ChatGPT as seen by Gilles Herman

ChatGPT as seen by Gilles Herman

A text by Gilles Herman publisher

How will AI affect book publishing?

When people talk about artificial intelligence (AI), it’s easy to get carried away and give in to the urge to see AI as either the end of critical thinking or the beginning of some kind of new cyberhumanism. Let’s keep our feet firmly on the ground and take an unbiased look at what’s happening.

Far from being conscious entities on the verge of awakening, AI solutions are simply new tools enabling our creativity. Yes, they’re advanced tools but this isn’t the first revolution that humanity has lived through.

To start with, the expression “artificial intelligence” is already a misnomer. Until Siri actually succeeds in playing the song I want the first time I ask, allow me to remain skeptical.

“Algorithmic intelligence” would be a better term. It helps us remember that behind all the hype, there’s nothing but computing power based on learning from human creative work. With that in mind, we can then wonder about the role those tools will play in our professional fields.


Book industry embraces technology

Contrary to what you may think, the book industry is very forward-looking and is often a leader in adopting new technologies. As early as 2008, a partnership between the Association nationale des éditeurs de livres (ANEL) and De Marque led to the creation of a digital warehouse that today distributes our literature all over the world, offering innovative options for digital lending and accessibility in public libraries and schools.

As you can see, we’re no stranger to technological challenges and we know how to take them on.


Testing the limits

ChatGPT has been a huge media story in recent months. Like many people, I was interested in testing its limits after I had played around with images on DALL·E. I asked ChatGPT to draft all sorts of texts ranging from bios to book blurbs to chapter summaries and I quizzed it on its factual knowledge.

It’s important to avoid being distracted by all the mistakes the chatbot makes. The key isn’t the accuracy of its answers (which ultimately depend only on its access to databases) but rather its understanding of the issue raised and its attempts to address it. The process is truly impressive and fascinating.


Should we be concerned?

Definitely not. Again, the deep learning hidden behind its prowess comes from the astronomical volumes of data that the algorithm can process. In the end, its creations are never more than regurgitation and mimicry of what truly creative human intelligence can achieve.

No matter how powerful they may be, these networks will never do original research.

As a history book publisher, I’m always looking for original information about characters, events and social trends to shed new light on something that’s little known or misunderstood. It’s no coincidence that this often comes after much thought and extensive archival research supported by countless hours of reading.


Searchability and discoverability

Nevertheless, algorithmic tools are already emerging to support research work. An example is software that’s able to decipher manuscripts with illegible handwriting. Like Jean-François Champollion who cracked the code of Egyptian hieroglyphs, this tool helps us understand documents and supports our creative efforts.

What’s more, it scans the characters in thousands of documents, making them available for keyword searches.

Here’s another example: since 2018, a group of dedicated experts in Quebec has been developing a service called TAMIS designed to generate metadata by applying algorithmic approaches to document content. Basically, the TAMIS tools process the text and images and then come up with relevant keywords, descriptions and subjects relating to the content.

When the metadata is supplied to the search engines, discoverability is greatly enhanced.


Technical progress

This text was written first by a human, then reviewed using Antidote, the highly efficient spelling and grammar checker. But it will subsequently be re-read by an individual who will notice all sorts of discrepancies that will need to be corrected.

In the same way that tractors weren’t a threat to agriculture but instead transformed it, algorithmic tools will help us work better and more efficiently, playing a support role for our writing rather than being a panacea.



Gilles Herman is the general manager of Editions du Septentrion and vice-president of Copibec. Trained as an engineer, he entered the publishing world through technology and now tries to make the link between the past and the future.

Subscribe to our newsletter to continue learning about copyright and intellectual property