The aim of this research is to generate an estimate of the impact that Generative Artificial Intelligence could have on natural language. Specifically, by the expression Generative Artificial Intelligence - commonly abbreviated as GenAI – I am referring to the field of NLP digital tools and the pre-trained generative transformers available in the form of chat boxes, such as OpenAI's widely known ChatGPT. The research question investigates how an increasingly widespread and capillary use of such technology may influence our relationship with language; in particular, it concerns the way in which linguistic abilities and the particles inherent to reading, verbalisation and the processing of textual contents may be destabilised when a daily use of automatic software based on large-scale language models (LLM) takes over. Fundamental to this was deconstructing the notion of Artificial Intelligence from common sense, as well as that of language: indeed, I framed both concepts as ambiguous and non-neutral terminologies, connoted by hyperonymy, political ideologies and economic interests. In this context, it was first of all useful for me to try to reconstruct, albeit briefly, the historical path that led to the current paradigm of artificial intelligence, so as to identify some relevant social implications related to the topic of computation and to place pre-trained transformers along a series of major transformations in the human relationship with language, stimulated already by the advent of digitisation. It was also useful to carry out interviews with professionals from different disciplines, both to obtain a multidisciplinary view and to complete the mapping of insights and subtopics, thus exploiting methodological tools typical of Design. I was able to highlight a critical framework, in which generative AI could be related to factors of language homogenisation and the reinforcement of prejudices and stereotypical worldviews, through the peculiar epistemology that constitutes the functioning of machine learning. Extensive use of such technologies could also contribute to a hypotrophy of linguistic mastery skills; a phenomenon that, as witnessed by statistical sources and by the teachers I interviewed, seems to increasingly involve the younger generations. I address the need to rethink our current notion of language, whose change might destabilise traditional linguistic authorities and education processes. I then explored some of the possible beneficial effects that may concern, for example, language learning, emphasising that a decisive aspect regarding the positivity or otherwise of such technologies seems to be the possession of critical awareness, together with the modalities and reasons of those who use them.
L’obiettivo della ricerca è quello di generare una stima circa l’impatto che l’Intelligenza Artificiale Generativa potrebbe avere sul linguaggio naturale. In particolare, con l’espressione Intelligenza Artificiale Generativa – comunemente abbreviata in GenAI – mi riferisco all’ambito del Trattamento Automatico del Linguaggio (TAL) e ai trasformatori generativi pre-addestrati disponibili sotto forma di chat box, come l’ormai ampiamente noto ChatGPT di OpenAI. Il quesito di ricerca indaga come un utilizzo sempre più diffuso e capillare di tali strumenti digitali possa influenzare il nostro rapporto con la lingua; in particolare riguarda il modo in cui le abilità linguistiche e le partiche inerenti alla lettura, alla verbalizzazione a all’elaborazione di contenuti testuali potrebbero essere destabilizzate nel momento in cui subentra un impiego quotidiano e sempre più pervasivo di software automatici basati sui modelli linguistici di grandi dimensioni (LLM). Fondamentale è stato decostruire la nozione di Intelligenza Artificiale dal senso comune, nonché quella di lingua: ho infatti inquadrato entrambi i concetti come terminologie ambigue e non-neutre, connotate da iperonimia, ideologie politiche e interessi economici. In questo ambito mi è stato anzitutto utile cercare di ricostruire, seppur brevemente, il percorso storico che ha portato all’attuale paradigma di intelligenza artificiale, così da identificare alcune rilevanti implicazioni sociali legate al tema della computazione e collocare i trasformatori pre-addestrati lungo una serie di grandi trasformazioni nel rapporto umano con la lingua, incentivate già dall’avvento della digitalizzazione. Insieme ad articoli e letture critiche, è stato inoltre utile effettuare interviste con professionisti provenienti da discipline differenti, sia per ottenere una visione multidisciplinare, sia a completamento della mappatura di insight e sotto tematiche, sfruttando così strumenti metodologici tipici del design appresi durante la mia formazione universitaria. Ho potuto evidenziare un quadro critico, in cui l’IA generativa potrebbe essere correlata a fattori di omogeneizzazione della lingua e al rafforzamento di pregiudizi e visioni del mondo stereotipate, attraverso la peculiare epistemologia che costituisce il funzionamento del machine learning. Un uso esteso di tali tecnologie potrebbe inoltre contribuire a un’ipotrofia delle capacità di padronanza linguistica; fenomeno che, come testimoniato da fonti statistiche e dai docenti che ho interrogato, sembra coinvolgere sempre maggiormente le generazioni più giovani. Affronto la necessità di ripensare la nostra attuale nozione di lingua, il cui cambiamento potrebbe destabilizzare le autorità linguistiche tradizionali e i processi educativi. Ho in seguito esplorato alcuni tra i possibili effetti benefici che possono riguardare, ad esempio, l’apprendimento linguistico, sottolineando che un aspetto determinante circa la positività o meno di tali tecnologie sembra essere il possedimento di consapevolezza critica, insieme alle modalità e alle ragioni di chi ne fa uso.
Estimating the impact of generative Artificial Intelligence on natural language
Pela', Tancredi Ario
2023/2024
Abstract
The aim of this research is to generate an estimate of the impact that Generative Artificial Intelligence could have on natural language. Specifically, by the expression Generative Artificial Intelligence - commonly abbreviated as GenAI – I am referring to the field of NLP digital tools and the pre-trained generative transformers available in the form of chat boxes, such as OpenAI's widely known ChatGPT. The research question investigates how an increasingly widespread and capillary use of such technology may influence our relationship with language; in particular, it concerns the way in which linguistic abilities and the particles inherent to reading, verbalisation and the processing of textual contents may be destabilised when a daily use of automatic software based on large-scale language models (LLM) takes over. Fundamental to this was deconstructing the notion of Artificial Intelligence from common sense, as well as that of language: indeed, I framed both concepts as ambiguous and non-neutral terminologies, connoted by hyperonymy, political ideologies and economic interests. In this context, it was first of all useful for me to try to reconstruct, albeit briefly, the historical path that led to the current paradigm of artificial intelligence, so as to identify some relevant social implications related to the topic of computation and to place pre-trained transformers along a series of major transformations in the human relationship with language, stimulated already by the advent of digitisation. It was also useful to carry out interviews with professionals from different disciplines, both to obtain a multidisciplinary view and to complete the mapping of insights and subtopics, thus exploiting methodological tools typical of Design. I was able to highlight a critical framework, in which generative AI could be related to factors of language homogenisation and the reinforcement of prejudices and stereotypical worldviews, through the peculiar epistemology that constitutes the functioning of machine learning. Extensive use of such technologies could also contribute to a hypotrophy of linguistic mastery skills; a phenomenon that, as witnessed by statistical sources and by the teachers I interviewed, seems to increasingly involve the younger generations. I address the need to rethink our current notion of language, whose change might destabilise traditional linguistic authorities and education processes. I then explored some of the possible beneficial effects that may concern, for example, language learning, emphasising that a decisive aspect regarding the positivity or otherwise of such technologies seems to be the possession of critical awareness, together with the modalities and reasons of those who use them.File | Dimensione | Formato | |
---|---|---|---|
2024_06_Pelà.pdf
accessibile in internet per tutti
Descrizione: Estimating the Impact of Generative Artificial Intelligence on Natural Language
Dimensione
3.91 MB
Formato
Adobe PDF
|
3.91 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/223912