An application of artificial intelligence is radically changing humanity’s relationship with writing. The very famous ChatGPT3. His success has been such that his website has collapsed. And in a few weeks it has become paid. And it has generated unexpected mutations, such as God in a Box, which allows dialogue with the algorithm on WhatsApp, or Radio GPT, with automatically broadcast music and texts read by synthetic voices. And for a few days now the most advanced ChatGPT4 has been working.
GPT deep learning neural networks, developed by the company OpenAI since 2015, were in the minority until late last year. Trained with thousands of texts, they are capable of generating natural language. Each text they produce after receiving an instruction is different, without even a phrase copied from the internet. But GPT2 and GPT3 only generate short texts, with no connection between one and the next. With ChatGPT3, on the other hand, they have acquired continuity. The way has been opened so that soon they will be able to write hundreds of coherent pages. An essay, a collection of poems, a novel.
Since the Washington Post began publishing sports news written by an artificial intelligence called Heliograph in 2016, algorithms have been introduced into our daily writing. The option to auto-complete a word or a sentence, when writing instant messages or emails, coexists with chatting with customer service bots.
During those same seven years, books have been published that used these new technologies of artificial writing. In 2018, the American author Ross Goodwin published 1 The Road , which transcribed the conversations and the geographic coordinates that were captured by the sensors of a car during a trip. The popularization of GPT2 in 2019 and GPT3 the following year led to the sale of the first titles written in collaboration with algorithms on Amazon: from Pharmako-AI , by K. Allado McDowell, in English, to Non siamo mai stati sulla Terra, by Rocco Tanica, in Italian. But no more than a couple dozen books co-written with artificial intelligence existed until February. Then ChatGPT3 caused an explosion of nearly 200 self-published titles on Amazon. They don’t stop growing.
If the GPT2 could only write surreal texts, the GPT3, on the other hand, is already capable of writing sensible texts, just like its Google antagonist, LaMDA. GPT4 means a new qualitative leap, along with other neural networks that we still don’t know what they will be capable of. It is impossible to know how far we are from the generation of powerful literary texts, with a metaphorical dimension. In parallel, the automatic production of images is altering the world of work for designers and illustrators. Everything points to the coexistence of two circuits and parallel markets: that of works signed by men and women and that of artificial creation. Machines have been better than us at chess and go for a long time, but that has not meant the disappearance of games and championships between humans.
The confusion caused by the irruption of generative writing in professional fields such as journalism or content for social networks and web pages has updated the already classic equivalence between the Internet revolution and that of the printing press. “It’s a dangerous idea”, the prestigious professor, essayist and expert in the history of books Roger Chartier tells me, passing through Barcelona to present Imaginary Geographies (16th-18th Centuries) (Ampersand), a study on maps in editions Don Quixote, Robinson Crusoe and other modern classics. “Because it is completely different to read on paper or to read on a screen, it means going from a spatial and physical logic, in which there is continuity and there is a whole, to a topographical or thematic logic, by a classifying order of information, without a clear context ”.
The same is true between human and artificial writing. Of the two previous great cultural revolutions in the history of humanity, the transition from orality to writing and from handwritten books to printed ones, we only have the second well documented, so it is natural that we resort to it to understand better what is happening now. Perhaps we rushed into thinking that the internet was the great turning point and artificial intelligence will be instead.
Chartier has explained in his reference titles, such as The Order of Books (Gedisa), that the printing press did not suppress the circulation of hand-copied books, that they were cheaper and allowed additions or dodging censorship. I ask him if he believes that coexistence is also going to be typical of the 21st century: “Historians are the worst prophets, because they can only prophesy the past”, but “to ensure this coexistence, I think that educational, cultural and political institutions must defend the book on paper, publishers, bookstores and libraries, if we want those devices that are characterized by their architecture and by their nature of memory and legacy to continue to exist”. Not everything is online. And no large structure there guarantees long-term preservation of documents. This is why traditional libraries and archives remain essential.
“Have you tried ChatGPT3?” I ask. And he answers no, because it scares him, but that he has been interested in automatic translation: “They have sent me a translated book, for popularization, and, to my surprise, I have only found two errors, that is on the one hand fascinating, but on the other hand it will be catastrophic for the profession of translation”. Undoubtedly for the translation of manuals and best-selling novels. Complex literature and poetry cannot yet be translated by machines. Nor written. That is why perhaps the future of journalism is literary. Not only at the level of words and images, but also at the level of the gaze or the structure. Perhaps it is necessary to unite, as occurs in the metaphor, two concepts or two referents or two unexpected themes. Like artificial intelligence and an expert historian in the books of the 16th, 17th and 18th centuries. Because that wouldn’t do a machine. At least for now.
