File | Artificial intelligence | Faced with AI, the publishers’ dilemma (3 articles)

Vigilance, a little worry, a relative openness, but above all a lot of questions. Here is the anthology of feelings shared by various players in Quebec literary publishing with regard to the imminent instillation of artificial intelligence in their field.


Moreover, a few drops have already percolated: manuscripts written with the support of this technology have recently been placed on the desks of some of them, such as at Boréal. “We have received some, but it’s only a handful, maybe two or three. Perhaps manuscripts were also sent to us without the use of artificial intelligence having been revealed,” indicates Renaud Roussel, co-director of the publishing house.

PHOTO FRANÇOIS COUTURE, PROVIDED BY BORÉAL

Renaud Roussel

The writings of candidates assuming the use of an AI were read and considered, but quickly discarded, failing to meet the publisher’s standards. “We have therefore never had the opportunity to think about the question for a quality manuscript, but it is certain that internal discussions are taking place, because this is likely to happen in the future,” anticipates M. . Roussel, convinced that AI will inevitably find its nest in contemporary literature, “because it defines the moment we are living in and the authors are there to grasp this reality”.

Would publishers therefore be inclined to publish a very good text composed entirely or in part with this technology? As is often the case, the answer is: it depends…

Creative actions and gestures

“I would never accept a manuscript generated in whole or in part by AI,” says Antoine Tanguay, publishing director at Alto, who dismisses it as a “finished product”; without slamming the door. “The creative process that it offers as a tool seems much more interesting to me. If AI is a creative companion, which generates leads or ideas, we will live better with a phenomenon already at work,” he says. Moreover, a perfect example has just been published under his banner, with the translation of Do you remember your birth? by Sean Michaels, where artificial intelligence is one of the central characters.

This approach resonates with Leméac, where literary director Pierre Fillion remains receptive to the phenomenon, but without feeding on illusions.

PHOTO MARCO CAMPANOZZI, LA PRESSE ARCHIVES

Pierre Fillion, literary director at Leméac

If it is an advance in the literature, we will consider it. However, for the moment, we are subject to a lot of trial and error, experimentation, trial and error, but also deception.

Pierre Fillion, literary director at Leméac

He says he is ready to consider the publication of a work built with the support of an AI, provided that the author’s part of creativity is particularly clear and distinctive, raising everything to a literary level. He draws a parallel with an author who would use Wikipedia, while reworking this raw material. “If he used sources from artificial intelligence, which are in fact amalgams of a multitude of sources, and if his intervention on the work he recovered is sufficiently notable and of good quality, it could be considered,” he illustrates.

Renaud Roussel, for his part, makes sure to play the balancing act, between duty of vigilance and reflection on possible artistic potential, placing the author’s approach at the center of the chessboard. If he perceives “a creative gesture” behind the partial integration of AI into a composition, he will not hesitate to seriously evaluate the proposal. “In this question of artificial intelligence, it is always important to put humans at the center. »

AI: artistic incompetence?

And it is precisely in humans that publishers have faith, unconvinced (for the moment) of the capacity of software to produce convincing literary material. Pierre Fillion evokes the example of the dismal failure of an AI which was asked to produce a piece in the style of Michel Tremblay. Result: the text was full of coronations, even though the author has not used them for decades. The robot had mechanically drawn on old, outdated writings by Tremblay.

It didn’t take much to persuade Caroline Fortin, president of Quebec America.

I am convinced that AI is very bad at literature.

Caroline Fortin, president of Quebec America

Mme Fortin, however, does not rule out the possibility that software will make progress in this area. She also doubts its usefulness as creative support: “I have a lot of difficulty saying that a literary author would use AI to enable him to move forward. I have a hold on that. On the other hand, if we are talking about a reference work, I think it is interesting to see how to use artificial intelligence to help obtain keywords, or title formulations, more quickly to market books. »

Pierre Fillion, from Leméac, agrees. “Writers are already complete in their writing autonomy, they have no need to deal with these amalgamations of texts,” he adds. But in the event that a writer, in collaboration with an AI to develop his idea, produces an exceptional work, Mme Fortin will not necessarily turn a blind eye. “Let’s just say I’ll have a talk.” But above all, the important thing in all of this is that the work is exceptional,” she warns, referring to “a gray area” whose guidelines are constantly under discussion.

A few acres of traps

Until a solid legal framework is put in place, caution and concern remain. As proof, several publishers invoked a specter named Anne Hébert, declaring that they feared the resurgence of a bad trick. Explanations. In 1998, a journalist from The Press sent 12 Quebec publishers a copy of the manuscript of Wooden rooms of the said writer, by altering the names of the author and the characters. None retained it – some arguing a “dated” text. Still marked by this hoax, some contemporary publishers told us that they were experiencing the same sleight of hand, but this time with a manuscript written entirely by AI, the true nature of which would only be revealed after a hypothetical publication. Especially since, to date, they do not have any software-produced text detection tool…

Pierre Fillion also mentions the recent case of the Japanese novelist Rie Kudan, winner of the most prestigious Japanese literary prize. During the presentation of the distinction, she revealed that 5% of her text had been written by ChatGPT, without retouching, leading to contrasting reactions in the community. And what about the publication of children’s books published by Amazon, composed largely with the same technology? Caroline Fortin gets the chills. “It’s the biggest threat in publishing. Not necessarily artificial intelligence per se, but how others will use it. Also, who are those who direct its development, what ethics do they have? » The director of Boréal tempers, declaring to rely on the relationship of trust with his authors, most of whom have fueled a long-standing collaboration. “Yes, we can publish books with AI. But I publish works rather than books,” says Mr. Fillion, from Leméac. “Artificial intelligence is very attractive, but it is full of dangers…”


reference: www.lapresse.ca

Leave a Comment