This thesis examines the evolution of automatically generated literature as a long continuum in which authorship, creativity, and human linguistic expression are repeatedly redefined. It traces the origins of automated writing to early combinatorial philosophies: Lullo’s Ars Magna, which envisioned a mathematics of thought; Leibniz’s attempt to build an “alphabet" of human reason; and Swift’s satirical writing machine in Gulliver’s Travels, which mocked mechanically produced knowledge. These speculative ideas became concrete in the 20th century with the advent of the computer, giving rise to early literary experiments. In Italy, Nanni Balestrini used computational recombination to amplify rather than replace authorial creativity; Primo Levi explored the unsettling implications of machine authorship; and Italo Calvino conceptualized the writer as a “writing machine” shaped by combinatorial processes. Looking beyond Italy, movements such as OuLiPo, the German stochastic school, and early computational works like Strachey’s Love Letters collectively challenged traditional authorship by externalizing creative constraints. A major shift occurred with the rise of modern AI: from symbolic, rule-based systems to data-driven statistical models culminating in the Transformer architecture. The thesis assesses the ethical and cultural consequences of this transition through the lens of «algoretica», addressing algorithmic bias, opacity, and the diffusion of responsibility. It also considers AI’s ambivalent role in human creativity—as both an enhancer and a potential source of homogenization—and engages with debates about whether LLMs merely parrot syntax or exhibit emergent behaviors suggesting a form of non-human semantic processing. The linguistic analysis introduces “IA-taliano,” a new variety of Italian shaped by generative models, marked by stylistic formulaicity, register inconsistency, Anglophone interference, and recurrent lexical patterns. These features are illustrated through machine-generated texts and human–AI collaborations, from Rocco Tanica’s interaction with GPT-2 to Giuseppe Croci’s stylistic use of modern LLMs. Ultimately, the thesis argues that the progression from Lullo’s combinatorial wheels to contemporary LLMs is not only technological but conceptual: a continual re-negotiation of what it means to write, create, and express human thought in an age of automated language.
L’attuale entusiasmo per le IA generative, come ChatGPT, non è un fenomeno improvviso ma il risultato di una lunga storia culturale che affonda le sue radici nelle antiche aspirazioni umane a creare macchine capaci di pensare e scrivere. Da Raimondo Lullo e la sua Ars Magna, passando per l’alfabeto del pensiero umano teorizzato da Leibniz e per la satira di Swift sulla macchina generatrice di discorsi, fino alla rivoluzione avuta grazie al computer nel XX secolo, si costruisce un percorso che porta alle prime forme di letteratura automatica. Con la macchina di Turing e i primi computer, la scrittura generativa diventa realtà, inaugurando un filone sperimentale basato su combinazioni e algoritmi. In Italia, figure come Balestrini, Levi e Calvino offrirono contributi decisivi, interrogando il rapporto tra automazione e autorialità. Questa fase pionieristica, pur controllata dagli autori, prepara la svolta successiva: la transizione dall’IA simbolica ai moderni modelli linguistici di grandi dimensioni. Questi ultimi, fondati sulla predizione statistica e non sulla comprensione semantica, producono testi fluidi, ma sollevano interrogativi etici, culturali e linguistici. La diffusione dei LLM richiede oggi di ripensare concetti come creatività, responsabilità e autore, soprattutto alla luce dell’emergere dell’IA-taliano, un nuovo standard linguistico che ridefinisce l’italiano scritto. La tesi si propone quindi di analizzare la storia della letteratura generativa e le sue radici filosofiche e artistiche; la rivoluzione tecnologica dei LLM e le sue implicazioni etiche e culturali; le caratteristiche linguistiche dell’italiano generato dalle IA e il suo impatto sulla produzione letteraria contemporanea. In definitiva, lo studio vuole offrire strumenti critici per comprendere come l’automazione della scrittura stia trasformando il rapporto tra linguaggio, creatività e identità autoriale.
«Io leggo dunque "esso" scrive». Considerazioni sulla narrativa generata dai modelli linguistici di grandi dimensioni
FIORENTE, ILARIA
2024/2025
Abstract
This thesis examines the evolution of automatically generated literature as a long continuum in which authorship, creativity, and human linguistic expression are repeatedly redefined. It traces the origins of automated writing to early combinatorial philosophies: Lullo’s Ars Magna, which envisioned a mathematics of thought; Leibniz’s attempt to build an “alphabet" of human reason; and Swift’s satirical writing machine in Gulliver’s Travels, which mocked mechanically produced knowledge. These speculative ideas became concrete in the 20th century with the advent of the computer, giving rise to early literary experiments. In Italy, Nanni Balestrini used computational recombination to amplify rather than replace authorial creativity; Primo Levi explored the unsettling implications of machine authorship; and Italo Calvino conceptualized the writer as a “writing machine” shaped by combinatorial processes. Looking beyond Italy, movements such as OuLiPo, the German stochastic school, and early computational works like Strachey’s Love Letters collectively challenged traditional authorship by externalizing creative constraints. A major shift occurred with the rise of modern AI: from symbolic, rule-based systems to data-driven statistical models culminating in the Transformer architecture. The thesis assesses the ethical and cultural consequences of this transition through the lens of «algoretica», addressing algorithmic bias, opacity, and the diffusion of responsibility. It also considers AI’s ambivalent role in human creativity—as both an enhancer and a potential source of homogenization—and engages with debates about whether LLMs merely parrot syntax or exhibit emergent behaviors suggesting a form of non-human semantic processing. The linguistic analysis introduces “IA-taliano,” a new variety of Italian shaped by generative models, marked by stylistic formulaicity, register inconsistency, Anglophone interference, and recurrent lexical patterns. These features are illustrated through machine-generated texts and human–AI collaborations, from Rocco Tanica’s interaction with GPT-2 to Giuseppe Croci’s stylistic use of modern LLMs. Ultimately, the thesis argues that the progression from Lullo’s combinatorial wheels to contemporary LLMs is not only technological but conceptual: a continual re-negotiation of what it means to write, create, and express human thought in an age of automated language.| File | Dimensione | Formato | |
|---|---|---|---|
|
TESI - FIORENTE.pdf
accesso aperto
Dimensione
1.11 MB
Formato
Adobe PDF
|
1.11 MB | Adobe PDF | Visualizza/Apri |
È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/32521