Generative Artificial Intelligence: between Borges and the return of orality

di Enrico Nardelli

(versione italiana qua)

Everyone — or at least I hope everyone — is familiar with Jorge Luis Borges’ famous short story The Library of Babel (one of the most beautiful intersections of literature and mathematics). In the story, the brilliant Argentine writer imagines a library that extends infinitely in every direction, containing an infinite number of books with every possible combination of characters. In this way, it holds all the knowledge of the world, even that of which we are not yet aware, along with all its variations and contradictions.

During a recent interdisciplinary panel organized by colleagues at the University of Milano-Bicocca, where we discussed generative artificial intelligence (GenAI) and higher education, as well as how to handle the presence of GenAI in the production of theses, I first observed that over the past twenty years the process of writing theses has evolved. With the ubiquitous spread of digital technologies, students have shifted from spending long days in libraries to navigating the web through search engines. This was a huge leap forward, putting the libraries of the entire world at our fingertips. I then added that we are now witnessing another leap, at least of the same magnitude and probably even greater, since GenAI — of which ChatGPT is the most well-known example — not only brings us this knowledge but is also capable of extracting, summarizing, and linking it together into coherent arguments

In a sense, with generative AI, it's as if we now have access to an "active" library with superpowers. This is a transition similar to the one I described in my book La rivoluzione informatica, which occurred when we moved from the passive knowledge contained in books to the active knowledge in software programs—what I called "knowledge in action." With books, it was people who, after absorbing the content, could put that passive knowledge into action in the world. With computer programs, anyone can achieve the same effect by simply running the program on a digital device, even without having assimilated the knowledge it contains. Now, with generative AI, you can produce new texts on subjects you know little or nothing about.

Unfortunately, there's a problem: GenAI sometimes fabricates facts and cites non-existent works. This phenomenon has been called "hallucination," by analogy with the same term used for people, and it represents the main Achilles' heel of this powerful technology, demanding careful attention when it is used in educational settings. In other respects, I think Borges might have appreciated generative AI's ability to mix and combine fragments of existing texts, producing endless variations. I'm sure that creatives in every field, despite being rightly concerned about the thorny copyright issues generative AI is raising, appreciate these capabilities.

During the panel discussion, my colleagues and I debated what changes were needed in university teaching methods. The situation for schools is different, and I'll address that in a future article. From my perspective, which views computer systems — including GenAI — as "cognitive machines" that can amplify our purely rational and logical abilities, I don't think it makes sense to forbid a student from using a tool that reduces the mental effort of producing a certain piece of work, as long as they have mastered the specialized knowledge of the subject.

Provided, and I emphasize this because it is the central point, that the student has mastered the subject. Only then can a student recognize if something the GenAI has produced is an "hallucination." At a high level of abstraction, using GenAI systems is not very different from using a library. The ultimate responsibility for a text always lies with the author, regardless of their source material—be it a standard library or an "active" library with superpowers. If a student fails to find the most appropriate articles or volumes in a library, or if they copy text verbatim without proper citation, they have still done their work poorly. Similarly, if they use generative AI without the critical oversight that comes from a deep knowledge of the subject, they are making the same kind of mistake. Perhaps it's an even more serious mistake, because the tool is more powerful, and as a famous comic book character often says, «with great power comes great responsibility».

However, we agreed that, especially for bachelor's degrees, the traditional final paper or thesis may no longer make sense. The effort of putting it together from source material—the very process that helps students assimilate and internalize the knowledge—can now be almost entirely delegated to generative AI, which removes the potential benefits of the exercise.

Instead, I believe it's important to shift the emphasis back on a person’s ability to demonstrate, “on the spot,” the skills that the thesis was traditionally meant to prove. As the moral of Aesop’s fable reminds the braggart who boasted of a miraculous jump in Rhodes, «Hic Rhodus, hic salta» (= Here is Rhodes, jump here). If you can truly do it, you can do it anywhere.

I am convinced that the advent of GenAI marks a historical moment that, in some ways, mirrors the one that occurred about 2,500 years ago when writing overtook speech as the primary means of communication. That transition, which we can read about in memorable pages of Plato's Dialogues from the fourth century B.C., was the first step toward the spread of culture and the progress of humanity. However, in the Phaedrus, the Athenian philosopher recounts Socrates' view that written language is inferior to spoken language, especially when speech is used dialectically. Socrates argued that a written text cannot respond to questions from a reader and essentially serves only as a reminder for those who already know things, whereas spoken words are "in the soul" of the speaker and are thus capable to help those who need to learn.

I believe the availability of generative AI tools can restore power and value to orality, given that these tools enable anyone to produce written texts that seem to certify the possession of any skill whatsoever. If everyone can create these texts, their mere existence no longer proves anything, much like Socrates's critique of the written word. But if the author is questioned and challenged, it will become clear who truly possesses knowledge "in the soul" and who can only repeat "the same thing over and over." This is why I think academia will return, to some extent, to the very tradition practiced in the first Academy in history.

Of course, writing, especially with the incredible boost from Gutenberg's invention of the printing press in the fifteenth century, will always play a vital role in spreading knowledge.

But I find it beautiful that, to some degree, the latest and most revolutionary innovations in digital technology are giving new value to something very old—something only people can have: human connection and oral dialogue.

--
The italian version has been published by "StartMAG" on 14 October 2023.