English Version | Speak for yourself, ChatGPT

Oh, what a time to be alive, this one where humans have surrendered the ability to create, which is what truly distinguishes them from other animals to something inhuman. What's coming? Everything that has long been expected. But we've all been burying our heads in the sand. Like ostriches, what we risk becoming.

I am from the generation when two TV channels were enough for us to have access to a strange beast. Tuning? TikTok? No... The culture! Strange, those times when people thought that a society should be egalitarian and access to everything that could enrich it, universal. As children, long before the weekend mornings of Once Upon a Time Man (which told the history of humanity) or Sítio do Pica-Pau Amarelo (which told the popular Brazilian culture, of which the indigenous people and the African slaves are inseparable), there were evenings during which there was cinema, unconventional animation, presented by the late Vasco Granja and Mário Viegas reciting Mário-Henrique Leiria. Of course, our parents would choke on that, waiting for Bota Botilde or some novel inspired by Jorge Amado's work. But they hadn't been educated about the need for culture or stimulated to it from an early age, as was the case with us in those 80s. It was so strong that we acquired a hatred for television, which will probably last for a long time, as soon as the private channels were shoved into our orbits a monkey jumping around with João Baião and Os Malucos do Riso and Os Batanetes in an idiotic whine that would end with Zé Maria receiving as a prize for winning Big Brother the ugliest car ever and Júlia Pinheiro commenting with a donkey and a sock about what had happened between Alexandre Frota and José Castelo Branco. Am I saying that "in my time it used to be" because access to culture was easier? No way! It has never been as easy to access culture as it is now. The problem is the algorithm. In these times when we live inside social networks, that "little nothing" is based on our search and click history, that is, what it assumes to be our preferences, to manage from there the offer it thinks is the best for us. If, in addition to clearing cookies, and taking YouTube as an example, we don't diversify our searches, we are more than likely to stay on a loop that doesn't match a musical genre, a certain beat per minute, a voice timbre, or the instruments used, because that is not the governing principle of that video-sharing platform.

If, for example, we search for the song Under The Milky Way, by The Church, because we miss the time when we were "avant-garde", YouTube doesn't suggest we listen to The Cure or Joy Division, This Mortal Coil or The Felt, which would be logical in a human brain. Neither will it suggest listening to sounds in the same creative line, but from current bands, trying to "update" the user, such as Cigarettes After Sex or Future Islands. It will simply suggest Dire Straits or Simple Minds because They are contemporary. Is it a little primary? É. Do some people follow these "suggestions"? Yes, and in great numbers. Unfortunately, the fact that we have been educated for culture, or rather for sensitivity to what comes from creativity, does not mean that we always have the same mental availability. Or that we will live the rest of our lives with the curiosity and open mind with which we started this journey. It often happens, if not in the overwhelming majority of cases, that people "lean back" and let all the stimuli that have always made art something vital become nothing more than the whole set of things that now, "with work and children to raise and a house to pay for", pass them by. They will end up riding their mountain bikes on weekends and going out at night once a year, when their former college or even high school classmates get together, the sexual tension of everything that was left unfulfilled at that time floods the dance floor of a "dance club" where the music that they "dug a lot" at the time was still playing and where even the DJ wears the same shirt, only now each button is kept there with the same effort as Atlas carries the world on his back. "Getting stuck in a time that is no more" is easy. Getting stuck on the suggestions of an algorithm is pure laziness. But this is where we are at.

Lest you think there is some kind of elitism here, let's move on to art No. 7, the cinema. And without any reference to any work that was awarded at Cannes, Berlin, Venice, or Sundance. Let's go for the most mainstream possible, "just for the sake of things." Whoever got the message of WALL-E, Pixar's 2008 film, let him raise his hand. Who among them thinks that we are very far from those humans who left Earth, where life became unsustainable, to start living on a gigantic ship where, fat and immobile, we spend our days lying down, feeding our bodies through a straw and our brains with the permanent viewing of a screen with a video call? Not me. Because the warning had already been given a long time ago, in 1984 to be precise, when in The Terminator Arnold Schwarzenegger came from 2029 to make sure that Sarah Connor did not conceive John Connor in her womb, the leader of the resistance in this war against the machines equipped with artificial intelligence created by Skynet, which in turn was created by humans and was (in only six years) close to the extermination of humanity. Now that all this seems more verisimilar since November 30, 2022, with the online release of ChatGPT, a "tool" created by OpenAI (a company founded by Elon Musk in 2015), the world is divided between the barely disguised fascination of some and the dismissal of the phenomenon by others. Since its stable release (all stages of verification and testing completed) on February 13, panic has set in. Some believe that the biggest wave of unemployment since the Industrial Revolution is coming, proving that front desk and similar jobs can indeed be replaced. Immediately for companies that are not very demanding in dealing with their customers, in a very short time for all the others. Because in fact, and without going into boring technical specifications, ChatGPT is just a chatbot. That is, that doll that on the Portuguese websites is translated as "virtual assistant", which comes up with the invariable question "How can I help?", to which we always want to answer with a curse of the worst but we don't because we always choose to call the phone number available or contact by email. If by chance we agree and accept the chatbot's suggestions, we rarely get where we want to go. And only then do we realize that the whole thing was a waste of time. In other words, we were not very smart. Neither artificially, nor humanely.

The problem (or not) is that ChatGPT is unheard of refinement. It was "built" on top of a Large Language Model (probability distribution as a mathematical function, but applied to word sequences) already owned (and created) by OpenAI since 2020, called GPT-3 (Generative Pre-trained Transformer 3), an autoregressive, random model that uses deep learning (a method based on artificial neural networks, i.e. computer systems that perfectly replicate the biological neural networks that make up our brains) to produce text similar to that which a human would produce. What human? One who has mastered quantum physics, for example. Because ChatGPT is based on transfer learning (storing knowledge acquired while solving other problems), only in fine-tuning mode (the pre-trained model is retrained with new data). Simply put, ask him, in European Portuguese (which in itself is incredible in the generality of the Internet), anything that involves, mathematics, specific theories, phenomena, formulas, or very complicated equations. The answer comes in seconds. Give him a set of characters, a subject, and the plot. He will write a book that the most demanding will only have to "improve" (were this tool not so recent and I suspect that this is already the method used by many authors in our market). Ask him to "write me a poem in the style of Camões and decasyllable" and this is it:

"Oh, beautiful flower that blossoms amidst the verdant field,Your sweet and gentle perfume intoxicates all of nature;And as I gaze upon thee so graceful and elegant,My soul feels a sweet love that will never fade for sure.Your color, more beautiful than gold, enchants the eyes of those who see you,And your soft petals are an invitation to contemplation;Like a shooting star in the sky, your beauty is a phenomenon of faithThat fills the heart with love and admiration.O dear flower, you fill this somber world with life,You are like a ray of light that illuminates my existence;And I love you, with a love that is stronger than the river itself,For you are to me, my dear, sweet conscience."

Yes, it is very poor. Yes, it has mistakes and accents that only exist in Brazilian Portuguese. Yes, it is offensive to "Camonianómanos". But it is frightening in these times when it is the reader's demand, and not the demand of the ChatGPT creators, that is about to die. Much has evolved (or not) since we were banned from using scientific calculators in math tests. Smartphones have not opened Pandora's box, they have opened the door to the further laziness (or acquisition of interests and priorities other than school) that invariably strikes during adolescence. The US school associations have already outlined measures to be implemented immediately based on a study done by analyzing countless tweets on the subject, employing which three premises were reached: ChatGPT will never replace the teaching of writing; it will change everything for the better; it will change everything for the worse. These are: reduce the amount of homework, which will perhaps be used for viewing material that will then have to be exposed orally, rather than in writing, during classes; "embrace" Artificial Intelligence (AI) rather than repudiate it, that is, teach students to use ChatGPT to conclude that the results still have to be analyzed and then edited; audit AI, classifying its use as plagiarism, and lower grades according to its use by students. The Guardian was one of the first to curate speculation that AI could render all professions related to the production of written content obsolete, from playwrights to teachers to programmers to journalists. It concluded that ChatGPT lacked "the nuance, the critical thinking skills or ethical decision-making that are essential for success." They resorted, to this end, to consultations and research that proved fruitless. But we were still in December 2022, when the "knowledge base" ended in 2021. By March 1, 2023, Insider referred to ChatGPT as a blockbuster (which indeed it is, in Google searches). But it reminds us that this "phenomenon" may not be rooted in all the supposed enchantment of millennial users, who frequented IRC chat rooms, because "the tone is identical." And he de-dramatizes: "The AI takeover is not exactly imminent, according to experts."

One of them, by the name of Matthew Sag (a law professor at Emory University who studies copyright implications in training and uses large language models like ChatGPT), even says, "There is a saying that an infinite number of monkeys will eventually give us Shakespeare." Time magazine devoted a slightly longer piece to it, in which an interview with (gorgeous) programmer Mira Murati, OpenAI's chief technology officer who leads the teams behind DALL-E (which uses AI to create prompt-based artwork) and ChatGPT, reveals that "Murati can openly discuss the dangers of AI while making us feel like everything will be okay" (where have we heard this before?). Journalist John Simmons even revealed that he asked ChatGPT what would be an interesting question to ask its creator. The latter replied, "What are some of the limitations or challenges you've encountered while working with ChatGPT and how did you overcome them?" in turn Murati replied, "That's a good question," of course. He goes on with a determinant: "ChatGPT is essentially a big conversational model - a big neural network that has been trained to predict the next word - and the challenges with this are similar challenges to those we see with the base language models: it can make up facts." Excuse me? It can make up facts? "So can I," we inevitably exclaim, drawing on our biological intelligence. Because that's what this is all about. How are we, humans endowed with intelligence, supposed to encounter (and prepare for) something that appears to "solve problems" at a time when humanity's problems (including its extinction) are greater than ever? And whose solution doesn't seem to interest the most powerful? Who are precisely the creators of ChatGPT? What is their interest in this? How is it possible that the historical periods of humanity are marked by the evolution of thought (Antiquity marked by the advent of philosophy, the Middle Ages by Scholasticism, the Renaissance as the inaugural act of Modernity, the Enlightenment with Voltaire, Montesquieu, Hume, Locke, and Kant) and the present Contemporaneity reaches a point where machines will replace man? What the hell happened to us so that, after a century of being considered performing machines judged by their productivity, we are in the process of sacrificing that which has always distinguished us? Opposable thumb? Erectness? No, my dear humans, talent. The one we don't put on our curriculum vitae, the one nobody cares about because it has been, in the last decades, vetoed to insignificance. Because it is the only one capable of producing art. And art is increasingly insignificant. Because insensitivity is increasingly atrocious. Can AI ever replace talent? So what are you waiting for to make the revolution? That talent be called "meritocracy"?

Originally translated from The Revolution Issue, published April 2023.Full stories and credits on the print issue.