When ChatGPT challenges libraries

The new chatbot ChatGPT that Microsoft wants to integrate with its Bing search engine is an important step in expanding the library paradigm. He promises to answer any question with credible answers himself. Although it can raise fear, as explained in The world Artificial Intelligence (AI) specialist Laurence Devillers (1), regarding his ability to create mass fake news, has threatened to return Google’s librarianship and its answer lists to the Stone Age. By combining billions of data and cognitive processes, it achieves on an unprecedented scale what the Library of Alexandria pioneered: the accumulation and classification of knowledge, the globalization and socialization of knowledge. The question of the role of libraries is being posed with new intelligence.

Many librarians already asked this question in the 1970s and 1980s when the computerization and networking of catalogs opened the prospect of an extension to the contents themselves and their uses. Moreover, I remember that I organized at the BM in Lyon, in 1988, a two-week meditation with computer scientists, neuroscientists, philosophers and artists, around the “second AI revolution” and networks of artificial neurons which will lead to the success of deep learning.

I am a librarian in a city that marked the history of printing through its economic and technical dynamism, in a library designed by the great book historian Henri-Jean Martin. What interests me is to expand on the latter’s often very daring reflections (especially in his later writings) and imagine what my profession will be in the dawning digital age. Hence the title of these meetings: nature of thought (2), that is, thinking as a natural phenomenon, at the same time technical, cognitive and collective, far from an idealistic approach that returns the reader and reader to an illusory sovereignty of self-awareness.

35 years later, this illusion continues to lose ground as the digital ecosystem takes on a larger human component and highlights its own redundancies and its capacity for invention. No need to have expertise in AI and human-machine hybridization to see the effects. Even the field of culture, however, eager for diversity, cannot hide its repetitive mechanics (of which cultural marketing is only one aspect). In fact, it is increasingly difficult to escape the Mallarmean feeling that culture inhabits and seems to swallow itself, at the same time, moreover, as the earth system in the process of artificialization. Paradoxically we are entering an era of all possibilities and extreme limitations that can only question our individual or collective use of thought.

Does this mean, as is evidently suggested in his latest novel Aurélien Bellanger (3), that we are doomed to nostalgia for a 20th still a creative century in which great bibliomaniacs like Walter Benjamin were able to save the true “aura” of artwork from mechanization? In truth, the library experience has always trained us not to fall into the trap of authenticity, to compensate for the inevitable melancholy of accumulation and repetition through the excitement of a constant initiation of exploration. our own limitations. But, this wisdom of the librarian who always knows that the search for truth in texts (or data) is endless needs to be recreated on the scale of the fully librarianized world of ChatGPT.

So what can our modest neighborhood libraries do? It is clear that the project of changing many knowledge documents thanks to indexing devices is increasingly going beyond libraries, including university libraries depending on a digital offer beyond them. But the globalized information technostructure that replaces it does not deliver free-wheeling super-knowledge, detached from everyday life. On the contrary, it has not been enriched by the most down to earth and best interactions, far from the esotericism of the old monastic libraries that Umberto Eco loved. This is why we now speak of an ecosystem, meaning a dynamic whole based on the interaction of its parts, of all its parts including … libraries.

They are not the remnants of an evolution that would make them obsolete. They occupy a strategic position if they have the ambition to do so, at the interface between readers (very different from their expectations or their levels of expertise), seasoned librarians and the vast ocean of information and Culture.

First of all, libraries obviously have to rely on books. They have admirably been able to adapt to the general conversation through various interposed media and play in this arena the role of important relays, at the same time sources, digests and reminders of exchanges. Then, they know how to fire any wood, moving all means of expression and all mediations. Last but not least, they are public spaces where individual relationships in information networks are almost part of a collective, open approach. More than school perhaps, they are places of adaptation to the new information ecosystem.

This adaptation is exactly what Laurence Devillers calls for by suggesting not to shame ChatGPT but to learn how to use it to avoid its pitfalls and take advantage of it. On the other hand, in the same columns (4), Eric Sadin digs the Rousseauist furrow of a state of the nature of language, indescribable in paradox, which would be radically incompatible with its mechanization (through the necessary evil powers). His method of opposing the irreducibility of thought to its technical externalization is not new, since the Platonic critique of artifice. But, it becomes less plausible in light of the discoveries made by thought about itself by studying its own biological mechanisms and by implementing itself in technical or organizational device (for example, libraries). Faced with this liberating evolution, nostalgia for the lost paradise of pure creativity is as empty as any tautology.

There is no question, however, of denying the dangers of new applications like ChatGPT. They can, through their ability to imitate, not replace people but, on the contrary, allow them to express their darkest passions. It is surprising, moreover, that the recent “reconstruction” of The world and IRCAM, thanks to an AI, the voice of General de Gaulle reciting the speech on June 18 that no one cares ever does not raise more questions. However, this opens the door to less holy recreations of historical fact. This example shows the need to educate oneself in the critical use of the new cognitive engineering, as recommended by Bronner reports in Enlightenment in the digital age (2022).

Beyond engineering, it is also about our environment, an environment that is becoming increasingly mentalized and humanized. We shape it in our image as we shape the landscapes we inhabit. It can result from this mise en abyme of a philosophical sense of a “what’s the point” and the temptation to return to a very personal wisdom, to practice a kind of spiritual descent. It is to fail to see that the most intimate thinking comes from the dynamics of linguistic, cultural, social and technical networks. However, if there are organizations at the center of this dynamic, they are indeed libraries, these institutions where personal thought is fed by everything else. True, their very powerful paradigm now extends beyond their places of origin, but it is precisely for this reason that these places, even modest ones, in the articulation of the everyday that life and the high sea of ​​data, maintains a very special responsibility.

Also, the sophistication of ChatGPT should not prevent librarians from joining readers in its rational taming, for finding reliable information, or even for content creation. Only by dealing with this type of cyber-tools can libraries continue to be windows to the world and schools of knowledge.

(1) GPT chat. Laurence Devillers in The world from January 22-23, 2023

(2) Nature of thought. Acts of the days organized by the municipal library of Lyon

(3) Aurelien Bellanger: Twentieth centuryGallimard, 2023

(4) The illusion of a language natural. Eric Sadin on The world from January 22-23, 2023

Leave a Reply

Your email address will not be published. Required fields are marked *