse, listen), as well as write (and even read), our irrelevance becomes even more undisguised. The machine invites us to play a supporting role in our own history. And speaks for silicon elbows.
* Eugene Bucci He is a professor at the School of Communications and Arts at USP. Author, among other books, of The superindustry of the imaginary (authentic).
Originally published in the newspaper The State of S. Paul.
The A Terra é Redonda website exists thanks to our readers and supporters.
Help us keep this idea going.
Click here and find how
