It is now clear how fast society is changing due to the exponential technological development of recent years. We need only to think of the number of technological innovations that have emerged in the last 30 years with the potential to radically change the way we are used to living, working, and interacting with others. If we want to look for the origin of this for sure, and it is now almost taken for granted, we would come across the birth of the Internet, based on which mobile internet and phones, social networks, and cloud computing were born, to name just a few of the innovations that have most changed everyday life. Nowadays, however, we are likely witnessing a new radical moment in human history, comparable to the birth and the growth of the Internet. Artificial intelligence (AI) represents one of the latest evolving technologies with enormous potential to revolutionize and deeply transform the foundations of human society. What is truly groundbreaking, however, is how this technology is recently going through a process of democratization and widespread dissemination. As much as a field of study long known by engineers in the field, advances with generative AI and large language models like GPT-3 and GPT-4, have made AI widely accessible. This significant development has ignited the democratization process one of the most powerful technological innovations in recent years, putting AI abilities directly into the hands of the public. As a result, decision-makers in all sectors must now consider how best to integrate AI-based tools to reimagine and optimize processes to keep pace with today’s rapidly shifting landscape and not lose competitive advantage to other players in the market. We can indeed assume that this technology, with its ease of accessibility and implementation through Natural Language Processing-based tools, has the potential to revolutionize the ways decision-making dynamics take place, especially during the initial, and more strategic, phases of projects. The whole thesis work presented here aims to challenge, and potentially validate this assumption. The work has been structured into three sections: the theoretical background chapter, the research chapter, and lastly the results chapter. The first one gathers the findings of a comprehensive literature review on the subject, aiming to give an overview on design thinking, with a focus on the early phases of it, generative AI and LLMs, and to contextualize the opportunities and implications of implementing them at different stages of the design process. The second unpacks the experimentation process and methodology followed to challenge and verify the initial assumption. LLMs have been subjected to several tests aimed at testing their creative capabilities and to what extent they could either substitute humans or team with humans, augmenting their cognitive efforts during creative processes. Finally, during the third section, the results of the experimentation phase are analyzed, and the opportunities, challenges, and guidelines for the integration of LLMs in the design process are outlined.
È ormai evidente la velocità con cui la società sta cambiando grazie all'esponenziale sviluppo tecnologico degli ultimi anni. Basti pensare al numero di innovazioni tecnologiche emerse negli ultimi 30 anni, potenzialmente in grado di cambiare radicalmente il modo in cui siamo abituati a vivere, lavorare e interagire con gli altri. Se volessimo cercarne con certezza l'origine, ormai quasi scontata, ci imbatteremmo nella nascita di Internet, sulla base della quale sono nati internet e i telefoni cellulari, i social network e il cloud computing, per citare solo alcune delle innovazioni che più hanno cambiato la vita quotidiana. Oggi, tuttavia, stiamo probabilmente assistendo a un nuovo momento radicale nella storia dell'umanità, paragonabile alla nascita e allo sviluppo di Internet. L'intelligenza artificiale (AI) rappresenta una delle ultime tecnologie in evoluzione con un enorme potenziale per rivoluzionare e trasformare profondamente le basi della società umana. Ciò che è veramente innovativo, tuttavia, è il modo in cui questa tecnologia sta recentemente attraversando un processo di democratizzazione e di diffusione capillare. Per quanto si tratti di un campo di studio conosciuto da tempo dai ricercatori del settore, i progressi dell'AI generativa e dei large language models (LLMs), come GPT-3 e GPT-4, hanno reso l'AI accessibile su larga scala. Questo sviluppo significativo ha innescato il processo di democratizzazione di una delle innovazioni tecnologiche più potenti degli ultimi anni, mettendo le capacità dell'AI direttamente nelle mani del pubblico. Di conseguenza, i decision-maker in tutti i settori devono ora valutare come integrare al meglio gli strumenti basati sull'AI per ripensare e ottimizzare i processi, al fine di tenere il passo con il panorama odierno in rapida evoluzione e non perdere il vantaggio competitivo nei confronti di altri attori sul mercato. Possiamo infatti ritenere che questa tecnologia, con la sua facilità di accesso e di implementazione attraverso strumenti basati sull'elaborazione del linguaggio naturale, abbia il potenziale per rivoluzionare il modo in cui avvengono le dinamiche decisionali, soprattutto durante le fasi iniziali e più strategiche dei progetti. L'intero lavoro di tesi qui presentato mira a testare e potenzialmente a convalidare questa ipotesi. Il lavoro è stato strutturato in tre sezioni: un capitolo sul background teorico, un capitolo sulla ricerca sperimentale e infine un capitolo riportante i risultati. Il primo raccoglie i risultati di un'ampia revisione della letteratura scientifica sull'argomento, con l'obiettivo di fornire una panoramica sul Design Thinking, con particolare attenzione alle sue prime fasi, sull'AI generativa e sui LLMs, e di contestualizzare le opportunità e le implicazioni della loro implementazione nelle diverse fasi del processo di progettazione. I LLM sono stati sottoposti a diversi test volti a verificare le loro capacità creative e in che misura potessero sostituire gli esseri umani o collaborare con loro, supportando i loro sforzi cognitivi durante i processi creativi. Infine, nella terza sezione si analizzano i risultati della fase di sperimentazione e si delineano le opportunità, le sfide e le linee guida per l'integrazione dei LLM nel processo di progettazione.
LLMs facing up to human creativity: investigating the creative capabilities of large language models and their role in design thinking
Martini, Giacomo
2022/2023
Abstract
It is now clear how fast society is changing due to the exponential technological development of recent years. We need only to think of the number of technological innovations that have emerged in the last 30 years with the potential to radically change the way we are used to living, working, and interacting with others. If we want to look for the origin of this for sure, and it is now almost taken for granted, we would come across the birth of the Internet, based on which mobile internet and phones, social networks, and cloud computing were born, to name just a few of the innovations that have most changed everyday life. Nowadays, however, we are likely witnessing a new radical moment in human history, comparable to the birth and the growth of the Internet. Artificial intelligence (AI) represents one of the latest evolving technologies with enormous potential to revolutionize and deeply transform the foundations of human society. What is truly groundbreaking, however, is how this technology is recently going through a process of democratization and widespread dissemination. As much as a field of study long known by engineers in the field, advances with generative AI and large language models like GPT-3 and GPT-4, have made AI widely accessible. This significant development has ignited the democratization process one of the most powerful technological innovations in recent years, putting AI abilities directly into the hands of the public. As a result, decision-makers in all sectors must now consider how best to integrate AI-based tools to reimagine and optimize processes to keep pace with today’s rapidly shifting landscape and not lose competitive advantage to other players in the market. We can indeed assume that this technology, with its ease of accessibility and implementation through Natural Language Processing-based tools, has the potential to revolutionize the ways decision-making dynamics take place, especially during the initial, and more strategic, phases of projects. The whole thesis work presented here aims to challenge, and potentially validate this assumption. The work has been structured into three sections: the theoretical background chapter, the research chapter, and lastly the results chapter. The first one gathers the findings of a comprehensive literature review on the subject, aiming to give an overview on design thinking, with a focus on the early phases of it, generative AI and LLMs, and to contextualize the opportunities and implications of implementing them at different stages of the design process. The second unpacks the experimentation process and methodology followed to challenge and verify the initial assumption. LLMs have been subjected to several tests aimed at testing their creative capabilities and to what extent they could either substitute humans or team with humans, augmenting their cognitive efforts during creative processes. Finally, during the third section, the results of the experimentation phase are analyzed, and the opportunities, challenges, and guidelines for the integration of LLMs in the design process are outlined.File | Dimensione | Formato | |
---|---|---|---|
LLMs_Facing_Up_To_Human_Creativity-Giacomo_Martini.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
14.21 MB
Formato
Adobe PDF
|
14.21 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/10589/215157